![\"AINOW](\"/api/uploads/2026/10/ed2b4cf2306a586b346738c4652d4750.webp\")
North Macedonia-based AINOW Society has released version 0.95 of its free AI Literacy Platform, introducing a new development-focused learning environment alongside improvements to its existing Education and Business applications.
\nThe update expands the platform beyond AI fundamentals and professional use into the technical aspects of building AI-powered applications.
\n
The new environment, called AINOW DEV, introduces learning materials covering AI skills, Model Context Protocol (MCP) connectors and practical workflows. It is designed for students, developers and professionals who want to understand how AI systems interact with external tools and services.
\nAINOW Society, which also publishes AI Balkans, developed the platform as an open-source educational resource available in Macedonian, Albanian and English.
\nThe AINOW AI Literacy Platform originally focused on helping teachers and educational institutions understand artificial intelligence and use it responsibly.
\nIts Education environment currently includes 19 learning chapters, 1,106 educational prompts, 57 tools and a glossary of 60 AI-related terms. Teachers can use these resources to prepare lessons, develop classroom activities and create assessments.
\n![\"AINOW](\"/api/uploads/2026/10/1ac4ae88d8f6ae92342fbaa1aba65d06.webp\")
The platform expanded into professional applications with version 0.94, which introduced AINOW Business.
\nThat environment now provides 3,270 prompts across 33 professional fields, 2,190 AI agent definitions and 453 practical scenarios. Its resources cover areas including management, administration, finance, communication and software development.
\n![\"AINOW](\"/api/uploads/2026/10/c568c09589ed51fa982055eb4027aec8.webp\")
With version 0.95, users can access a third environment focused on understanding and developing AI applications.
\nThe new DEV environment includes a 13-chapter learning guide, seven AI tool profiles and three reference libraries containing:
\nMCP, or Model Context Protocol, is a technical standard that allows AI applications to communicate with external tools and data sources.
\n![\"AINOW](\"/api/uploads/2026/10/28f37df2ad0143eb22c66521ee93014e.webp\")
As AI assistants become capable of performing tasks beyond generating text, understanding these connections is increasingly relevant to developers and organisations exploring automation.
\nDEV also provides practical exercises, tools for examining original AI skill files and a calculator for estimating token and context requirements.
\nThe reference libraries are intended for education and exploration. They do not represent automatically installed connectors or immediately executable integrations.
\nVersion 0.95 also introduces a unified interface across Education, Business and DEV.
\nThe update includes improved navigation, more consistent language controls, updated prompt libraries and better reading options on mobile devices.
\nUsers can now work with longer prompts through improved reading views, while mobile users benefit from a full-screen interface that makes reviewing and editing content easier.
\nThe release also improves local note-taking, settings and backup management. Business and DEV now provide separate local workspaces for saving notes and preferences.
\nTechnical changes address problems involving multilingual search, browser navigation and data synchronisation. The project's published release notes also report fixes for three previously identified GitHub CodeQL findings.
\nThe downloadable version passed automated verification checks, although additional physical-device and screen-reader testing was still pending at publication.
\nOne of the platform's distinguishing features is its approach to accessibility.
\nAll three learning environments are free to use and do not require an AINOW account. Users can access the online version or download the application for local use without installing a separate server.
\nDownloaded educational materials can also be used offline, although external AI services may still require internet access and separate accounts.
\nFor users in North Macedonia and the wider Western Balkans, the availability of learning materials in Macedonian and Albanian provides an alternative to relying entirely on English-language technical resources.
\nThe platform also offers practical material for schools, smaller organisations and individuals who may not have the budget for commercial training platforms.
\nIts open-source model allows developers and institutions to inspect the application code, adapt it to their needs and contribute improvements, subject to the project's licensing requirements.
\nThe new DEV environment extends AINOW's original educational focus toward technical AI literacy, giving learners a way to explore how AI-powered applications are assembled, connected and maintained.
\nThe platform is available at edu.ainow.mk. Further information about the update is available in the official AINOW Society announcement, while the downloadable application and technical release notes can be found on GitHub.
","featured_image_id":425,"intro_image_id":425,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":7,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":4,"ordering":0,"published_at":"2026-10-26 22:51:57","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-26 22:51:57","updated_at":"2026-10-26 22:51:57","key_takeaways":"[\"AINOW Society has released version 0.95 of its free AI Literacy Platform, introducing DEV as a third learning environment alongside Education and Business.\",\"AINOW DEV includes 13 learning chapters, 1,000 AI skills, 1,000 MCP references and 200 workflows designed to help users understand AI development and integrations.\",\"The platform is open source, supports Macedonian, Albanian and English, and offers offline access to downloaded learning resources without mandatory registration.\"]","faq":"[{\"q\":\"What is new in AINOW AI Literacy v0.95?\",\"a\":\"The update introduces AINOW DEV, a learning environment focused on AI development, MCP connectors, reusable skills and workflows. It also improves navigation, prompt reading, mobile usability and security across the platform.\"},{\"q\":\"Who can use the AINOW AI Literacy Platform?\",\"a\":\"The platform serves educators, students, professionals, organisations and developers through three dedicated environments: Education, Business and DEV.\"},{\"q\":\"Is AINOW AI Literacy free and available offline?\",\"a\":\"Yes. The platform is free, open source and accessible without registration. Downloaded learning materials can be used offline, although external AI services may require internet access.\"}]","sources":"[{\"title\":\"AINOW AI Literacy Platform v0.95 Introduces DEV\",\"url\":\"https://ainow.mk/blog/ainow-ai-literacy-platform-v0-95-introduces-dev\"}]","category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Aleksandra Todorovikj","author_slug":"aleksandra-todorovikj","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/ffc8dd720f66d531239cbc548799965c.webp","featured_image_alt":"Ainow literacy","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/ffc8dd720f66d531239cbc548799965c.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"AINOW v0.95 Adds DEV to Free AI Learning Platform","meta_description":"AINOW v0.95 launches DEV with 1,000 AI skills, 1,000 MCP references and 200 workflows. Free AI education in three languages.","robots":"index","canonical_url":"","og_title":"","og_description":""},"breaking":[{"id":41,"title":"OpenAI’s GPT-5.6 Sol Arrives","slug":"openai-s-gpt-5-6-sol-arrives","short_description":"","content":"OpenAI has unveiled GPT-5.6 Sol, its newest flagship artificial intelligence model, alongside two additional models—Terra and Luna—in a limited preview for selected developers and organizations. The new model family represents OpenAI's latest step toward AI systems capable of handling complex, long-running tasks with greater accuracy, efficiency, and reliability.
\nRather than releasing a single model designed for every use case, OpenAI is introducing a three-tier lineup tailored to different workloads. The company says the approach gives developers and businesses greater flexibility by allowing them to choose the right balance between capability, speed, and cost.
\nThe new GPT-5.6 family consists of:
\nGPT-5.6 Sol, OpenAI's most capable model, built for advanced reasoning, software engineering, cybersecurity, scientific research, and complex agentic workflows.
\nGPT-5.6 Terra, designed to deliver strong performance while reducing operational costs for enterprise applications.
\nGPT-5.6 Luna, the fastest and most affordable model in the lineup, optimized for large-scale deployments where speed and efficiency are the primary priorities.
\nAccording to OpenAI, the three-model strategy reflects the growing diversity of AI workloads. Instead of relying on a single model for every task, organizations can now select a model that better matches their technical and financial requirements.
\nOne of the biggest themes of GPT-5.6 is the continued evolution of agentic AI.
\nUnlike traditional AI assistants that respond to individual prompts, agentic systems are designed to complete extended workflows involving planning, coding, research, tool usage, and multi-step reasoning while maintaining context over longer periods.
\nOpenAI says GPT-5.6 Sol demonstrates significant improvements in software development, cybersecurity analysis, scientific reasoning, and autonomous task execution, making it suitable for developers and organizations building increasingly sophisticated AI-powered applications.
\nOpenAI notes that GPT-5.6 Sol underwent one of the company's largest safety evaluation programs before entering preview.
\nThe model was tested using more than 700,000 A100 GPU-equivalent hours of automated evaluations, combined with weeks of expert red-teaming focused on identifying potential security and misuse risks.
\nThe company says these evaluations contributed to stronger cybersecurity capabilities while improving safeguards intended to reduce harmful or unsafe behavior.
\nUnlike previous GPT releases, GPT-5.6 is not yet broadly available.
\nOpenAI has launched the model through a limited preview, giving access to selected developers, researchers, and enterprise partners while gathering feedback before a wider rollout.
\nThe preview also allows the company to continue evaluating performance and safety as increasingly capable frontier AI models move toward public deployment.
\nOpenAI has also published API pricing for the new models.
\nThe pricing structure reinforces OpenAI's strategy of offering multiple models for different performance and budget requirements.
\nThe introduction of GPT-5.6 reflects a broader trend across the artificial intelligence industry. Leading AI companies are no longer competing solely to build the most capable chatbot—they are developing specialized models optimized for different types of work.
\nFor software teams, researchers, and enterprises, this means selecting an AI model based not only on intelligence but also on latency, operating cost, and workload requirements.
\nAs organizations increasingly integrate AI into business operations, this model specialization could improve efficiency while making advanced AI more accessible across industries.
\nThe launch of GPT-5.6 highlights how rapidly artificial intelligence is moving beyond simple conversational interfaces. The next generation of AI competition is increasingly centered on digital workers capable of planning projects, writing software, analyzing complex datasets, conducting research, and collaborating with humans over extended periods.
\nFor the Western Balkans, this evolution presents both opportunities and challenges. Technology companies, startups, universities, and public institutions throughout the region are beginning to adopt AI for practical business and research applications. Access to specialized models such as Sol, Terra, and Luna could help organizations choose solutions that better fit their needs instead of relying on a single high-end model for every task.
\nHowever, the announcement also underscores another emerging reality: access to the most advanced AI systems is becoming more gradual and carefully managed. Limited previews, extensive safety testing, and staged deployments are likely to become standard practice as frontier AI models grow more powerful.
\nFor businesses across the Balkans, the message is becoming increasingly clear. Success with AI will depend less on simply using the newest model and more on understanding which model best supports a specific workflow, team, or industry.
\nNorth Macedonia-based AINOW Society has released version 0.94 of its AI literacy platform, adding a new professional AI library alongside its existing education product and expanding its resources for universities.
\nThe release reorganises edu.ainow.mk around two applications: AINOW Education for schools and higher education, and the new AINOW Business platform for professionals and organisations.
\nBoth remain free to use, require no user account and have been designed to operate without a permanent internet connection.
\nThe move broadens a project that was originally focused primarily on helping teachers understand and use artificial intelligence. With v0.94, AINOW is extending the same model to professional work while keeping the software open source and available in Macedonian, Albanian and English.
\nAINOW Education now operates under /edu/ and retains the learning material and tools developed for teachers and school administrators.
The platform includes a 19-chapter AI literacy guide covering subjects from basic AI concepts and prompting to retrieval, performance, responsible use and safety.
\nIts education library now contains 1,106 prompts built around work carried out in schools and universities, including lesson preparation, communication with parents, individual education plans, administrative documentation and institutional evaluation.
\nVersion 0.94 makes Higher Education a dedicated category alongside Teachers and Administration.
\nThe new section provides material for lecturers, researchers, doctoral supervision, student services, quality assurance and university leadership, with 60 prompts per supported language across the six areas.
\nThe education platform also includes a 60-term AI glossary and a directory of 57 AI tools.
\nThe expansion reflects a wider problem across the region: much of the practical material available for using generative AI in education is still developed primarily in English and for larger education systems. AINOW's decision to maintain the platform in Macedonian and Albanian gives local institutions material that can be used without translating an English-language resource first.
\n![\"\"](\"/api/uploads/2026/08/e3e154ec6f8507994a78089916cce54f.webp\")
The biggest addition to v0.94 is AINOW Business, a separate application built around professional AI use.
\nIt launches with 3,270 prompts across 33 fields, including software development, law, medicine, finance, agriculture, journalism, public health, logistics and management.
\nThe platform also contains 2,190 predefined AI agent configurations.
\nThese provide structured roles, objectives and instructions that can be copied into compatible AI systems rather than requiring users to construct an agent configuration themselves.
\nAINOW has added several ways of navigating that material.
\nA Situations section currently provides eight guided scenarios for tasks such as negotiations and incident response. Users describe the circumstances they are dealing with and the platform helps construct a relevant prompt.
\n![\"AINOW](\"/api/uploads/2026/08/459950cd1a6714a1d3274bf9cf88150f.webp\")
\nPlaybooks provide more structured, step-by-step workflows, while an Agent Builder generates customised instructions based on information supplied by the user.
\nThe Business application itself is free. Consulting, training and organisational AI adoption services are offered separately through AINOW Society's technology partner, CTRLAB.
\nOne of the more distinctive aspects of the project is its approach to distribution.
\nBoth applications are designed to run locally without a conventional server installation. A copy of the platform can be stored on a computer or USB drive and opened locally through its index.html file.
That makes the architecture relevant beyond software design.
\nSchools and organisations across the Balkans do not operate with identical infrastructure. Older computers, restricted installation permissions and inconsistent connectivity remain practical barriers in some institutions, particularly outside major urban centres.
\nThe Business application originally used a more conventional React and bundler-based architecture. For the v0.94 release, its modules and dependencies were adapted so the final application could operate from local files.
\nReact remains part of the application, but dependencies needed by the software are stored locally instead of being retrieved from third-party services when the application runs.
\nThe platform's fonts are now hosted locally as well.
\nThe same architecture supports AINOW's privacy model.
\nNeither Education nor Business requires registration. Language preferences, favourites and scratchpad information remain in the user's browser rather than being stored in a central AINOW account.
\nAccording to the organisation, the applications do not use analytics, tracking pixels or external telemetry to monitor how people use the platform.
\nThat design is particularly relevant for the education product, where teachers may be working with information connected to students, but AINOW has applied the same model to its professional platform.
\nIt is a different approach from the increasingly common model of AI software built around mandatory cloud accounts and persistent user profiles.
\nThe release was not limited to new features.
\nAINOW says it also conducted an audit of its platform statistics and multilingual interface after finding inconsistencies that had accumulated during development.
\nThe Business interface had previously advertised 3,370 prompts across 34 areas. An audit of the underlying content found the correct figures to be 3,270 prompts across 33 areas, which are now used throughout the platform.
\nThe team also found obsolete translation entries left behind after features had been removed, as well as around 40 interface placeholders that still appeared in English for Macedonian and Albanian users.
\nFollowing the cleanup, all three languages now use the same 901 active interface strings.
\nAnother discrepancy concerned the education guide, which was listed as having 20 chapters in one part of the application and 19 elsewhere. The verified number is 19.
\nPublishing those corrections alongside the larger feature release is a relatively small detail, but a useful one for an open-source project whose credibility depends partly on the accuracy of the resources it distributes.
\nAINOW Education and AINOW Business are released under the GNU GPLv3 licence, allowing the software to be inspected, modified and redistributed under the licence terms.
\nFor schools in particular, downloadable releases mean the hosted website is not the only way to access the project. A local copy can be distributed between computers without requiring each device to maintain internet access.
\n![\"AINOW](\"/api/uploads/2026/08/7f96bcebd4bd86893586e185f9dc076e.webp\")
\n\n
With v0.94, AINOW is testing whether the model it developed for AI literacy in education can also work for professional adoption.
\nThe project now covers classrooms, universities and workplaces from the same multilingual, offline-capable base.
\nThe more important test comes next: whether organisations and contributors outside AINOW's existing education community begin using and extending it.
\nThe full AINOW Society release announcement is available on the organisation’s official blog
","featured_image_id":242,"intro_image_id":242,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":1,"show_table_of_contents":1,"read_time_minutes":6,"view_count":45,"ordering":0,"published_at":"2026-08-16 16:22:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-16 16:22:19","updated_at":"2026-08-16 16:30:56","key_takeaways":null,"faq":null,"sources":null,"category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/9a8e84f250b054e8c7e59735b92aed7d.webp","featured_image_alt":"AINOW Society webpage offers free Education and Business platforms, including 3,270 business prompts.","featured_image_width":1487,"featured_image_height":696,"intro_image_path":"/api/uploads/2026/08/9a8e84f250b054e8c7e59735b92aed7d.webp","intro_image_width":1487,"intro_image_height":696,"seo_title":"AINOW v0.94 Adds Business AI Platform With 3,270 Prompts","meta_description":"AINOW v0.94 adds a Business platform with 3,270 prompts and 2,190 AI agents while expanding higher education and offline access.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":94,"title":"Europe’s AI Act Now Has Investigators","slug":"europe-s-ai-act-now-has-investigators","short_description":"","content":"For years, Europe’s approach to artificial intelligence was discussed mainly through draft laws, risk categories, voluntary codes and future deadlines.
\nThat phase is ending.
\nFrom August 2, 2026, the European Commission’s AI Office and national authorities begin exercising key enforcement powers under the EU AI Act. Brussels is also strengthening the teams responsible for investigating possible violations, examining powerful general-purpose AI models and responding when AI systems create serious risks.
\nThe AI Act is no longer only a rulebook. It is becoming an enforcement system.
\nThe European Commission’s AI Office was established to coordinate implementation of the AI Act across the European Union.
\nIts most important direct responsibility concerns general-purpose AI models, including the foundation models behind many chatbots, coding assistants, content generators and AI agents.
\nUntil now, most public attention has focused on what companies must disclose and which systems fall into different risk categories. The harder questions begin now.
\nWho decides whether a model presents a systemic risk? What evidence must a company provide? How will regulators test systems that are constantly updated? And what happens when a model behaves differently in real-world use than it did during internal testing?
\nThe AI Office includes teams focused on regulation, compliance and AI safety. Their work will include examining possible infringements, assessing model risks and supporting enforcement across the EU.
\nThe Commission has also appointed an independent scientific panel to assist regulators with model evaluation and systemic-risk assessment.
\nThat matters because enforcing AI rules cannot depend only on lawyers reviewing company documents. Regulators will increasingly need technical experts capable of testing models, examining their capabilities and challenging claims made by developers.
\nThe EU’s attention is not limited to visible deepfakes or chatbots that fail to identify themselves.
\nRegulators are also concerned about broader risks linked to advanced models, including their potential use in cyberattacks, manipulation, illegal content generation and other activities that may affect public safety or fundamental rights.
\nThe European Commission has direct authority over certain general-purpose AI providers, while national authorities remain responsible for many AI systems deployed within individual member states.
\nThis shared structure reflects how modern AI systems operate.
\nA model may be developed by a company in the United States, hosted in another country, integrated into software by a European business and then used by citizens across several jurisdictions.
\nNo single national regulator can easily supervise that entire chain.
\nEurope is therefore building a system in which central institutions examine the most powerful models, while national authorities supervise how AI is used in workplaces, public services, healthcare, education, policing and other sensitive areas.
\nThe AI Office also operates a whistleblower mechanism through which people connected to AI providers can report suspected violations.
\nEmployees, contractors and business partners may have access to internal evaluations, safety failures or company decisions that never become public.
\nExternal researchers and journalists usually see only the final product. Insiders may know whether a company ignored a serious risk, concealed a failed test or released a model before safety concerns were resolved.
\nA confidential reporting channel gives regulators another way to identify problems before they grow into larger incidents.
\nIts success, however, will depend on trust. People must believe that reports will be treated seriously, their identities will be protected and the AI Office has the capacity to investigate the information it receives.
\nPassing a law is easier than enforcing it against some of the world’s largest technology companies.
\nMajor AI providers have strong legal teams, large technical departments and substantial financial resources. Their systems are also complex enough that regulators may struggle to independently reproduce company evaluations or determine whether a safety measure is genuinely effective.
\nEurope must avoid two opposite failures.
\nThe first would be weak enforcement, where companies submit large amounts of documentation but face little technical scrutiny.
\nThe second would be enforcement that becomes so slow and bureaucratic that smaller European developers carry a heavier burden than the global corporations the law was designed to oversee.
\nThe credibility of the AI Act will depend on whether regulators can distinguish between real risk and minor administrative mistakes.
\nA startup that makes an honest reporting error should not be treated in the same way as a powerful provider that hides a serious capability or repeatedly ignores safety obligations.
\nProportionate enforcement will be one of the biggest challenges.
\nWestern Balkan countries are outside the European Union, but they will not remain outside the influence of this enforcement system.
\nRegional companies provide software, outsourcing, data services, technical support and digital products to clients across the EU. Some operate AI tools that can be accessed by European users. Others integrate models developed by global providers into services used by banks, schools, public institutions and businesses.
\nThe question will no longer be only whether an organisation uses AI.
\nCompanies may increasingly need to know which model they use, where data is processed, how outputs are reviewed, what risks were identified and who is responsible when the system fails.
\nThis will matter particularly for organisations working in regulated sectors or participating in European projects.
\nIt will also matter for governments in the region. Candidate countries seeking closer alignment with the EU will face growing pressure to develop compatible laws, institutions and oversight mechanisms.
\nCopying the language of the AI Act will be the easy part.
\nFinding qualified inspectors, technical auditors, legal experts and independent researchers will be much harder.
\nThe Balkans already face shortages of specialised digital expertise in many public institutions. AI enforcement will require even more advanced knowledge, combining law, cybersecurity, data protection and machine-learning evaluation.
\nWithout that capacity, national laws may exist mostly on paper.
\nThe new AI disclosure labels will be the most visible change for ordinary users.
\nPeople may notice a warning under a synthetic video, a notice attached to AI-generated content or a chatbot explaining that it is not human.
\nBut the deeper change will happen away from the screen.
\nEurope is building institutions that can request information, investigate companies, examine advanced models and impose consequences when obligations are ignored.
\nWhether those institutions will be fast enough, technically capable enough and independent enough remains uncertain.
\nThe AI Act has already given Europe one of the world’s most ambitious legal frameworks for artificial intelligence.
\nFrom August 2, the more difficult phase begins: proving that the rules can actually be enforced.
","featured_image_id":126,"intro_image_id":126,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":1,"show_table_of_contents":0,"read_time_minutes":6,"view_count":50,"ordering":0,"published_at":"2026-08-01 11:12:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-01 11:12:23","updated_at":"2026-10-01 18:06:40","key_takeaways":null,"faq":null,"sources":null,"category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/2604f5de8c500cc8eec9eb5ede7801c1.webp","featured_image_alt":"The European Union flag waves against a blue, clouded sky.","featured_image_width":1600,"featured_image_height":936,"intro_image_path":"/api/uploads/2026/07/2604f5de8c500cc8eec9eb5ede7801c1.webp","intro_image_width":1600,"intro_image_height":936,"seo_title":"Europe’s AI Act Now Has Investigators","meta_description":"Europe moves from AI regulation to enforcement as the EU AI Office gains powers to investigate companies, advanced models and systemic risks.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":53,"title":"AI Cyber Threats in the Western Balkans","slug":"ai-cyber-threats-in-the-western-balkans","short_description":"","content":"The real story for the Western Balkans is not about Five Eyes capitals, but about how AI is accelerating cyber threats in a region that is already under pressure. As governments, businesses, and media systems digitize, attackers are gaining new ways to automate phishing, deepen disinformation, and target public institutions with greater speed and precision.
\nThis matters because the region’s cyber resilience is still uneven. Recent regional and EU reporting points to persistent weaknesses in incident response, coordination, staffing, and protection of digital infrastructure, even as support programs and policy attention increase.
\nThe Western Balkans face a combination of high digital dependence and fragile trust ecosystems. Cybercrime reporting, evidence handling, and private-sector support remain priorities, which shows that the region is still working to strengthen the basics of cyber defense.
\nAt the same time, disinformation and hybrid interference continue to shape the digital information space. Regional analysis of elections in the Western Balkans warns that weak regulation, media capture, low media literacy, and bot networks create fertile ground for manipulation. AI makes those tactics cheaper, faster, and more believable.
\nAI does not create every cyber problem in the Balkans, but it makes existing ones harder to contain. It can help criminals write convincing multilingual phishing messages, generate fake identities, clone voices, and produce synthetic images or videos that look authentic enough to mislead citizens and institutions.
\nThat is especially dangerous during elections, crises, and public controversies. In a region where trust in institutions is already limited, a single believable fake can spread quickly and distort public debate before fact-checkers or officials can respond.
\nThe public sector is a prime target because governments hold valuable data and run essential services. The EU has already expanded cybersecurity support for Albania, Montenegro, and North Macedonia, while also funding a broader regional program for the Western Balkans to improve prevention, preparedness, and response.
\nThat support is important, but it also shows the scale of the challenge. Digital public services are expanding faster than institutional capacity in many places, and that gap creates opportunities for AI-assisted attacks on government networks, identity systems, and critical service provision. In practical terms, this means ministries, municipalities, hospitals, and schools need stronger defenses now, not later.
\nOne of the most serious risks is AI-powered manipulation of political communication. Regional research warns that the Western Balkans remain highly exposed to disinformation and foreign or domestic influence operations that undermine democratic trust. AI can amplify those operations by producing large amounts of content that looks local, emotional, and credible.
\nThis is not only a technology issue. It is a democratic resilience issue. When citizens cannot easily distinguish real from fake, public debate becomes noisier, institutional credibility weakens, and polarization deepens.
\nThe response should start with basic cyber hygiene: multifactor authentication, patching, backup discipline, employee training, and incident response planning. Regional reporting has made clear that cybercrime response still needs better reporting, evidence handling, and technical support for the private sector.
\nSecond, the region needs faster coordination. Governments, CERT teams, media organizations, and civil society groups should have shared procedures for deepfakes, impersonation attacks, and crisis communication. Third, AI and cyber literacy should become part of public education, journalism training, and civil-service development, because technical defenses alone will not stop social manipulation.
\nThe Western Balkans do not need to copy larger powers to respond well. They need to build a regional model based on shared vulnerabilities, practical cooperation, and realistic investment in resilience. The same digital transformation that creates risk can also strengthen institutions if it is paired with security, transparency, and public trust.
\nThat is the key message for AI Balkans: the threat is local, the impact is local, and the solution must be local too. The region should treat cyber-capable AI as a now-problem, not a future one.
\n","featured_image_id":110,"intro_image_id":110,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":18,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":1,"show_table_of_contents":1,"read_time_minutes":4,"view_count":31,"ordering":0,"published_at":"2026-06-29 11:12:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-29 11:06:03","updated_at":"2026-06-29 11:12:15","key_takeaways":null,"faq":null,"sources":null,"category_name":"Cybersecurity","category_slug":"cybersecurity","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/450f25cf08495900e6d8275f13db2a88.webp","featured_image_alt":"Cyber threats photo illustration","featured_image_width":739,"featured_image_height":415,"intro_image_path":"/api/uploads/2026/06/450f25cf08495900e6d8275f13db2a88.webp","intro_image_width":739,"intro_image_height":415,"seo_title":"Cyber-Capable AI Models Are Changing Balkan Security","meta_description":"AI-powered cyber threats are rising fast, and the Western Balkans must strengthen resilience, digital trust, and cybersecurity now.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":37,"title":"AI Regulation Becomes the Balkans’ Next EU Test","slug":"ai-regulation-becomes-the-balkans-next-eu-test","short_description":"","content":"Artificial intelligence is quickly becoming a new test of European integration for the Western Balkans. As the European Union moves ahead with the AI Act and a broader AI policy framework, countries in the region are being pushed to think not only about innovation, but also about regulation, responsibility and trust.
\nThis is a major shift. For years, digital transformation in the Balkans was mostly discussed through e-government, broadband access, startup development and public-sector modernization. Now, AI is forcing governments to answer deeper questions: Who is responsible when an algorithm makes a harmful recommendation? How should high-risk AI systems be supervised? Can public institutions use AI without weakening citizens’ rights? How can smaller economies regulate technology without blocking innovation?
\nThe first structured regional dialogue on artificial intelligence among the Western Balkans Six shows that the region understands the need for coordination. AI policy leads from the six economies met in Podgorica to discuss responsible and sustainable AI adoption, shared priorities and common challenges. This matters because no Balkan country can solve AI governance alone.
\nEU alignment will be one of the strongest drivers of AI regulation in the region. Countries that want closer integration with the EU will need to adapt their legal systems to European standards. That includes risk-based regulation, transparency obligations, human oversight, data protection and safeguards for high-risk AI systems.
\nFor businesses, this creates both pressure and opportunity. Companies building AI tools for European clients will increasingly need to understand compliance. Those that prepare early may gain a competitive advantage. Those that ignore regulation may find it harder to sell into EU markets or work with regulated industries such as healthcare, finance, education and public administration.
\nFor governments, the challenge is more complex. Public institutions want to use AI to improve services, reduce administrative delays and analyze data more efficiently. But public-sector AI is sensitive because it can affect access to benefits, permits, healthcare, education and other essential services. In these areas, mistakes are not only technical failures; they can become rights violations.
\nThis is why AI regulation in the Balkans cannot simply copy European documents without building local capacity. Laws are necessary, but they are not enough. Countries need regulators who understand AI systems, public officials who can manage digital projects responsibly, and courts or oversight bodies that can evaluate algorithmic harm.
\nThe region also needs regulatory sandboxes. These controlled testing environments can help startups, public institutions and researchers experiment with AI while staying within clear safety rules. For smaller markets, sandboxes may be especially useful because they allow innovation without immediately imposing heavy compliance burdens on early-stage companies.
\nAnother important issue is public trust. Many Balkan citizens already have low confidence in institutions. If AI is introduced into public services without transparency, it could increase suspicion rather than improve efficiency. Citizens need to know when AI is being used, what data it relies on and how decisions can be challenged.
\nThe Balkans also face a risk of regulatory fragmentation. If each country develops different AI rules, companies and institutions may struggle to scale across the region. A regional AI policy roadmap could help create more consistency, especially in areas such as public procurement, education, data governance and innovation support.
\nAt the same time, regulation must not become a barrier that only large companies can afford. If compliance costs are too high, local startups and small businesses may avoid AI development or depend entirely on foreign platforms. The region needs a balanced approach: strong enough to protect citizens, but flexible enough to support innovation.
\nAI regulation is therefore becoming more than a legal topic. It is becoming part of the region’s economic strategy, EU integration path and democratic development. Countries that manage it well will be better positioned to attract investment, protect citizens and build trustworthy digital services.
\nThe next stage for the Balkans will be practical implementation. The region does not only need AI strategies and conferences. It needs working institutions, clear rules, trained regulators and responsible adoption by companies and governments.
\nIf the Western Balkans can align with European AI standards while preserving space for local innovation, the region could turn regulation into an advantage. Instead of being late to AI, it could become a region known for practical, trustworthy and EU-compatible AI adoption.
","featured_image_id":75,"intro_image_id":75,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":3,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":1,"show_table_of_contents":1,"read_time_minutes":4,"view_count":15,"ordering":0,"published_at":"2026-06-25 18:15:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 18:15:47","updated_at":"2026-06-25 18:34:06","key_takeaways":null,"faq":null,"sources":null,"category_name":"Insights","category_slug":"insights","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/e357e22a7b081def9e3675d7580849e5.webp","featured_image_alt":"AI text centered in a ring of EU stars between two blue circuit-patterned human profiles.","featured_image_width":1600,"featured_image_height":995,"intro_image_path":"/api/uploads/2026/06/e357e22a7b081def9e3675d7580849e5.webp","intro_image_width":1600,"intro_image_height":995,"seo_title":"AI Regulation Becomes the Balkans’ Next EU Test | AI Balkans","meta_description":"Artificial intelligence is quickly becoming a new test of European integration for the Western Balkans.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":32,"title":"North Macedonia Joins Europe’s AI Supercomputing Network","slug":"north-macedonia-joins-europe-s-ai-supercomputing-network","short_description":"","content":"North Macedonia is taking a major step toward joining Europe’s artificial intelligence infrastructure network through VEZILKA, its first national artificial intelligence center. The initiative is designed to connect the country with European AI Factories and provide access to advanced computing resources that are essential for modern AI development.
\nThe project is especially important because artificial intelligence is no longer driven only by software talent. To build powerful AI systems, countries need access to large datasets, specialized expertise and high-performance computing infrastructure. For smaller economies, building all of this independently is expensive and difficult. VEZILKA gives North Macedonia a way to connect with a wider European network instead of developing in isolation.
\nThrough the European Union’s EuroHPC AI Factory Antennas initiative, VEZILKA is expected to act as a bridge between North Macedonia and AI-optimized supercomputing resources across Europe. The center is also connected strategically with the PHAROS AI Factory in Greece, creating a regional link that could help Macedonian researchers, startups and institutions access stronger technical capacity.
\nThis connection matters because high-performance computing is one of the foundations of serious AI work. Training and testing advanced models requires far more computing power than ordinary computers can provide. Without access to such infrastructure, local AI teams are often limited to smaller experiments or forced to depend entirely on foreign commercial platforms.
\nFor North Macedonia’s universities, VEZILKA could create new opportunities for scientific research. Researchers working on language technologies, healthcare, engineering, energy, agriculture or public policy could use better AI infrastructure to develop more ambitious projects. This could also help the country participate more actively in European research programs.
\nFor startups, the center could reduce one of the biggest barriers to innovation: the cost of computing power. Many young technology companies have strong ideas but cannot afford the infrastructure needed to train or test AI models. If VEZILKA provides practical access, local startups may be able to move faster from concept to prototype and from prototype to market.
\nThe project also has clear value for the public sector. Artificial intelligence could help public institutions improve services, analyze data, detect inefficiencies and support better decision-making. However, this requires secure infrastructure and responsible governance. By working within a European AI framework, North Macedonia can benefit from technical support while also following standards for trustworthy and ethical AI.
\nThe regional connection with Greece is also significant. Instead of treating AI development as a purely national effort, VEZILKA places North Macedonia inside a cross-border innovation ecosystem. This could encourage cooperation between universities, companies and institutions in the wider region, especially in areas where countries face similar challenges.
\nA key part of the project will be skills development. Infrastructure alone is not enough if there are not enough engineers, researchers, public officials and business leaders who understand how to use it. VEZILKA will need to support training, education and practical cooperation so that AI becomes useful beyond a small group of specialists.
\nThe center could also help North Macedonia avoid dependence on imported AI systems that do not understand local needs. By building domestic capacity and connecting it with European infrastructure, the country can develop solutions that are more suitable for Macedonian language, institutions, companies and citizens.
\nStill, the success of the project will depend on implementation. VEZILKA must become accessible to real users, not only exist as a formal institution. Startups, researchers and public bodies will need clear procedures, support programs and practical guidance on how to use the resources available through the center.
\nNorth Macedonia’s connection to Europe’s AI supercomputing network shows that the country is trying to position itself inside the next phase of digital transformation. If VEZILKA becomes an active and useful platform, it could help the country move from AI ambition to real AI development.
\nIn the coming years, the impact of the project will be measured by results: new research, better public services, stronger startups, Macedonian-language AI tools and practical applications that citizens and businesses can feel. The launch of VEZILKA is only the beginning; the real test will be how effectively North Macedonia uses this European connection to build its own AI future.
","featured_image_id":69,"intro_image_id":69,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":1,"show_table_of_contents":1,"read_time_minutes":4,"view_count":19,"ordering":0,"published_at":"2026-06-25 12:30:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 12:30:55","updated_at":"2026-06-25 18:35:15","key_takeaways":null,"faq":null,"sources":null,"category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/b0a3a11ee448f39c7f3c996c4a36bcfa.webp","featured_image_alt":"Rows of AI supercomputers fill a brightly lit data center.","featured_image_width":1024,"featured_image_height":682,"intro_image_path":"/api/uploads/2026/06/b0a3a11ee448f39c7f3c996c4a36bcfa.webp","intro_image_width":1024,"intro_image_height":682,"seo_title":"North Macedonia Joins Europe’s AI Supercomputing Network","meta_description":"North Macedonia is taking a major step toward joining Europe's artificial intelligence infrastructure network through VEZILKA.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":30,"title":"Albania Plans AI-Powered Healthcare by 2026","slug":"albania-plans-ai-powered-healthcare-by-2026","short_description":"","content":"Albania is preparing to introduce artificial intelligence into its national healthcare system in 2026, marking one of the country’s most important steps toward digital transformation in public health. The announcement was made by Prime Minister Edi Rama, who said that the 2026 healthcare budget will prioritize both artificial intelligence and infrastructure investments.
\nThe plan comes as Albania continues to modernize hospitals, emergency centers and medical services across the country. According to the government’s direction, artificial intelligence will not be treated as a separate technology experiment, but as part of a wider reform of the healthcare system. The goal is to improve diagnostics, connect health institutions digitally and raise the quality of care closer to European standards.
\nOne of the main areas where AI is expected to be used is medical imaging. Artificial intelligence can support doctors by analyzing scans, detecting patterns and identifying possible health risks faster than traditional manual processes alone. This does not mean that AI will replace doctors, but it can become a powerful support tool for specialists, especially in cases where fast and accurate interpretation is important.
\nAnother planned area is predictive analysis. With better digital systems, Albania could use health data to identify trends, anticipate pressure on hospitals, improve planning and support preventive medicine. Predictive tools can help health institutions understand where demand may increase, which services need more resources and how patients can be treated earlier.
\nThe government has also linked the AI healthcare plan with interoperability. This means that different parts of the healthcare system should be able to communicate with each other digitally. Hospitals, clinics, laboratories and central health institutions often work with separate systems, which can slow down care and create administrative problems. A national digital ecosystem could make patient information easier to manage and improve coordination between institutions.
\nThe project is expected to be developed through cooperation with a European AI healthcare partner and supported with EU funding. This is important because healthcare AI requires high standards for safety, privacy and reliability. Albania will need to ensure that any AI tools introduced into hospitals are tested, supervised and compatible with European health and data protection principles.
\nThe announcement also fits into Albania’s broader goal of improving healthcare infrastructure by 2030. Rama has mentioned hospital modernization projects, including the new Korçë Hospital, the Durrës Emergency Center and municipal hospitals in Mati and Tropojë. These physical investments are now being connected with digital modernization, showing that the government wants to combine better buildings, better equipment and smarter technology.
\nFor citizens, the promise of AI in healthcare is faster service, better diagnostics and more efficient public health management. For doctors and medical staff, AI could reduce pressure by helping with repetitive tasks, data analysis and decision support. For the state, it could improve planning and reduce inefficiencies in the healthcare system.
\nHowever, the reform will also bring serious challenges. Health data is highly sensitive, and any AI system used in medicine must be transparent, secure and carefully monitored. Albania will need strong rules on who can access patient data, how algorithms are tested and what happens if an AI-supported decision is wrong.
\nThe success of the project will depend not only on technology, but also on training. Doctors, nurses, administrators and IT professionals will need to understand how these systems work and how to use them responsibly. Without proper training, even advanced technology can fail to deliver real improvements.
\nAlbania’s plan to introduce artificial intelligence into healthcare shows that AI is moving from political discussion into practical public services. If implemented carefully, it could help modernize the country’s health system and bring it closer to European digital standards. The real test will be whether the new tools improve everyday care for patients, not only whether they sound impressive on paper.
","featured_image_id":66,"intro_image_id":66,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":9,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":1,"show_table_of_contents":1,"read_time_minutes":4,"view_count":15,"ordering":0,"published_at":"2026-06-25 11:45:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 11:45:54","updated_at":"2026-06-25 18:34:46","key_takeaways":null,"faq":null,"sources":null,"category_name":"Albania","category_slug":"albania","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/6ae9f253a90dde2799b3dcbeb6b55836.webp","featured_image_alt":"A scientist in a white coat examines medical data and a brain scan on a large digital display.","featured_image_width":1600,"featured_image_height":1066,"intro_image_path":"/api/uploads/2026/06/6ae9f253a90dde2799b3dcbeb6b55836.webp","intro_image_width":1600,"intro_image_height":1066,"seo_title":"Albania Plans AI-Powered Healthcare by 2026 | AI Balkans","meta_description":"Albania is preparing to introduce artificial intelligence into its national healthcare system in 2026, marking one of the country’s most important…","robots":"index","canonical_url":"","og_title":"","og_description":""}],"latest":[{"id":179,"title":"AINOW v0.95 Expands Free AI Learning With DEV","slug":"ainow-v0-95-expands-free-ai-learning-with-dev","short_description":null,"content":"North Macedonia-based AINOW Society has released version 0.95 of its free AI Literacy Platform, introducing a new development-focused learning environment alongside improvements to its existing Education and Business applications.
\nThe update expands the platform beyond AI fundamentals and professional use into the technical aspects of building AI-powered applications.
\n
The new environment, called AINOW DEV, introduces learning materials covering AI skills, Model Context Protocol (MCP) connectors and practical workflows. It is designed for students, developers and professionals who want to understand how AI systems interact with external tools and services.
\nAINOW Society, which also publishes AI Balkans, developed the platform as an open-source educational resource available in Macedonian, Albanian and English.
\nThe AINOW AI Literacy Platform originally focused on helping teachers and educational institutions understand artificial intelligence and use it responsibly.
\nIts Education environment currently includes 19 learning chapters, 1,106 educational prompts, 57 tools and a glossary of 60 AI-related terms. Teachers can use these resources to prepare lessons, develop classroom activities and create assessments.
\n
The platform expanded into professional applications with version 0.94, which introduced AINOW Business.
\nThat environment now provides 3,270 prompts across 33 professional fields, 2,190 AI agent definitions and 453 practical scenarios. Its resources cover areas including management, administration, finance, communication and software development.
\n
With version 0.95, users can access a third environment focused on understanding and developing AI applications.
\nThe new DEV environment includes a 13-chapter learning guide, seven AI tool profiles and three reference libraries containing:
\nMCP, or Model Context Protocol, is a technical standard that allows AI applications to communicate with external tools and data sources.
\n
As AI assistants become capable of performing tasks beyond generating text, understanding these connections is increasingly relevant to developers and organisations exploring automation.
\nDEV also provides practical exercises, tools for examining original AI skill files and a calculator for estimating token and context requirements.
\nThe reference libraries are intended for education and exploration. They do not represent automatically installed connectors or immediately executable integrations.
\nVersion 0.95 also introduces a unified interface across Education, Business and DEV.
\nThe update includes improved navigation, more consistent language controls, updated prompt libraries and better reading options on mobile devices.
\nUsers can now work with longer prompts through improved reading views, while mobile users benefit from a full-screen interface that makes reviewing and editing content easier.
\nThe release also improves local note-taking, settings and backup management. Business and DEV now provide separate local workspaces for saving notes and preferences.
\nTechnical changes address problems involving multilingual search, browser navigation and data synchronisation. The project's published release notes also report fixes for three previously identified GitHub CodeQL findings.
\nThe downloadable version passed automated verification checks, although additional physical-device and screen-reader testing was still pending at publication.
\nOne of the platform's distinguishing features is its approach to accessibility.
\nAll three learning environments are free to use and do not require an AINOW account. Users can access the online version or download the application for local use without installing a separate server.
\nDownloaded educational materials can also be used offline, although external AI services may still require internet access and separate accounts.
\nFor users in North Macedonia and the wider Western Balkans, the availability of learning materials in Macedonian and Albanian provides an alternative to relying entirely on English-language technical resources.
\nThe platform also offers practical material for schools, smaller organisations and individuals who may not have the budget for commercial training platforms.
\nIts open-source model allows developers and institutions to inspect the application code, adapt it to their needs and contribute improvements, subject to the project's licensing requirements.
\nThe new DEV environment extends AINOW's original educational focus toward technical AI literacy, giving learners a way to explore how AI-powered applications are assembled, connected and maintained.
\nThe platform is available at edu.ainow.mk. Further information about the update is available in the official AINOW Society announcement, while the downloadable application and technical release notes can be found on GitHub.
","featured_image_id":425,"intro_image_id":425,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":7,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":4,"ordering":0,"published_at":"2026-10-26 22:51:57","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-26 22:51:57","updated_at":"2026-10-26 22:51:57","key_takeaways":"[\"AINOW Society has released version 0.95 of its free AI Literacy Platform, introducing DEV as a third learning environment alongside Education and Business.\",\"AINOW DEV includes 13 learning chapters, 1,000 AI skills, 1,000 MCP references and 200 workflows designed to help users understand AI development and integrations.\",\"The platform is open source, supports Macedonian, Albanian and English, and offers offline access to downloaded learning resources without mandatory registration.\"]","faq":"[{\"q\":\"What is new in AINOW AI Literacy v0.95?\",\"a\":\"The update introduces AINOW DEV, a learning environment focused on AI development, MCP connectors, reusable skills and workflows. It also improves navigation, prompt reading, mobile usability and security across the platform.\"},{\"q\":\"Who can use the AINOW AI Literacy Platform?\",\"a\":\"The platform serves educators, students, professionals, organisations and developers through three dedicated environments: Education, Business and DEV.\"},{\"q\":\"Is AINOW AI Literacy free and available offline?\",\"a\":\"Yes. The platform is free, open source and accessible without registration. Downloaded learning materials can be used offline, although external AI services may require internet access.\"}]","sources":"[{\"title\":\"AINOW AI Literacy Platform v0.95 Introduces DEV\",\"url\":\"https://ainow.mk/blog/ainow-ai-literacy-platform-v0-95-introduces-dev\"}]","category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Aleksandra Todorovikj","author_slug":"aleksandra-todorovikj","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/ffc8dd720f66d531239cbc548799965c.webp","featured_image_alt":"Ainow literacy","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/ffc8dd720f66d531239cbc548799965c.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"AINOW v0.95 Adds DEV to Free AI Learning Platform","meta_description":"AINOW v0.95 launches DEV with 1,000 AI skills, 1,000 MCP references and 200 workflows. Free AI education in three languages.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":178,"title":"Sweden Blocks ChatGPT and Claude on Government Work Computers","slug":"sweden-blocks-chatgpt-and-claude-on-government-work-computers","short_description":null,"content":"Sweden's Government Offices are restricting access to ChatGPT and Claude on employees' work computers from October 26, directing staff toward approved artificial intelligence services over concerns about information security.
\nThe decision, first reported by Swedish public broadcaster Sveriges Radio on October 23, affects the Government Offices (Regeringskansliet), the administrative organisation supporting the prime minister and government ministries.
\nEmployees will no longer be able to access the public web interfaces of OpenAI's ChatGPT and Anthropic's Claude from government-issued computers. Microsoft Copilot remains available as an approved alternative.
\nHowever, the restriction does not amount to a complete ban on either AI model. Some employees will retain access through an approved AI platform operating within the government's controlled environment.
\nThe Swedish Government Offices want employees to use AI services that have been formally reviewed and approved for official work.
\nAccording to Sveriges Radio, the decision follows internal concerns about staff using AI platforms without appropriate agreements or security controls.
\nMikael Wettercrantz, the Government Offices' director of digitalisation, warned in an internal communication that using AI services without agreements could have consequences for information security.
\nThe concern is particularly relevant to government institutions handling confidential documents, internal correspondence, policy proposals and potentially sensitive personal information.
\nWhen employees enter information into publicly accessible AI services, the organisation must understand how that information is processed, stored and protected.
\nWithout appropriate contractual arrangements and technical safeguards, institutions may have limited control over where information goes and who can access it.
\nThe Government Offices have reviewed Microsoft Copilot and strengthened its security configuration. However, officials have declined to disclose the technical details of those changes, citing security considerations.
\nThe decision does not establish that Copilot is inherently more secure than ChatGPT or Claude. It reflects the Government Offices' assessment of the services and configurations it has approved for workplace use.
\nAn important detail distinguishes Sweden's decision from an outright prohibition.
\nIn a statement provided to Swedish broadcaster SVT, the Government Offices confirmed that employees would retain access to ChatGPT and Claude language models through Intric, an approved AI platform.
\nIntric is a Swedish technology company specialising in AI infrastructure for public institutions and organisations handling sensitive information.
\nIn July 2026, the company announced that it had won a contract to supply its AI platform to the Swedish Government Offices.
\nThe platform allows organisations to develop and operate AI assistants using their own information under defined governance arrangements.
\nIntric says its infrastructure operates in Sweden, supports deployment within customers' own environments and provides institutions with greater control over their data.
\nThe arrangement illustrates the distinction between accessing an AI model through a public website and using the same underlying technology through an institutionally managed service.
\nIn Sweden's case, the restriction focuses on the access method and the organisation's ability to control information handling, rather than rejecting the technology itself.
\nThe Government Offices are not the first Swedish public institution to limit access to AI services.
\nAccording to Sveriges Radio's investigation, the Swedish Tax Agency (Skatteverket) and the National Board of Institutional Care (SiS) have already introduced similar restrictions.
\nThe Swedish Enforcement Authority (Kronofogden) is also reviewing the issue before deciding whether to introduce additional controls.
\nEarlier concerns emerged at the Swedish Transport Administration, where an internal audit identified widespread and potentially risky use of several AI services.
\nThe audit raised concerns about both the sensitivity of information employees entered into AI tools and how they subsequently handled AI-generated responses.
\nThese findings suggest that the problem extends beyond selecting an approved chatbot. Public institutions must also consider employee behaviour, access permissions and the reliability of information produced by AI systems.
\nSweden's Agency for Digital Government, Digg, has already issued guidance recommending that public authorities keep generative AI use under organisational supervision.
\nIts recommendations emphasise protecting confidential information, controlling access to institutional data and establishing clear rules about which information employees may process using AI services.
\nSweden's decision comes as public administrations across Europe are exploring how generative AI can support administrative work without exposing sensitive information or weakening accountability.
\nThe Swedish approach demonstrates one possible response: restricting uncontrolled access while maintaining AI capabilities through approved services.
\nFor government IT departments, this requires more than blocking websites. Institutions need clear policies governing permitted tools, data classification, supplier agreements, access controls and staff responsibilities.
\nThere is also a practical distinction between the security of an AI model and the security of the environment in which that model operates.
\nAn organisation may use the same underlying language model through different services, each with its own contractual arrangements, permissions and information-handling controls.
\nThis is particularly relevant to public institutions considering AI adoption without having established formal procurement procedures or internal usage policies.
\nThe Swedish decision raises a broader question for administrations elsewhere in Europe: how can employees benefit from increasingly capable AI systems while institutions retain effective control over sensitive information?
\nFor Sweden, the immediate change takes effect on October 26. What remains less clear is how the new restrictions will affect everyday government work and whether additional Swedish authorities will adopt comparable measures.
","featured_image_id":424,"intro_image_id":424,"featured_image_credit":"Meet Sweden, the AI incubator. Getty Images","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":15,"ordering":0,"published_at":"2026-10-26 08:33:46","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-26 08:33:46","updated_at":"2026-10-26 08:33:46","key_takeaways":"[\"Sweden's Government Offices are blocking direct browser access to ChatGPT and Claude on work computers from October 26, 2026. The restriction does not apply to the general public.\",\"Microsoft Copilot remains an approved workplace service, while access to ChatGPT and Claude models continues through the Intric platform. The decision therefore concerns how AI services are accessed and governed, rather than a complete prohibition of the models.\",\"Other Swedish authorities have introduced similar restrictions or are considering them, reflecting broader concerns about confidentiality, employee AI use and information-security controls.\"]","faq":"[{\"q\":\"Has Sweden banned ChatGPT and Claude?\",\"a\":\"No. Sweden's Government Offices are restricting direct access to the public web interfaces of ChatGPT and Claude on employees' work computers. The restriction is not a nationwide ban.\"},{\"q\":\"Why are the Swedish Government Offices blocking these AI services?\",\"a\":\"The stated reason is information security. Officials want employees to use approved services with appropriate contractual arrangements and security controls.\"},{\"q\":\"Can Swedish government employees still use AI?\",\"a\":\"Yes. Microsoft Copilot remains approved, and employees can access ChatGPT and Claude language models through Intric, an AI platform used by the Government Offices.\"},{\"q\":\"Are other Swedish government agencies introducing similar restrictions?\",\"a\":\"Yes. The Swedish Tax Agency and National Board of Institutional Care have introduced restrictions, while the Swedish Enforcement Authority is considering additional measures.\"}]","sources":"[{\"title\":\"Sveriges Radio, original investigation, 23 October 2026\",\"url\":\"https://www.sverigesradio.se/artikel/chat-gpt-och-claude-forbjuder-regeringskansliet\"},{\"title\":\"SVT Nyheter, government confirmation, updated 24 October\",\"url\":\"https://www.svt.se/nyheter/inrikes/ai-tjanster-sparras-pa-regeringskansliet\"},{\"title\":\"Swedish Agency for Digital Government, guidance on confidentiality and generative AI\",\"url\":\"https://www.digg.se/kunskap-och-stod/regler-och-rekommendationer/regler-och-rekommendationer/sakerstall-sekretess-vid-anvandning-av-generativ-ai\"}]","category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/38e2e86534210d0213839c4518c9443c.webp","featured_image_alt":"Sweden","featured_image_width":1440,"featured_image_height":960,"intro_image_path":"/api/uploads/2026/10/38e2e86534210d0213839c4518c9443c.webp","intro_image_width":1440,"intro_image_height":960,"seo_title":"Sweden Blocks ChatGPT and Claude on Government PCs","meta_description":"Sweden restricts ChatGPT and Claude on government work computers from October 26, citing security concerns. Approved AI services remain available.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":177,"title":"Claude Takes Software Security Beyond Bug Detection","slug":"claude-takes-software-security-beyond-bug-detection","short_description":null,"content":"Anthropic has demonstrated that its Claude AI models can generate mathematical proofs of specific software security properties, offering a new approach to detecting vulnerabilities and verifying that code behaves as intended.
\nIn research published on October 22, Anthropic reported that Claude, working with software verification company Theorem, produced machine-checked proofs covering 239 test entry points across 123 open-source libraries.
\nThe work goes beyond conventional AI-assisted vulnerability detection. Instead of searching for bugs and reporting potential weaknesses, Claude generated mathematical proofs showing that certain programs cannot crash under precisely defined conditions.
\nThe findings suggest that a security technique historically reserved for highly specialised software projects could become more accessible to developers.
\nTraditional software security testing involves running programs, examining their behaviour and attempting to identify weaknesses.
\nAutomated tools can execute millions or billions of tests, but they cannot guarantee that every possible input or execution path has been examined.
\nFormal verification takes a different approach.
\nRather than testing a program against selected inputs, developers define mathematical rules describing how the software should behave. A formal proof then establishes whether those rules hold for every possible execution within the specified conditions.
\nThe difficulty has always been the work required to produce these proofs. They can be considerably larger and more complicated than the software being examined.
\nAnthropic's research explored whether AI could handle much of that complexity.
\nWorking with Theorem, researchers used Claude to generate proofs for software included in Google's OSS-Fuzz project, which continuously tests open-source libraries for security problems.
\nThe researchers compiled the software into RISC-V machine instructions and used mathematical models of the processor and Linux operating system to establish the conditions that needed to be verified.
\nClaude then generated proofs using Rocq, a formal proof system that independently checks whether mathematical arguments are valid.
\nThis distinction matters. The security result does not depend solely on Claude claiming that the software is correct. A separate mathematical proof checker evaluates the generated evidence.
\nAcross the experiment, Claude produced approximately 180 million lines of proof code.
\nAccording to Anthropic, the verified test entry points cannot experience the specified crashes or memory-safety failures for inputs smaller than 64 MB, provided the assumptions used in the verification model hold.
\nThe process also revealed vulnerabilities.
\nAnthropic reported discovering hundreds of bugs while attempting to establish the proofs. Some had remained undetected despite extensive automated testing.
\nOne example involved a decompression library.
\nClaude identified a counter underflow that could cause an out-of-bounds memory write, but only when the program processed an input of approximately 20 MB with a particularly unusual arrangement.
\nSuch conditions can be difficult for conventional fuzzing tools to discover because they typically concentrate on smaller inputs and previously observed execution paths.
\nThe researchers also described memory-access problems in image-processing software and the rediscovery of a vulnerability in the widely used libexpat XML parsing library, identified as CVE-2026-56132.
\nDuring the experiments, Claude was allowed to modify code when it encountered problems that prevented a proof from being completed.
\nThe verification process then continued against the modified software.
\nThat does not mean every discovered issue has already been corrected in the upstream open-source projects. Anthropic has withheld some proofs that might disclose unpatched vulnerabilities.
\nThe findings nevertheless demonstrate an additional use for AI in cybersecurity: attempting to establish a proof can expose weaknesses that existing testing methods have failed to identify.
\nNo, and Anthropic explicitly acknowledges this limitation.
\nThe research does not establish that all 123 libraries are free from vulnerabilities.
\nThe proofs cover particular test entry points, specific security properties and a defined execution environment.
\nThe primary experiments involved single-threaded, statically linked RISC-V programs running under a mathematical model of Linux with a fixed memory budget.
\nMost production software operates in more complicated environments, commonly using x86 or ARM processors, additional dependencies and different operating-system configurations.
\nA successful proof under one set of assumptions does not automatically extend to those environments.
\nThere is another important limitation: formal verification can establish that software satisfies a specification, but it cannot guarantee that the specification describes everything the software should do.
\nA program might satisfy its mathematical requirements while still containing vulnerabilities outside the properties being checked.
\nThe reliability of the underlying processor and operating-system models, as well as the proof checker itself, also remains essential.
\nFor these reasons, formal verification should be considered an additional security mechanism rather than a replacement for code reviews, penetration testing, vulnerability management or existing secure-development practices.
\nThe most significant implication concerns how software security could eventually become part of routine development.
\nToday, developers commonly rely on automated tests, static analysis, dependency scanning and security reviews.
\nFormal verification has generally remained outside ordinary development workflows because of its cost and complexity.
\nIf AI can reliably generate proofs that independent tools can verify, developers could begin applying mathematical security guarantees to selected components without requiring every team to employ specialists in formal methods.
\nThis may be particularly valuable for libraries that process untrusted inputs, including document parsers, image decoders, compression utilities and network components.
\nA vulnerability in one widely used library can affect thousands of applications that depend on it.
\nVerifying critical properties of shared software dependencies could therefore offer benefits beyond individual projects.
\nHowever, the approach is not yet inexpensive or universally applicable.
\nAnthropic acknowledges that generating complex proofs can consume hundreds of millions of AI tokens, making computational cost an important obstacle.
\nThe company is investigating ways to reuse existing proofs and reduce the work required for new projects.
\nTogether with the research, Anthropic and Theorem have released a selection of proofs and supporting verification components for public inspection.
\nTheorem has also made resources available for open-source maintainers interested in applying the approach to their own software.
\nThe next test will be whether developers outside the research collaboration can reproduce the results, extend verification to more common production environments and integrate the process into everyday development.
\nFor now, the research establishes a narrower but important result: AI-generated code security proofs can be checked independently, and the attempt to create those proofs can uncover real software defects.
","featured_image_id":423,"intro_image_id":423,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":18,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":6,"view_count":15,"ordering":0,"published_at":"2026-10-26 08:19:37","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-26 08:19:37","updated_at":"2026-10-26 08:19:37","key_takeaways":"[\"Anthropic reports that Claude generated mathematical security proofs for 239 test entry points across 123 open-source libraries, with the results checked independently using Rocq.\",\"The research uncovered hundreds of bugs, including memory-safety issues that conventional testing had not detected. However, the proofs apply only to specific conditions and do not establish that the libraries are entirely secure.\",\"AI-assisted formal verification could eventually help developers establish stronger security guarantees for critical software components, although computational costs and technical limitations remain significant.\"]","faq":"[{\"q\":\"What did Claude prove about the 123 open-source libraries?\",\"a\":\"Claude generated proofs that 239 selected test entry points across 123 libraries cannot experience specified crashes under defined conditions, including input-size and execution-environment restrictions.\"},{\"q\":\"How is formal verification different from vulnerability scanning?\",\"a\":\"Vulnerability scanning searches for known or potential security weaknesses. Formal verification uses mathematical proofs to establish that defined properties hold across all executions covered by the verification model.\"},{\"q\":\"Can Claude guarantee that software contains no vulnerabilities?\",\"a\":\"No. The proofs establish particular properties under specific assumptions. Software may still contain vulnerabilities outside the verified scope or behave differently in real-world environments.\"},{\"q\":\"Can developers use this technology today?\",\"a\":\"Anthropic and Theorem have published selected verification resources, and Theorem provides an open-source project enrollment process. However, broad adoption still requires specialised tooling, computational resources and careful specification of the properties being verified.\"}]","sources":"[{\"title\":\"Anthropic, LLMs Are Making Formal Software Verification Feasible at a Much Larger Scale, October 22, 2026\",\"url\":\"https://www.anthropic.com/research/formal-verification\"}]","category_name":"Cybersecurity","category_slug":"cybersecurity","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/ab55c0a3fc619bc0af554b7c60012f30.webp","featured_image_alt":"Claude code scan","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/ab55c0a3fc619bc0af554b7c60012f30.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Claude Uses Mathematical Proofs to Verify Software Security","meta_description":"Anthropic says Claude verified security properties across 123 open-source libraries and uncovered bugs missed by traditional testing.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":176,"title":"AI Balkans Is Now Available on Android","slug":"ai-balkans-is-now-available-on-android","short_description":null,"content":"AI Balkans is now available on Android, bringing regional artificial intelligence news, research, policy developments and practical resources directly to mobile devices.
\nDeveloped by AINOW Society, the official AI Balkans application is available to download for free from Google Play. It gives readers across the Western Balkans a dedicated mobile experience for following AI developments in North Macedonia, Albania, Kosovo, Montenegro, Serbia and Bosnia and Herzegovina.
\nThe release extends AI Balkans beyond its website, making regional reporting and reference resources accessible through a dedicated Android application.
\n![\"07](\"/api/uploads/2026/10/729a8dcc3a95a2b92b054e07cf41c635.webp\")
![\"06](\"/api/uploads/2026/10/ea840fc9dbb563f74b3e267d048adb38.webp\")
![\"05](\"/api/uploads/2026/10/a11a089c7273c4593241086684ca5d8d.webp\")
![\"04](\"/api/uploads/2026/10/b4aa285c0740cd2a7cb85e4499a4cd86.webp\")
![\"03](\"/api/uploads/2026/10/19cb565460e5cc41cb3cdfc5fcc87c0f.webp\")
![\"02](\"/api/uploads/2026/10/47fe189b6b7993ecfa8ffc50dae44503.webp\")
![\"01](\"/api/uploads/2026/10/b8dc60991ef137cd1a91fd81f8aee297.webp\")
The application brings together the latest AI news from across the region, alongside international developments that matter to Balkan readers.
\nIts mobile interface includes breaking news, editor's picks and most-read stories. Readers can follow specific countries and topics, save articles for later and continue reading where they left off.
\nReading preferences include light, dark and sepia themes, with adjustable text size and spacing.
\nPersonalisation does not require creating an account. Saved articles, followed topics and reading preferences are stored on the user's device.
\nThe application also includes Lumi, the AI assistant introduced with AI Balkans 2.0.
\nReaders can ask Lumi questions about artificial intelligence developments covered by AI Balkans and receive answers supported by sources.
\nLumi also allows users to listen to responses, copy information and share answers.
\nThe feature connects readers with the publication's existing reporting and reference material, making it easier to explore topics beyond individual headlines.
\nAlongside daily reporting, the Android application provides access to several AI Balkans resources.
\nThe EU AI Act Tracker follows regulatory and policy developments across the six Western Balkan countries, with links to official sources.
\nReaders can also explore the AI Tools directory, AI Glossary and regional AI Directory, which provides information about organisations working in artificial intelligence.
\nA dedicated Android home-screen widget displays the latest headlines without requiring users to open the application.
\nThe app is free to use, contains no advertisements and does not require registration.
\nAI Balkans was established by AINOW Society to improve access to reliable information about artificial intelligence and give regional developments greater visibility.
\nSince its launch, the publication has expanded beyond news reporting to include research, policy monitoring, practical guides and resources for professionals, educators, researchers and the wider public.
\nThe Android application brings these resources into a mobile experience designed for everyday use.
\nReaders can continue accessing AI Balkans through its website, while Android users now have another way to follow the developments shaping artificial intelligence across the region.
\nThe official AI Balkans application is available on Google Play.
\nDownload AI Balkans on Google Play
\nReaders are also welcome to share feedback and suggestions as the application continues to develop.
","featured_image_id":422,"intro_image_id":422,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":3,"view_count":16,"ordering":0,"published_at":"2026-10-25 09:39:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-25 09:39:27","updated_at":"2026-10-25 10:09:55","key_takeaways":"[\"AI Balkans has launched its official Android application on Google Play, providing mobile access to artificial intelligence news from all six Western Balkan countries.\",\"The application includes Lumi, personalised reading, saved articles, regional AI resources and the EU AI Act Tracker.\",\"The app is free, contains no advertisements and does not require users to create an account.\"]","faq":"[{\"q\":\"Is the AI Balkans Android app free?\",\"a\":\"Yes. The application is free to download from Google Play, contains no advertisements and requires no registration.\"},{\"q\":\"What can I do with the AI Balkans app?\",\"a\":\"You can read AI news, follow countries and topics, save articles, ask Lumi questions, explore regional AI resources and access the EU AI Act Tracker.\"},{\"q\":\"Is AI Balkans available for iPhone?\",\"a\":\"The announced release is for Android. An iOS version has not been confirmed.\"},{\"q\":\"Who developed the AI Balkans application?\",\"a\":\"The application was developed by AINOW Society, the organisation behind the AI Balkans publication.\"}]","sources":null,"category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/52f040254899795e1cb762d1f0ac40d6.webp","featured_image_alt":"Aibalkans android","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/52f040254899795e1cb762d1f0ac40d6.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"AI Balkans Android App Now Available on Google Play","meta_description":"Download the official AI Balkans Android app for regional AI news, Lumi assistant, EU AI Act Tracker and personalised reading.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":175,"title":"Anthropic Launches Claude Duet, a Proactive AI Assistant","slug":"anthropic-claude-duet-proactive-ai-agent","short_description":null,"content":"Anthropic introduced Claude Duet on October 21, 2026, a new AI assistant designed to monitor connected applications, identify tasks that need attention and complete work without requiring users to remain in an active conversation.
\nUnlike conventional AI chatbots that wait for instructions, Claude Duet can initiate contact, prepare daily briefings and continue working while users are away from their computers.
\nThe product is entering a limited research preview for Claude Pro and Max subscribers in the United States, with access managed through a waitlist. It is available as a separate application for iOS, Android and the web.
\nClaude Duet is powered by Anthropic's Claude Opus 5.5 model and is designed primarily for professionals and small business owners.
\n![\"Claude](\"/api/uploads/2026/10/b1b733c177d8a3984f65cd549db539f5.webp\")
Each morning, the assistant prepares a briefing based on information from connected applications, including email, calendars and workplace communication tools. It identifies upcoming commitments, unfinished tasks and information that may require a decision.
\nThroughout the day, Duet checks connected accounts every few hours and contacts users when it identifies something important.
\nFor example, an assistant connected to a business email account and calendar could notice an upcoming client meeting, collect relevant information and prepare a briefing before the meeting takes place.
\nIt could also identify unpaid invoices, draft payment reminders or prepare a follow-up message after a customer interaction.
\nUsers can assign more substantial tasks, including creating financial forecasts, building dashboards, developing prototypes and generating web applications.
\nThese capabilities move Claude beyond answering questions and generating content. The assistant is intended to take responsibility for completing parts of a workflow.
\nDuet operates through a dedicated virtual computer in the cloud, equipped with its own browser.
\nThis allows the assistant to continue working even when the user's laptop or smartphone is switched off.
\nRather than requiring users to manage multiple conversations, Duet maintains a continuing interaction that incorporates information from their existing Claude account, including memory, skills and connected applications.
\nAnthropic says the assistant can work with more than 6,000 connectors and plugins available through the Claude Marketplace.
\n\nThese include services such as Gmail, Slack, QuickBooks, Stripe, HubSpot, Canva and Shopify.
\nA business owner could theoretically ask Duet to review outstanding invoices, prepare reminders, update customer records and compile a report using information from several connected services.
\nHowever, the availability of integrations does not guarantee that every workflow can be completed reliably or without additional configuration.
\nGiving an AI assistant access to business applications introduces security and accountability questions.
\nAn agent that can read correspondence, access financial records and prepare transactions has considerably more operational responsibility than a chatbot generating text.
\nAnthropic says Claude Duet requires explicit approval before performing sensitive actions, including sending messages, making payments, signing documents and deleting files.
\nCertain operations, such as trading securities or transferring money between accounts, remain prohibited even with user approval.
\nThe company also says Duet cannot directly access passwords or authentication tokens and uses additional safeguards intended to prevent sensitive information from being disclosed.
\nThese protections include a dedicated security classifier, defenses against malicious instructions embedded in external content and penetration testing of the system's infrastructure.
\nNevertheless, such safeguards do not eliminate the risks associated with autonomous agents.
\nAn assistant could misunderstand instructions, incorrectly prioritize information or produce an inaccurate document. Malicious content encountered through connected applications could also attempt to influence its behavior.
\nAnthropic acknowledges that the preview version can misinterpret conversations and make mistakes when deciding which issues require immediate attention.
\nFor businesses, this makes human review particularly important when AI-generated work involves customers, financial transactions or confidential information.
\nClaude Duet is not yet available to subscribers in North Macedonia or other Balkan markets.
\nThe initial research preview is restricted to US-based Pro and Max subscribers. Team and Enterprise plans are not included.
\nAnthropic has not announced a timetable for broader international availability.
\nDuring the preview, Duet shares the usage limits of the user's existing Claude subscription.
\nIts background checks consume those limits, meaning that additional activity by the assistant can reduce the capacity available for other Claude tasks.
\nDuet also does not continuously inspect connected applications in real time. Its periodic checks mean that important information may be detected some time after it arrives.
\nThe company describes the product as an early version, with some capabilities still under development.
\nAlthough Claude Duet is currently unavailable in the region, its introduction illustrates how AI tools are moving toward more active participation in business operations.
\nFor smaller companies, the potential appeal is straightforward. Tasks such as following up on invoices, preparing meeting materials, reviewing communications and compiling reports consume working hours even when they do not require specialist expertise.
\nDelegating parts of that work to an AI assistant could reduce administrative effort.
\nBut practical adoption requires more than connecting an account.
\nBusinesses would need to determine which information an AI agent may access, which tasks it can perform independently and which actions require human authorization.
\nThey would also need appropriate controls for personal data, confidential business records and access permissions.
\nThese considerations are particularly relevant for organizations that already depend on Microsoft 365, Google Workspace, customer management systems and cloud-based financial software.
\nThe immediate question for regional businesses is not whether they should adopt Claude Duet, which they currently cannot access, but how they would govern similar autonomous tools once they become available.
\nAnthropic has not provided a timeline for expanding Duet beyond the United States. Its performance during the research preview, particularly its reliability and ability to handle sensitive workflows, will be important factors to watch.
","featured_image_id":413,"intro_image_id":413,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":17,"ordering":0,"published_at":"2026-10-22 16:22:48","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-22 16:22:48","updated_at":"2026-10-22 16:22:48","key_takeaways":"[\"Anthropic launched Claude Duet on October 21, 2026, introducing a proactive AI assistant that monitors connected applications and works in the background.\",\"Duet can prepare briefings, identify unfinished tasks, create reports and complete multi-step workflows, while requiring approval for sensitive actions.\",\"The research preview is limited to US Pro and Max subscribers. Availability in the Balkans has not been announced.\"]","faq":"[{\"q\":\"What is Claude Duet?\",\"a\":\"Claude Duet is an AI assistant developed by Anthropic that can proactively identify tasks, work across connected applications and complete assignments in the background.\"},{\"q\":\"How is Claude Duet different from regular Claude?\",\"a\":\"Regular Claude primarily responds to user requests during conversations. Duet can check connected applications periodically, initiate contact when something needs attention and continue working on tasks without an active conversation.\"},{\"q\":\"Can Claude Duet send emails or make payments automatically?\",\"a\":\"Duet can prepare these actions, but it requires user approval before sending messages, making payments or performing other sensitive operations.\"},{\"q\":\"Is Claude Duet available in North Macedonia?\",\"a\":\"No. As of October 22, 2026, the research preview is available by waitlist only to Claude Pro and Max subscribers in the United States.\"}]","sources":"[{\"title\":\"Anthropic, Introducing Claude Duet, October 21, 2026\",\"url\":\"https://claude.com/resources/articles/claude-duet\"},{\"title\":\"Claude Duet, Official Product Page\",\"url\":\"https://claude.com/product/claude-duet\"},{\"title\":\"Claude Release Notes, October 2026\",\"url\":\"https://support.claude.com/en/articles/12138966-release-notes\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/10a217a592dfbbe51b4fda2e842053df.webp","featured_image_alt":"Claude Duet","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/10a217a592dfbbe51b4fda2e842053df.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Claude Duet Is Anthropic's New AI Agent","meta_description":"Anthropic launches Claude Duet, an AI agent that manages tasks in the background. Explore its features, security controls and availability.","robots":"index","canonical_url":"anthropic-claude-duet-proactive-ai-agent","og_title":"","og_description":""},{"id":174,"title":"PHP2Shell Sparks Global Security Alert for Websites","slug":"php2shell-sparks-global-security-alert-for-websites","short_description":null,"content":"Two critical vulnerabilities in PHP have triggered emergency security updates worldwide. WordPress sites, online stores and institutional portals could be exposed, while AI-assisted research played a central role in uncovering the danger.
\nTwo critical security vulnerabilities in PHP, the programming language powering much of the internet, prompted an emergency security release on 21 October 2026. The flaws, collectively known as PHP2Shell, could allow attackers to execute malicious code on vulnerable web servers without needing a username or password.
\nThe PHP security team released fixes for both vulnerabilities, identified as CVE-2026-109618 and CVE-2026-109619. Each received a critical severity score of 9.2 under CVSS v4.0.
\nThe warning extends well beyond one website platform. WordPress, Joomla, Drupal, Magento and many custom web applications depend on PHP. A vulnerable PHP installation can expose an application even when its own code is properly maintained.
\nFor the Western Balkans, where PHP-based content management systems are widely used by businesses, media organisations and public institutions, the incident raises an immediate operational question: who is responsible for updating the software running their websites?
\nPHP is a server-side programming language. When someone opens a PHP-based website, the server uses PHP to process requests, retrieve information and generate the content displayed in the visitor's browser.
\nPHP2Shell involves two separate weaknesses in how PHP manages computer memory and handles errors.
\nThe first vulnerability, CVE-2026-109618, concerns memory corruption during an out-of-memory condition. Under certain circumstances, PHP can mistakenly treat partially initialised memory as valid.
\nThe second, CVE-2026-109619, involves internal memory references that can remain accessible after the memory has been released.
\nThese weaknesses can create opportunities for attackers to execute commands on a vulnerable server. Researchers also demonstrated scenarios involving exposure of sensitive process memory, potentially including authentication tokens and other confidential information.
\nAn attacker does not necessarily need access to the website's administration panel. Depending on the vulnerable configuration and exploitation method, specially crafted web requests may be sufficient.
\nThe PHP security team considers Apache installations using mod_php particularly exposed because researchers demonstrated an attack that could execute before an application's own code runs. Other configurations, including PHP-FPM, php-cgi, FrankenPHP and LiteSpeed, also require attention.
Not exactly. This distinction matters.
\nWordPress is built using PHP, but the newly disclosed vulnerabilities exist in PHP itself, rather than WordPress's core software, themes or plugins.
\nOn 21 October, WordPress published an official security notice urging website owners to update PHP immediately. It confirmed that no WordPress update was required or planned specifically to address these vulnerabilities.
\nResearchers nevertheless demonstrated successful attacks against WordPress, Drupal and Joomla applications running on vulnerable PHP installations.
\nThis means a website can be running the latest WordPress release, with updated plugins and strong administrative passwords, while remaining exposed because its underlying PHP runtime has not been patched.
\nThe same principle applies to WooCommerce stores, PHP-based learning platforms, custom institutional portals and applications developed using frameworks such as Laravel.
\nThere is also an important distinction between self-hosted WordPress websites and managed services.
\nAutomattic confirmed that sites hosted on WordPress.com, WordPress VIP, WP Cloud, Pressable and Tumblr were already protected as of 21 October. Owners of independently hosted WordPress websites should verify their hosting provider's PHP patch status rather than assuming they received the same protection.
\nThe scale of PHP adoption makes this incident particularly relevant to regional website operators.
\nAccording to W3Techs measurements from October 2026, PHP is used by approximately 69% of websites whose server-side programming language can be identified. WordPress powers approximately 40% of surveyed websites globally.
\nThe regional figures reveal substantial dependence on these technologies.
\nPHP and WordPress usage across selected Balkan domain markets
\n| Domain market | \nWordPress share | \nPHP share | \n
|---|---|---|
| North Macedonia (.mk) | \n85.1% | \n82.6% | \n
| Serbia (.rs) | \n83.7% | \n90.0% | \n
| Croatia (.hr) | \n79.8% | \n85.1% | \n
| Slovenia (.si) | \n76.9% | \n86.5% | \n
| Montenegro (.me) | \n54.5% | \n58.4% | \n
Source: W3Techs, observations from September and October 2026.
\nHow to interpret the figures: WordPress percentages describe its share among websites with an identifiable content management system. PHP percentages describe usage among websites with an identifiable server-side programming language. These are different statistical populations. The measurements cover country-code domains, not every website operated within each country, and the figures are not estimates of vulnerable websites.
\nThe distinction is especially important for Montenegro because .me is widely used internationally and does not exclusively represent Montenegrin organisations.
Additional data from BuiltWith illustrates the regional presence of WordPress. Its October-accessed country reports identify approximately 10,400 WordPress websites associated with North Macedonia, 10,100 with Bosnia and Herzegovina, and 9,900 with Albania. These are technology-detection estimates, not complete national website inventories.
\nComparable recent figures could not be independently established for every Western Balkan economy, including Kosovo. There is also no verified regional count of PHP installations that remain unpatched.
\nStill, the available measurements establish the central point: PHP and WordPress are important components of the region's web infrastructure.
\nA vulnerability in a programming language creates a different security challenge from a flaw in an individual plugin or website feature.
\nA web development agency may have built dozens of websites using different CMS installations, all running on the same PHP version. Updating each website's application software does not necessarily address a vulnerability in the shared runtime.
\nFor smaller businesses and nonprofit organisations, PHP maintenance is frequently delegated to a hosting provider.
\nGovernment bodies, municipalities, universities and healthcare organisations may face a more complicated problem. Their websites can include older applications, custom integrations and services that cannot easily be moved to a newer PHP release.
\nIf an attacker gains remote code execution privileges, the consequences may extend beyond changing website content. Depending on the server's configuration and access permissions, the attacker could access confidential information, compromise application credentials or use the affected system as a starting point for further intrusion.
\nThere is currently no verified public evidence establishing the number of compromised websites in the Western Balkans because of PHP2Shell. The regional usage figures should not be interpreted as confirmed attacks or as proof that particular institutions are vulnerable.
\nThe immediate responsibility is to identify potentially affected systems and establish whether their PHP runtimes have received the security fixes.
\nThe PHP project has released four patched versions:
\n| PHP branch | \nMinimum fixed version | \n
|---|---|
| PHP 8.2 | \n8.2.35 | \n
| PHP 8.3 | \n8.3.36 | \n
| PHP 8.4 | \n8.4.27 | \n
| PHP 8.5 | \n8.5.12 | \n
Administrators should follow these steps:
\nIdentify every PHP installation. Check production websites, development environments exposed to the internet, subdomains and older applications. WordPress administrators can find their PHP version under Tools > Site Health > Info > Server.
\nApply the appropriate security update. Upgrade the PHP runtime to a fixed version, or install the relevant vendor-maintained package containing the security backport. Restart affected PHP services and workers where required.
\nContact the hosting provider. On shared or managed hosting, the provider often controls PHP updates. Ask for confirmation that both CVE-2026-109618 and CVE-2026-109619 have been addressed.
\nReview unsupported PHP versions. PHP 8.1 and older are no longer supported by the main PHP project. Some operating-system vendors provide backported security fixes, but organisations should plan migration to a supported PHP release.
\nMaintain additional protections. Keep web application firewall protections enabled, monitor unusual activity and preserve relevant logs if compromise is suspected. A firewall is an additional defence, not a replacement for patching.
\nBefore a major PHP version upgrade, administrators should also back up their applications and databases and test compatibility where possible.
\nIf a vulnerable public-facing system cannot be patched promptly, security authorities recommend considering temporary isolation from the internet until effective remediation is available.
\nCloudflare confirmed that it deployed additional web application firewall protections for PHP2Shell on 21 October, including protections for customers using its Free plan.
\nThe company also reported observing attack traffic targeting its customers and attempts to modify malicious requests to evade security controls.
\nThis confirms active attempts to target the vulnerabilities, but it does not establish how many attacks succeeded.
\nCloudflare has explicitly advised customers to update PHP even when its protections are enabled. The PHP security team similarly states that temporary mitigations cannot provide the same protection as fixing the underlying software.
\nWebsite owners should therefore treat firewall rules as supplementary protection while ensuring their hosting environments are updated.
\nThe discovery process is another reason this incident has attracted attention.
\nAccording to the PHP Foundation, the first indications of one underlying memory-handling problem appeared during security work in 2024. Developers recognised a potential issue but struggled to identify a practical exploitation method and resolve difficult compatibility concerns.
\nIn August 2026, security researchers demonstrated its more serious implications.
\nCalif researcher Ryan A. reported a working remote code execution vulnerability on 11 August, using advanced AI-assisted security research. OpenAI independently reported the same underlying problem that day, while Wiz subsequently contributed additional findings and attack paths.
\nOpenAI confirmed that Calif used GPT-5.6 Sol during the research process as part of its Patch the Planet initiative.
\nThe resulting coordinated response involved researchers, PHP maintainers, infrastructure providers and security organisations. PHP's own developers remained responsible for developing and releasing the fixes.
\nThe PHP Foundation also acknowledged that the earlier delay in resolving the underlying problem reflected limitations in its security processes and available engineering resources.
\nThis is an important distinction in understanding AI's contribution. The technology helped researchers demonstrate severe weaknesses, but validating the findings, producing compatible fixes and coordinating a safe release required substantial human expertise.
\nThe same AI-assisted capabilities could also help attackers analyse vulnerabilities and develop exploitation techniques more quickly.
\nThe PHP Foundation has urged organisations to install the emergency releases immediately and warned that no temporary mitigation offers complete protection.
\nThe incident also exposed weaknesses in coordinated vulnerability disclosure. Information about the vulnerabilities began circulating before the scheduled release, forcing maintainers to accelerate their plans.
\nFor website operators across the Balkans, the priority is practical rather than speculative: verify PHP versions, confirm patches and identify systems still dependent on unsupported software.
\nThe broader security implications will become clearer as hosting providers complete their updates and researchers publish additional findings about exploitation attempts.
\nFor now, the central question for any organisation running a PHP-based website is straightforward: has the server been patched?
","featured_image_id":411,"intro_image_id":411,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":18,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":9,"view_count":24,"ordering":0,"published_at":"2026-10-21 18:12:55","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-21 18:12:55","updated_at":"2026-10-21 18:12:55","key_takeaways":"[\"PHP2Shell affects PHP itself, not WordPress core. Two critical vulnerabilities could allow unauthenticated attackers to execute code on affected servers. WordPress, Drupal, Joomla and other PHP-based applications may be exposed if their underlying PHP installations remain unpatched.\",\"PHP and WordPress have a significant presence in the Balkans. W3Techs data shows strong adoption across several regional domain markets, including North Macedonia, Serbia, Croatia and Slovenia. These figures indicate technological dependence, not the number of vulnerable or compromised websites.\",\"Website owners must verify their PHP security updates. Emergency patches were released on 21 October 2026. Updating WordPress plugins or enabling a firewall does not replace patching PHP. Administrators should verify the installed version and obtain confirmation from their hosting provider.\"]","faq":"[{\"q\":\"What is PHP2Shell?\",\"a\":\"PHP2Shell refers to two critical PHP vulnerabilities, CVE-2026-109618 and CVE-2026-109619, disclosed in October 2026. Under certain conditions, they could allow attackers to execute malicious commands on vulnerable servers without authentication.\"},{\"q\":\"Does PHP2Shell affect every WordPress website?\",\"a\":\"WordPress uses PHP, so self-hosted installations running vulnerable PHP versions may be exposed. However, patched servers are protected against these vulnerabilities. WordPress.com and several other Automattic-managed hosting services have confirmed that their customers are protected.\"},{\"q\":\"Can updating WordPress fix PHP2Shell?\",\"a\":\"No. These vulnerabilities are in the PHP runtime, not WordPress core. Website owners must update PHP or ensure their hosting provider has applied the appropriate security fixes.\"},{\"q\":\"Which PHP versions are safe?\",\"a\":\"The official fixed releases are PHP 8.2.35, 8.3.36, 8.4.27 and 8.5.12. Organisations using older versions should consult their distribution or hosting vendor about security backports and plan an upgrade to a supported release.\"},{\"q\":\"How do I check my PHP version in WordPress?\",\"a\":\"Open the WordPress administration dashboard and navigate to Tools > Site Health > Info > Server. The PHP version should be displayed there. You can also check through your hosting control panel.\"},{\"q\":\"Are websites in the Western Balkans already being attacked?\",\"a\":\"Cloudflare has confirmed observing attack traffic associated with PHP2Shell internationally. However, verified figures establishing successful compromises specifically in the Western Balkans were not available at the time of this report.\"}]","sources":"[{\"title\":\"PHP Project: Critical Remote Code Execution Vulnerabilities.\",\"url\":\"https://www.php.net/archive/2026.php\"},{\"title\":\"PHP Foundation: Today's PHP Security Release\",\"url\":\"https://thephp.foundation/blog/2026/10/21/todays-php-security-release-and-how-were-strengthening-php-security/\"},{\"title\":\"WordPress: Critical PHP Security Release\",\"url\":\"https://wordpress.org/news/2026/10/critical-php-security-release\"}]","category_name":"Cybersecurity","category_slug":"cybersecurity","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/18258cdb232fc62815dc9dff161cf97e.webp","featured_image_alt":"PHP Vulnerability","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/18258cdb232fc62815dc9dff161cf97e.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"PHP2Shell Sparks Global Security Alert for Websites","meta_description":"Two critical PHP vulnerabilities threaten WordPress and other websites. See Balkan usage data and the PHP updates administrators must install.","robots":"index","canonical_url":"php2shell-php-wordpress-balkans-security","og_title":"","og_description":""},{"id":173,"title":"Solo Entrepreneur Toolkit 2026: 15 Tools You Can Start Using for Free","slug":"solo-entrepreneur-toolkit-2026-15-tools-you-can-start-using-for-free","short_description":null,"content":"Running a business alone means doing the work of several people. One day you're serving customers, the next you're building a website, preparing proposals, creating marketing materials or chasing unpaid invoices.
\nFortunately, you don't need an expensive collection of software subscriptions to handle these tasks.
\nAI tools and free online services can help freelancers, consultants, creators and small business owners manage everyday work with less effort.
\nWe've selected 15 practical tools that offer free access, covering everything from AI assistants and business organisation to marketing, invoicing and automation. Some are completely free, while others offer limited free plans.
\nHere's what each tool can do for your business.
\nWebsite: https://duck.ai
\nNeed help writing an email, preparing a proposal or brainstorming marketing ideas? Duck.ai provides free access to several AI models, including options from OpenAI, Anthropic and Mistral.
\n![\"Duck](\"/api/uploads/2026/10/005b5810f1099afd8caf01670ad8c49b.webp\")
Developed by DuckDuckGo, the service allows users to chat with AI without creating an account. Conversations are anonymised, and DuckDuckGo says they aren't used to train AI models.
\nBest for: Business writing, brainstorming, research assistance and everyday questions.
\nFree access: Available with daily usage limits. More advanced models and higher allowances require a subscription.
\nWebsite: https://carrd.co
\nA professional website doesn't have to cost hundreds of euros.
\n![\"Carrd](\"/api/uploads/2026/10/817bcc3322bed1be8d10362adc075742.webp\")
Carrd lets you build responsive, one-page websites without coding. It's useful for displaying services, introducing your business and providing customers with contact information.
\nYou can create up to three websites on its free plan.
\nBest for: Freelancers, consultants and entrepreneurs launching their first website.
\nFree access: Three websites. Custom domains and built-in contact forms require a paid plan, starting at $19 per year.
\nWebsite: https://www.notion.com
\nKeeping track of customers, projects, deadlines and documents becomes difficult when everything is stored in different applications.
\n![\"Notion\"](\"/api/uploads/2026/10/c608f412ced1c0dd75a5005ce1cd3017.webp\")
Notion provides a central workspace for managing tasks, customer information, business notes and project documentation.
\nIts AI features can help create documents, summarise information and organise content.
\nBest for: Project management, customer tracking and business organisation.
\nFree access: The core workspace is free for individuals. Most Notion AI features are available only through a limited trial before requiring a paid Business or Enterprise plan.
\nWebsite: https://tally.so
\nInstead of asking customers to explain their requirements through lengthy email exchanges, send them a form.
\n![\"Tally](\"/api/uploads/2026/10/819518aa55be40acdc506f10e8833b06.webp\")
Tally helps create contact forms, customer questionnaires, registration pages and feedback surveys without coding.
\nYou can also connect responses to Google Sheets to keep customer information organised.
\nBest for: Lead generation, customer enquiries and collecting project requirements.
\nFree access: Unlimited forms and submissions within fair-use limits, with many useful features included.
\nWebsite: https://calendar.google.com
\nScheduling meetings shouldn't require five emails.
\n![\"Google](\"/api/uploads/2026/10/ff5c6d1e2d50ea919ce3c078994f4ddc.webp\")
Google Calendar's appointment scheduling feature lets you define available times and share a booking link with customers.
\nA consultant, tutor or freelancer can use it to organise introductory calls and client appointments.
\nBest for: Consultations, customer meetings and appointment scheduling.
\nFree access: Basic appointment scheduling is available for eligible Google accounts. Advanced scheduling features may require a subscription.
\nWebsite: https://notebook.google
\nFormerly known as NotebookLM, Google's Gemini Notebook helps users understand documents, reports, websites and other research materials.
\n![\"Google](\"/api/uploads/2026/10/91217e56628d507995822e1b24f970d9.webp\")
Upload relevant sources and ask questions about their contents. The AI generates answers with references, allowing you to inspect the supporting information.
\nThis can be particularly useful when preparing for meetings or studying unfamiliar subjects.
\nBest for: Document analysis, research, learning and understanding complex information.
\nFree access: A free version is available with usage limits. Advanced capabilities require eligible paid plans.
\nWebsite: https://gamma.app
\nPreparing a professional presentation can take hours, especially when starting with notes or an unfinished proposal.
\n![\"Gamma](\"/api/uploads/2026/10/33442a52875a0fcc181595ea4669d9c4.webp\")
Gamma uses AI to turn ideas into presentations and visual documents.
\nDescribe what you need, generate an initial draft and customise the content before sharing it with customers.
\nBest for: Business proposals, presentations and startup pitch decks.
\nFree access: A free plan includes introductory AI credits. Additional generation capacity and advanced features require payment.
\nWebsite: https://www.napkin.ai
\nSometimes a diagram explains a business idea better than several paragraphs.
\n![\"Napkin](\"/api/uploads/2026/10/730a4b021cba0d3526a055d114033dca.webp\")
Napkin AI transforms written information into diagrams, workflows and other visual content.
\nYou can use it to illustrate a business process, explain a service or create graphics for a presentation.
\nBest for: Process diagrams, business proposals, infographics and educational materials.
\nFree access: The free plan includes 500 AI credits per week. Exports include Napkin branding, and some advanced features require payment.
\nWebsite: https://www.photopea.com
\nNeed to resize a product image, remove part of a background or prepare marketing graphics?
\n![\"PhotoPea\"](\"/api/uploads/2026/10/3b8ee1e6253c9eb3c9ec12f3caddc087.webp\")
Photopea is a browser-based image editor supporting layers, filters and common professional design formats.
\nIt handles many editing tasks normally associated with desktop graphics software.
\nBest for: Product photographs, promotional graphics and website images.
\nFree access: Core editing features are available free with advertising.
\nWebsite: https://squoosh.app
\nLarge images can slow down websites and negatively affect the browsing experience.
\n![\"Squoosh\"](\"/api/uploads/2026/10/6c723bc5f3757e7821313cb8074bf11b.webp\")
Squoosh lets you compress images while comparing the original and optimised versions.
\nProcessing happens locally in your browser, making it useful when you don't want to upload images to an external server.
\nBest for: Website images, blog illustrations and product photography.
\nFree access: Free to use, with no subscription required.
\nWebsite: https://tinywow.com
\nNeed to convert a PDF, resize an image or change a document's format?
\nTinyWow brings together hundreds of small utilities for PDFs, images, videos and file conversions, alongside AI-assisted tools.
\n![\"Tinywow\"](\"/api/uploads/2026/10/52236197ba6971432016e0c2e4f36a2a.webp\")
Many tasks can be completed without registering.
\nBest for: Document conversion, image editing and quick administrative tasks.
\nFree access: Many utilities are free, although advertising, usage restrictions and additional paid features may apply.
\nPrivacy consideration: Avoid uploading confidential business documents without checking the provider's data-handling policies.
\nWebsite: https://buffer.com
\nManaging social media manually can consume time better spent serving customers.
\n![\"Buffer\"](\"/api/uploads/2026/10/ff6c89fc3abd066d158f839b87e62743.webp\")
Buffer helps prepare and schedule content for supported social platforms, including LinkedIn, Instagram and Facebook.
\nIts AI assistant can also help draft and improve posts.
\nBest for: Social media scheduling and content management.
\nFree access: Connect up to three channels and schedule up to ten posts per channel at a time.
\nWebsite: https://www.brevo.com
\nEmail remains a useful channel for maintaining relationships with customers.
\n![\"Brevo\"](\"/api/uploads/2026/10/91b2509f5d29daab82d832adfb1d12b4.webp\")
Brevo helps entrepreneurs manage subscribers, create newsletters and send marketing campaigns.
\nA freelancer might use it to share monthly updates, useful advice or new service announcements.
\nBest for: Email newsletters, customer communication and basic email marketing.
\nFree access: Up to 300 email sends per day on the free plan. Additional features and larger sending allowances require payment.
\nRemember to obtain appropriate consent before sending marketing emails.
\nWebsite: https://www.zoho.com/invoice
\nPreparing invoices and tracking payments are essential business tasks, but they shouldn't require complicated software.
\n![\"Zoho](\"/api/uploads/2026/10/daf8ab980df1cba40b718adf3b1012bc.webp\")
Zoho Invoice provides tools for creating estimates, issuing invoices, recording expenses and tracking billable time.
\nIt can help independent professionals manage customer billing from one application.
\nBest for: Freelancers, consultants and service-based businesses.
\nFree access: Zoho Invoice offers a free invoicing platform.
\nImportant: Check whether generated invoices comply with your country's accounting and tax requirements.
\nWebsite: https://www.make.com
\nImagine receiving a customer enquiry and automatically saving the details to a spreadsheet while sending yourself a notification.
\n![\"Make](\"/api/uploads/2026/10/317e3a77eba723b2a72641af67967145.webp\")
Make allows users to connect applications and create automated workflows through a visual interface.
\nIt can connect forms, spreadsheets, email services and many other business applications.
\nBest for: Workflow automation, customer enquiries and reducing repetitive administration.
\nFree access: A free plan includes a monthly credit allowance. More demanding automations require paid capacity.
\nYou don't need to use all 15 tools.
\nA freelance consultant could start with Carrd for a website, Tally for customer enquiries and Google Calendar for appointments.
\nNotion could manage projects and deadlines, while Duck.ai helps prepare business communications. Gamma and Napkin AI could support client presentations.
\nFor marketing, Buffer and Brevo provide ways to reach customers. Zoho Invoice helps with billing, and Make can automate repetitive tasks once the business grows.
\nMost importantly, begin with the work you actually need to complete.
\nA free tool that solves a daily problem is often more valuable than an expensive AI subscription you rarely use.
\nThe advantage of these tools isn't that every feature is free forever. It's that entrepreneurs can experiment, establish working processes and identify what genuinely saves time before spending money.
\nSome businesses will eventually need additional storage, automation credits, advanced AI models or professional website features.
\nThose expenses make sense when they support actual business needs.
\nFor a solo entrepreneur, the best software is rarely the one with the longest feature list. It's the one that makes running the business easier.
","featured_image_id":394,"intro_image_id":394,"featured_image_credit":"Jakub Żerdzicki / Unsplash","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":14,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":8,"view_count":21,"ordering":0,"published_at":"2026-10-20 19:19:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-20 19:19:35","updated_at":"2026-10-20 19:20:59","key_takeaways":"[\"Solo entrepreneurs can manage websites, customer enquiries, marketing, invoicing and everyday administration using free or freemium software without investing in expensive business platforms.\",\"AI tools such as Duck.ai, Gemini Notebook, Napkin AI and Gamma can help with writing, research, presentations and visual content. Traditional utilities remain useful for tasks that don't require AI.\",\"Free plans often include usage limits, AI credits or restricted features. Start with the tools that solve immediate business problems and consider paid subscriptions only when necessary.\"]","faq":"[{\"q\":\"Are free AI tools safe for business use?\",\"a\":\"Security and privacy depend on the provider and the information being processed. Review privacy policies and avoid uploading confidential customer records or sensitive financial information without appropriate safeguards.\"},{\"q\":\"Do I need technical skills to use these tools?\",\"a\":\"Most are designed for beginners. Tools such as Carrd, Tally, Gamma and Napkin AI provide visual interfaces, while Make allows users to create many automated workflows without traditional programming.\"},{\"q\":\"Is Notion AI completely free?\",\"a\":\"No. Notion offers a free workspace for organising tasks, documents and projects, but its AI features are limited to trial usage on the free plan.\"},{\"q\":\"Can I run a one-person business using only free tools?\",\"a\":\"Many everyday tasks can be handled with free software, including building a basic website, managing projects, scheduling meetings and creating marketing materials. Some businesses will still need paid services for specific operational or legal requirements.\"},{\"q\":\"What are the best free AI tools for solo entrepreneurs in 2026?\",\"a\":\"Duck.ai is useful for AI-assisted writing, Gemini Notebook helps analyse documents, and Napkin AI creates diagrams and visual content. Gamma can generate presentations, while Notion provides business organisation features with limited free AI access.\"}]","sources":null,"category_name":"Tools & Guides","category_slug":"ai-tools","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/d9f265e308481038b67a862b6f6c0c9d.webp","featured_image_alt":"Laptop Device","featured_image_width":1600,"featured_image_height":1067,"intro_image_path":"/api/uploads/2026/10/d9f265e308481038b67a862b6f6c0c9d.webp","intro_image_width":1600,"intro_image_height":1067,"seo_title":"Solo Entrepreneur Toolkit 2026: 15 Free Tools","meta_description":"Discover 15 free and freemium tools for solo entrepreneurs in 2026. Build websites, manage clients, create content and automate everyday tasks.","robots":"index","canonical_url":"olo-entrepreneur-toolkit-2026-free-tools","og_title":"","og_description":""},{"id":172,"title":"Western Balkans AI Roadmap Presented, Awaiting Leaders' Endorsement","slug":"western-balkans-ai-roadmap-presented-awaiting-leaders-endorsement","short_description":null,"content":"A regional AI roadmap extending to 2030 was presented at the Western Balkans Digital Summit in Montenegro, where government representatives from all six Western Balkan economies discussed digital cooperation, artificial intelligence and closer alignment with European standards.
\nThe eighth Western Balkans Digital Summit, held on October 13–14 in Herceg Novi, brought together government officials, European Union representatives, international organisations and technology-sector participants.
\nAccording to the Regional Cooperation Council (RCC), the summit welcomed the AI Roadmap for the Western Balkans Six, developed by the RCC together with regional partners. The document is intended to establish a common direction for responsible AI development, governance, innovation and institutional readiness across the region.
\nHowever, an important step remains. The RCC's October 14 announcement states that the roadmap is expected to receive endorsement at the forthcoming Berlin Process Leaders' Summit.
\nThat distinction matters because the presentation and political welcome of a regional initiative do not necessarily constitute formal approval by every participating government.
\nThe summit was organised by Montenegro's Ministry of Public Administration in partnership with the RCC, the European Union and other regional and international partners, under Montenegro's 2026 chairmanship of the Berlin Process.
\nGovernment representatives from Albania, Bosnia and Herzegovina, Kosovo, Montenegro, North Macedonia and Serbia took part in the summit's ministerial panel.
\nNorth Macedonia was represented by Radoslav Nastasijević Vardarski, Deputy Minister of Digital Transformation, while Albania was represented by Enkelejda Muçaj, Deputy Minister of Infrastructure and Energy.
\nKosovo's representative was Lulzon Jagxhiu, Minister of Public Administration and Digitalization, and Montenegro was represented by Marash Dukaj, Minister of Public Administration.
\nBosnia and Herzegovina participated through Almir Badnjević, Director of the Agency for Identification Documents, Registers and Data Exchange (IDDEEA), while Serbia was represented by Slaviša Antić, State Secretary at the Ministry of Information and Telecommunications.
\nThe ministerial panel, moderated by the RCC's Danijela Gačević, addressed digital transformation, regional cooperation and alignment with European digital standards.
\nRCC Secretary General Amer Kapetanović also addressed the summit, alongside representatives of EU institutions and international partners.
\nTheir participation confirms that the meeting involved regional government representatives. It does not, by itself, establish that each government formally approved the AI roadmap.
\nDuring the ministerial discussions, North Macedonia's Deputy Minister Radoslav Nastasijević Vardarski highlighted Vezilka, the country's national AI project, as an initiative of national importance.
\nAnother representative of the Ministry of Digital Transformation, Ivan Ilioski, participated in a separate discussion focused on digital identity wallets, one of the country's priorities for modernising electronic public services.
\nThese initiatives reflect North Macedonia's national digital priorities, although their inclusion in summit discussions should not be interpreted as evidence of new commitments under the regional AI roadmap.
\nThe wider programme addressed digital public services, cybersecurity, infrastructure, AI readiness and digital skills.
\nAccording to the RCC, the AI Roadmap for the Western Balkans Six is designed to support coordinated development and responsible use of artificial intelligence.
\nIts publicly described priorities include improving institutional readiness, strengthening AI governance, encouraging innovation and bringing regional approaches closer to European standards.
\nThese areas are particularly relevant as Western Balkan governments expand their use of digital public services and consider how artificial intelligence could be incorporated into public administration and economic development.
\nA regional framework could help participating economies exchange expertise, coordinate approaches and identify shared challenges. Nevertheless, the publicly available summit announcements do not establish specific national funding obligations, binding implementation deadlines or enforceable AI governance measures arising from the roadmap.
\nThose details will be important in determining how the proposed framework moves beyond political agreement into implementation.
\nArtificial intelligence was one part of a wider digital agenda discussed during the two-day gathering.
\nThe summit's conclusions identified priorities including mutual recognition of digital identities, implementation of interoperable digital identity wallets, stronger cybersecurity cooperation, improved connectivity and efforts to reduce roaming costs between the Western Balkans and the European Union.
\nThe conclusions also envisaged continued RCC support for implementing the AI roadmap and coordinating regional digital initiatives.
\nIn his opening address, RCC Secretary General Amer Kapetanović emphasised that regional digital transformation should focus on connecting national systems rather than developing isolated services.
\nThe RCC has supported the Western Balkans Digital Summit since its establishment in 2018. The previous edition was held in Skopje in October 2025.
\nAlthough Montenegro hosted the summit, the roadmap is an RCC-led regional initiative developed with partners from across the Western Balkans.
\nMontenegro's Ministry of Public Administration described the roadmap in its October 18 announcement as having been presented and adopted at the summit. The RCC's official account uses more cautious language, saying that the roadmap was welcomed and that endorsement at the Berlin Process Leaders' Summit was still expected.
\nThe available statements therefore do not establish that all six governments have individually completed a formal endorsement process.
\nThe next development to watch is the leaders' decision under the Berlin Process, followed by publication of any country-specific commitments, implementation arrangements or financing measures.
\nFor the Western Balkans, the roadmap's practical significance will depend on what participating governments subsequently agree to implement and how progress is measured.
","featured_image_id":393,"intro_image_id":393,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":31,"ordering":0,"published_at":"2026-10-18 16:50:41","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-18 16:50:41","updated_at":"2026-10-18 16:50:41","key_takeaways":"[\"The Western Balkans AI Roadmap through 2030 was presented at the 2026 Digital Summit in Herceg Novi, Montenegro.\",\"Government representatives from all six Western Balkan economies participated in the summit's ministerial discussions.\",\"The Regional Cooperation Council developed the roadmap with regional partners, focusing on responsible AI adoption, governance, innovation and alignment with European standards.\",\"The RCC stated that formal endorsement was expected at the Berlin Process Leaders' Summit. Participation at the Digital Summit does not establish formal governmental approval.\",\"North Macedonia highlighted its Vezilka AI project and digital identity initiatives during the discussions.\"]","faq":"[{\"q\":\"What is the Western Balkans AI Roadmap 2030?\",\"a\":\"It is an RCC-led regional initiative intended to guide cooperation on AI readiness, governance, innovation and responsible adoption through 2030.\"},{\"q\":\"Which countries participated in the summit?\",\"a\":\"Albania, Bosnia and Herzegovina, Kosovo, Montenegro, North Macedonia and Serbia were represented at the ministerial panel.\"},{\"q\":\"Has the AI roadmap been formally adopted by all six governments?\",\"a\":\"The RCC reported that the roadmap was welcomed at the Digital Summit and expected endorsement at the forthcoming Berlin Process Leaders' Summit. Individual formal government approvals have not been established by the cited documents.\"},{\"q\":\"Who represented North Macedonia?\",\"a\":\"Deputy Minister of Digital Transformation Radoslav Nastasijević Vardarski participated in the ministerial discussions. Ivan Ilioski also participated in a separate panel on digital identity initiatives.\"},{\"q\":\"What happens after the summit?\",\"a\":\"The next expected step is endorsement through the Berlin Process, followed by clarification of implementation responsibilities, financing and any national commitments.\"}]","sources":"[{\"title\":\"RCC, October 14, 2026: Summit conclusions and AI Roadmap status\",\"url\":\"https://www.rcc.int/news/1083/western-balkans-digital-summit-from-digital-ambition-to-action\"},{\"title\":\"Official Western Balkans Digital Summit 2026 conclusions\",\"url\":\"https://www.see2020.rcc.int/docs/814/conclusions-of-the-western-balkans-digital-summit-2026\"}]","category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/638f26e3b54535227acc12d920ace4a3.webp","featured_image_alt":"WB RCC","featured_image_width":1600,"featured_image_height":1066,"intro_image_path":"/api/uploads/2026/10/638f26e3b54535227acc12d920ace4a3.webp","intro_image_width":1600,"intro_image_height":1066,"seo_title":"Western Balkans AI Roadmap Awaits Endorsement","meta_description":"Six Western Balkan economies joined AI roadmap discussions in Montenegro. The RCC says formal endorsement is expected at the Berlin Process summit.","robots":"index","canonical_url":"western-balkans-ai-roadmap-awaits-endorsement","og_title":"","og_description":""},{"id":171,"title":"DeepSeek Upgrades V4.1 Flash and Opens Search API","slug":"deepseek-upgrades-v4-1-flash-and-opens-search-api","short_description":null,"content":"DeepSeek has released an upgraded version of its V4.1 Flash artificial intelligence model and opened its proprietary web search infrastructure to developers. The October 17 announcement introduces DeepSeek-V4.1-Flash-1017, with improvements to coding and autonomous task execution, alongside two new APIs for retrieving information from the web.
\nThe updated model is already available through DeepSeek's website, mobile application and API. Existing developers can access it using the deepseek-flash model identifier without changing their integration or paying higher model rates.
The simultaneous release of web search capabilities gives developers another option for building AI assistants and applications that need current information rather than relying exclusively on knowledge stored during model training.
\nDeepSeek says V4.1 Flash 1017 completes tasks faster, produces shorter outputs and performs better across several technical evaluations than the previous version released in September.
\nThe improvements focus particularly on AI agents, systems that can use tools, execute instructions and complete sequences of tasks with limited human intervention.
\nAccording to the company's published benchmarks, the updated model achieved a score of 42.3 on Terminal-Bench 4.0, compared with 31.2 for the previous release. Its ProgramBench score increased from 20.3 to 29.1, while its DeepSWE v1.1 result rose from 74.2 to 77.0.
\nThese are company-reported results, rather than independent assessments of the new release. Performance in real development environments may differ depending on the tools, evaluation framework and complexity of the task.
\nDeepSeek has also optimized the model for agent frameworks including DeepSeek Harness, Tencent CodeBuddy and WorkBuddy. Developers using DeepSeek Harness are advised to upgrade to version 0.2.1-rc.1 or later.
\nOne area DeepSeek is emphasizing is the development of complete software projects from relatively short instructions.
\nThe company says it has used reinforcement learning with self-feedback to improve the model's ability to handle longer development workflows, including game creation and 3D modelling.
\nIn its release materials, DeepSeek demonstrated several projects developed with the model, including a browser-based racing game, a recreation of Portal 2 using the Godot game engine, a 3D model of Beijing's Forbidden City and an interactive historical scene called Valley of Ages.
\nThe examples illustrate the type of extended work the model is being trained to handle. They do not, by themselves, establish how consistently it performs across unfamiliar projects or production software environments.
\nAlongside the model upgrade, DeepSeek has launched public access to its Search and Fetch APIs.
\nThe Search API allows applications to query the web and receive structured results. The Fetch API retrieves the content of an individual webpage and returns it in Markdown format, making it easier for AI systems to process and summarize information.
\nAccording to DeepSeek, its internal search infrastructure already handles hundreds of millions of requests daily and indexes hundreds of billions of webpages.
\nThe company says its search engine supports multiple languages and uses filtering mechanisms intended to reduce search-engine spam, manipulation aimed at AI-generated answers and low-quality AI-produced material. These claims have not yet been independently established through publicly available comparative testing.
\nBoth interfaces can be used independently of the language model through authenticated API requests. They are particularly relevant to AI agents, research assistants and retrieval-augmented generation applications that need access to information published after a model's training period.
\nDeepSeek has priced both Search and Fetch at $0.75 per 1,000 successful requests.
\nCharges apply to requests that return usable results or webpage content. Failed requests, timeouts and searches without results are not billed under the published rules.
\nThe company is temporarily waiving these search charges for eligible server-side tool calls made through DeepSeek Harness, Claude Code and Codex using the DeepSeek API. Billing for those calls is scheduled to begin on October 31 at 16:00 UTC.
\nThe model itself retains its existing API pricing structure, including lower off-peak rates. Developers already using deepseek-flash automatically receive the upgraded model.
That distinction matters for development costs. Model inference continues to be billed according to token usage, while the new search tools are charged separately per successful request.
\nDeepSeek has published the V4.1 Flash 1017 model weights on Hugging Face and ModelScope, allowing developers and researchers to examine the release and explore deployments beyond the company's hosted API, subject to the applicable model terms and infrastructure requirements.
\nThe model builds on the V4.1 Flash architecture introduced in September, which combines a mixture-of-experts design with native visual understanding.
\nPublishing the updated weights gives developers a way to evaluate the model separately from DeepSeek's hosted service. However, the Search and Fetch APIs remain distinct hosted services, and releasing model weights does not mean the company's proprietary search index has been made publicly available.
\nFor organizations building coding assistants or web-connected AI tools, the October release introduces two practical changes: an upgraded model available through an existing API identifier and a separately priced route to live web information.
\nThe next test will be how the upgraded model performs under independent evaluation and whether DeepSeek's search service can consistently deliver the relevance, freshness and reliability claimed in its documentation. For developers using the temporarily free search integrations, the immediate deadline is October 31, when usage-based billing begins.
","featured_image_id":392,"intro_image_id":392,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":27,"ordering":0,"published_at":"2026-10-17 16:31:05","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-17 16:31:05","updated_at":"2026-10-17 16:31:05","key_takeaways":"[\"DeepSeek releases V4.1 Flash 1017. The October 17 upgrade improves company-reported coding and agent benchmarks while maintaining existing model API pricing.\",\"Search and Fetch APIs are now publicly available. Developers can access DeepSeek's proprietary web search infrastructure to retrieve current information and webpage content.\",\"Search pricing is $0.75 per 1,000 successful requests. Eligible server-side search calls through DeepSeek Harness, Claude Code and Codex remain temporarily free until October 31, 2026.\",\"The upgraded model weights are available. DeepSeek has published V4.1 Flash 1017 on Hugging Face and ModelScope, although its proprietary search infrastructure remains a separate hosted service.\"]","faq":"[{\"q\":\"What is DeepSeek V4.1 Flash 1017?\",\"a\":\"DeepSeek V4.1 Flash 1017 is an upgraded AI model released on October 17, 2026. It introduces improvements to coding, task execution and agent workflows, according to DeepSeek's published evaluations.\"},{\"q\":\"Is DeepSeek V4.1 Flash 1017 available now?\",\"a\":\"Yes. The model is available through DeepSeek's website, mobile application and API. Developers can access it using the existing deepseek-flash model identifier.\"},{\"q\":\"What is the DeepSeek Search API?\",\"a\":\"It is a web search interface that allows developers to retrieve structured search results from DeepSeek's proprietary search engine. A separate Fetch API retrieves webpage content in Markdown format.\"},{\"q\":\"How much does DeepSeek Search cost?\",\"a\":\"DeepSeek charges $0.75 per 1,000 successful Search API requests and the same rate for Fetch API requests. Eligible integrations through certain agent tools are temporarily exempt from these charges.\"},{\"q\":\"Is DeepSeek V4.1 Flash 1017 open source?\",\"a\":\"DeepSeek has publicly released the model weights on Hugging Face and ModelScope. Developers should review the applicable licensing terms before using or redistributing them. Its proprietary web search infrastructure is not part of that model-weight release.\"}]","sources":"[{\"title\":\"DeepSeek V4.1 Flash 1017 official release\",\"url\":\"https://api-docs.deepseek.com/news/news261017\"},{\"title\":\"DeepSeek V4.1 Flash 1017 on Hugging Face\",\"url\":\"https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/086832e75052ff4af4d3e01a309dbaec.webp","featured_image_alt":"DeepSeek","featured_image_width":1600,"featured_image_height":893,"intro_image_path":"/api/uploads/2026/10/086832e75052ff4af4d3e01a309dbaec.webp","intro_image_width":1600,"intro_image_height":893,"seo_title":"DeepSeek Upgrades V4.1 Flash and Opens Search API","meta_description":"DeepSeek releases V4.1 Flash 1017 with improved coding performance and opens Search and Fetch APIs priced at $0.75 per 1,000 requests.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":170,"title":"North Macedonia's AI Future Starts With Infrastructure","slug":"north-macedonias-ai-future-starts-with-infrastructure","short_description":null,"content":"Vladimir Stojkovski of GoNet Communications discusses the networks, data centres and technical investment needed to support artificial intelligence in North Macedonia.
\n![\"Vladimir](\"/api/uploads/2026/10/b45ab497cd37c5785343c3b983df3260.webp\")
\nArtificial intelligence may be transforming how businesses operate, but its growth depends on infrastructure that rarely receives the same attention as the technology itself. Reliable networks, cloud services, computing capacity and secure data connections all play a role in making AI systems accessible and dependable.
\nFor North Macedonia, the challenge is not simply adopting more AI applications. It is also developing the infrastructure needed to support them, particularly as businesses become more dependent on cloud platforms and data-intensive services.
\nIn an interview with AI Balkans, Vladimir Stojkovski of GoNet Communications, a Skopje-based telecommunications provider, discusses the country's readiness for local AI infrastructure, the growing importance of data sovereignty, and the changing role of telecommunications companies.
\nHe argues that North Macedonia has opportunities to develop specialised cloud and AI services without immediately attempting to compete with the world's largest data centre markets. The more realistic starting point, he says, is supporting local businesses with practical AI applications and infrastructure suited to their needs.
\n![\"Server](\"/api/uploads/2026/10/e1f2e441c1346ed99fb2a9421172e223.webp\")
![\"Server](\"/api/uploads/2026/10/55e0dcdad45d8ce059ccb26db475e4da.webp\")
GoNet is a telecommunications company focused on providing reliable and modern communication solutions for the business sector.
\nWith our own fibre-optic infrastructure and IP network, as well as connections to international telecommunications operators, we can provide high-quality internet and network services while maintaining a high degree of control over our infrastructure.
\nOur portfolio includes dedicated internet access, fibre-optic connectivity between business locations, private networks, VPN solutions, VoIP telephony, cloud services, server infrastructure, network security and monitoring systems.
\nOur goal is to be a long-term technology partner for businesses, providing solutions tailored to their actual needs rather than simply supplying an internet connection.
\nWhen discussing artificial intelligence, most attention goes to software and what it can do. But behind every AI application is substantial infrastructure.
\nAI systems depend on transferring large volumes of data, accessing server resources and maintaining stable communication between users, applications and data centres.
\nWithout reliable telecommunications infrastructure, even the most advanced AI solutions cannot reach their full potential.
\nI believe telecommunications operators therefore have an important role in AI development. We provide the connectivity that supports an increasing share of the modern digital economy.
\nThe main advantages of fibre-optic infrastructure are high capacity, stability and low latency.
\nAt GoNet, we develop network solutions that enable reliable communication between multiple business locations, server systems and cloud platforms.
\nThrough dedicated internet connections and private Layer 2 and VPN connectivity, clients can obtain infrastructure tailored to their requirements rather than relying exclusively on standard internet packages.
\nContinuous network monitoring is particularly important. It allows us to identify problems promptly and maintain service quality.
\nAs AI and cloud technologies grow, I expect these solutions to become increasingly important for businesses.
\nWe can already see a change, although I would not say that every new requirement is directly related to artificial intelligence.
\nCompanies are increasingly moving business processes to cloud platforms, using centralised applications and exchanging larger volumes of data between locations.
\nThis makes stability, symmetrical internet capacity, backup connections and communication security more important.
\nIn the past, the main question for many companies was how fast their internet connection was. Today, discussions increasingly focus on the overall availability and reliability of their infrastructure.
\nI expect AI adoption to accelerate this trend further.
\nNorth Macedonia has a foundation for further development of cloud and data centre services, but we need to distinguish between conventional hosting and infrastructure designed for advanced AI systems.
\nThe local market already offers options for standard business applications, virtual servers and private cloud solutions.
\nTraining large AI models, however, requires considerably more specialised infrastructure, particularly GPU servers, advanced cooling systems, reliable electricity supplies and substantial network capacity.
\nI believe the realistic starting point for the Macedonian market is developing local AI solutions for business applications, including the use of already-trained models, automation and data processing.
\nWe do not need to immediately build infrastructure for the world's largest AI models to create real value for the domestic economy.
\nAbsolutely. I believe local infrastructure will become increasingly important, particularly for companies and institutions working with sensitive or business-critical data.
\nInternational cloud platforms offer considerable advantages, but there are situations where storing and processing data locally provides greater control, makes certain regulatory requirements easier to meet and reduces dependence on external systems.
\nIt is important to emphasise that local hosting does not automatically guarantee security. Appropriate technical and organisational measures are still necessary.
\nI believe hybrid models will become dominant, with some infrastructure maintained locally and other services provided through international cloud platforms.
\nThis creates significant opportunities for domestic telecommunications and IT companies.
\nThe challenges are complex and cannot be reduced to a single factor.
\nInvestment is needed in modern data centres, energy capacity, network infrastructure and specialised server equipment.
\nAccess to qualified professionals is another particularly important factor because developing and maintaining these systems requires expertise from several different fields.
\nFrom a telecommunications perspective, North Macedonia should continue investing in fibre-optic networks, international connectivity and infrastructure redundancy.
\nBut economic viability matters alongside technology. Major investments require clear market demand and a long-term strategy.
\nAs digitalisation and AI adoption increase, security is becoming an inseparable part of telecommunications services.
\nProviding an internet connection is no longer enough. Business customers expect secure infrastructure, continuous monitoring and the ability to respond promptly to security incidents.
\nAt GoNet, we focus on protecting network infrastructure, controlling access, segmenting networks and monitoring traffic.
\nWith businesses becoming increasingly dependent on digital systems, even a short interruption can have serious operational consequences.
\nI believe telecommunications operators will gradually assume a greater role in their customers' overall digital security.
\nIn our daily operations, we already use automated systems for network monitoring, service availability tracking and timely alerts when problems arise.
\nThese systems do not necessarily use artificial intelligence, but they provide a good foundation for introducing it.
\nWe are particularly interested in the potential of AI to analyse network traffic, detect unusual activity and predict problems before they affect customers.
\nWe also see opportunities to automate technical support and optimise operational processes.
\nOur approach is practical. Technology should deliver measurable improvements in service quality rather than being introduced simply because it is popular.
\nI believe the traditional internet service provider model will gradually change.
\nInternet access will remain a core service, but businesses will expect much more from their telecommunications partners.
\nCloud infrastructure, cybersecurity, managed network services, private connectivity and integration with AI systems will become increasingly important.
\nTelecommunications companies that invest in their own infrastructure, technical expertise and high-quality support will have a significant advantage.
\nIn my view, the future is about providing comprehensive infrastructure solutions that enable businesses to grow, rather than simply selling faster internet connections.
\nI believe North Macedonia has potential, particularly for regional cloud services and specialised data centre operations.
\nOur geographic position and connectivity opportunities with neighbouring markets provide a good foundation, but they are not sufficient on their own.
\nAttracting larger investments requires competitive and predictable electricity costs, reliable power supplies, high-quality telecommunications connectivity and a stable business environment.
\nGreater cooperation between the private sector, institutions and the education system is also needed.
\nI see the realistic opportunity in gradually developing specialised services rather than attempting to compete immediately with the world's largest data centre markets.
\nOur priority remains the continuous expansion and modernisation of our fibre-optic and IP infrastructure, alongside the further development of services for business customers.
\nWe are placing particular emphasis on network reliability, increasing capacity, improving redundancy and developing cloud, server and security solutions.
\nIn the AI era, I see considerable potential in bringing telecommunications and advanced IT services closer together.
\n![\"\"](\"/api/uploads/2026/10/c9a31fc0cf21a67c7b9959f9ce3090a5.webp\")
Our ambition is for GoNet to continue developing as a company that provides complete technology infrastructure for modern businesses, rather than just internet access.
I believe the future belongs to companies that can offer reliable infrastructure, flexibility and genuine technical support.
\nFor us, artificial intelligence is an additional reason to keep investing in North Macedonia's digital infrastructure.
\nThis interview was conducted by AI Balkans. Responses were provided in Macedonian and translated into English, with light editing for clarity and readability.
","featured_image_id":387,"intro_image_id":387,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":27,"author_id":7,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":8,"view_count":79,"ordering":0,"published_at":"2026-10-16 14:28:34","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-16 14:28:34","updated_at":"2026-10-16 14:28:34","key_takeaways":"[\"North Macedonia can develop practical AI services using existing models without immediately investing in infrastructure for training the world's largest AI systems.\",\"Local hosting and hybrid cloud infrastructure could become increasingly important for organisations handling sensitive information, although local storage alone does not guarantee security.\",\"The country's potential as a regional data centre location will depend on electricity costs, reliable connectivity, specialist talent and sustained investment.\"]","faq":"[{\"q\":\"Is North Macedonia ready to host AI systems locally?\",\"a\":\"According to Vladimir Stojkovski, the country already has options for conventional cloud and server hosting, but infrastructure for training large AI models requires much more specialised investment.\"},{\"q\":\"Why might businesses choose local AI hosting?\",\"a\":\"Local hosting can offer greater control over business data, reduce dependence on external providers and help organisations meet certain regulatory requirements.\"},{\"q\":\"Could North Macedonia become a regional AI infrastructure hub?\",\"a\":\"Stojkovski sees potential in specialised cloud and data centre services, provided the country can offer reliable electricity, network connectivity, qualified professionals and a stable investment environment.\"},{\"q\":\"How could AI change telecommunications companies?\",\"a\":\"AI could help telecom providers improve network monitoring, detect unusual activity, predict outages and automate aspects of technical support.\"}]","sources":"[{\"title\":\"Official Go Net Communications LinkedIn\",\"url\":\"https://www.linkedin.com/company/go-net-communications\"},{\"title\":\"Go Net Communications Official Site\",\"url\":\"https://go.net.mk/\"}]","category_name":"Interviews","category_slug":"interviews","author_name":"Aleksandra Todorovikj","author_slug":"aleksandra-todorovikj","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/4a2f094d1022e341ac22ece7944d44c6.webp","featured_image_alt":"Gonet Server Room","featured_image_width":1448,"featured_image_height":1086,"intro_image_path":"/api/uploads/2026/10/4a2f094d1022e341ac22ece7944d44c6.webp","intro_image_width":1448,"intro_image_height":1086,"seo_title":"North Macedonia's AI Future Starts With Infrastructure","meta_description":"GoNet's Vladimir Stojkovski discusses AI infrastructure, local data centres, cybersecurity and North Macedonia's digital future.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":169,"title":"Mistral Releases Timestral for AI-Powered Forecasting","slug":"mistral-releases-timestral-for-ai-powered-forecasting","short_description":null,"content":"French AI company Mistral has released Timestral, a new family of artificial intelligence models designed to predict future trends from historical data, targeting applications in energy, finance, manufacturing and supply chain management.
\nAnnounced on October 15, 2026, Timestral introduces four models ranging from 5 million to 209 million parameters. Unlike general-purpose AI systems built for conversation, reasoning or content generation, these models specialise in time-series forecasting, analysing patterns in numerical data to estimate what may happen next.
\nAccording to Mistral, the largest Timestral model achieved the highest results among pretrained, zero-shot forecasting models across three major public benchmarks: GIFT-Eval, TIME and fev-bench.
\nThe company has released the model weights alongside an inference implementation, allowing developers and organisations to run the models on their own infrastructure, including standard CPUs.
\nTime-series forecasting is widely used in industries where decisions depend on understanding how values change over time.
\nElectricity providers forecast demand to balance production and consumption. Manufacturers analyse sensor readings to anticipate equipment failures. Retailers estimate future sales to manage inventory, while financial institutions use historical data to assess cash flows, trading volumes and market volatility.
\nThese tasks have traditionally relied on statistical methods or machine learning systems trained for specific datasets.
\nTimestral follows a different approach. As a time-series foundation model, it is pretrained on a broad collection of patterns and can generate predictions for previously unseen datasets without requiring task-specific training.
\nThis is known as zero-shot forecasting.
\nMistral says Timestral can analyse multiple related variables simultaneously, rather than treating each series independently. It also supports additional information that may influence future outcomes, such as known schedules, calendar events or other relevant measurements.
\nFor example, an electricity-demand model could consider historical consumption alongside temperature readings and information about upcoming holidays.
\nInstead of producing only a single prediction, Timestral generates probabilistic forecasts. These express a range of possible outcomes, giving organisations more information about uncertainty when planning operations.
\nMistral has released Timestral in four configurations.
\n| Model | \nParameters | \nMain advantage | \n
|---|---|---|
| Timestral-5M | \n5 million | \nLightweight, fast inference | \n
| Timestral-28M | \n28 million | \nLower latency with greater capacity | \n
| Timestral-123M | \n123 million | \nBalance of accuracy and efficiency | \n
| Timestral-209M | \n209 million | \nHighest reported forecasting accuracy | \n
All four models support historical input sequences of up to 32,768 time steps and generate probabilistic forecasts across 21 quantiles.
\nDespite their different sizes, they share the same underlying architecture and training approach.
\nMistral reports that the smallest model can complete a representative forecasting task in approximately 10 milliseconds on an eight-thread server CPU. The largest completes the same task in approximately 140 milliseconds.
\nThese figures come from the company's own testing and will vary depending on hardware, dataset size and forecasting requirements.
\nThe ability to operate without dedicated graphics processors could make Timestral attractive for businesses that need forecasting capabilities without investing in expensive AI computing infrastructure.
\nMistral evaluated Timestral against existing forecasting systems using three established public benchmarks.
\n![\"Mistral](\"/api/uploads/2026/10/3652e4909604b0165837e95a5658d8f6.webp\")
GIFT-Eval measures forecasting performance across datasets from different domains. TIME evaluates predictions across a broad collection of forecasting tasks, while fev-bench includes additional real-world forecasting scenarios and supporting variables.
\nAccording to the company's published results, Timestral-209M ranked first among pretrained and zero-shot models across all three benchmarks.
\nMistral also reports that its 28-million-parameter model outperformed the substantially larger Chronos-2 model on GIFT-Eval and TIME, while Timestral-209M achieved stronger results than TimesFM-3 and Falcon-2.0 using fewer parameters.
\nThe results suggest that model size alone does not determine forecasting quality.
\nHowever, the comparisons apply to the evaluated models and benchmark configurations. They do not establish that Timestral will outperform every specialised forecasting system or conventional statistical method in production.
\nMistral has submitted results to public leaderboards and published model information to support reproducibility. A full technical report was still forthcoming at the time of the announcement.
\nA notable aspect of Timestral is its training process.
\nMistral says most of the training data was synthetically generated rather than collected directly from real-world systems.
\nThe company created artificial time-series data containing trends, seasonal patterns, noise and relationships between variables.
\nTimestral was initially pretrained on synthetic data before undergoing a second training stage combining synthetic and real-world datasets.
\nThis approach allows developers to create large and varied training collections without depending entirely on historical records from individual companies or industries.
\nIt can also help expose models to patterns that may be uncommon in available real-world datasets.
\nNevertheless, synthetic training data cannot guarantee reliable performance when actual market conditions, industrial processes or customer behaviour differ significantly from the patterns represented during training.
\nMistral has made all four Timestral model weights available through Hugging Face under a Modified MIT licence.
\nThe company has also published a GitHub repository containing its inference implementation, released separately under Apache 2.0.
\nDevelopers can download the models and run forecasts using Python and PyTorch, either on CPUs or supported GPUs.
\nFor organisations working with commercially sensitive information, Timestral supports deployment within private infrastructure, including on-premises systems and isolated environments.
\nMistral is additionally offering model adaptation through Mistral Forge, its enterprise training platform.
\nThis enables organisations to fine-tune Timestral using proprietary datasets, adapt its architecture to industry-specific variables and update models as forecasting conditions change.
\nSuch capabilities may be particularly important in sectors where operational data cannot easily be transferred to external cloud services.
\nThe release places Mistral in a growing field of companies developing foundation models for specialised data rather than focusing exclusively on language and multimodal AI.
\nForecasting models are less visible to consumers than conversational assistants, but their performance can have direct financial and operational consequences.
\nAn inaccurate prediction of electricity demand can affect energy costs. Poor inventory forecasts can increase storage expenses or create shortages. In industrial environments, failures to anticipate changes in sensor measurements can contribute to equipment downtime.
\nThese applications require more than strong benchmark results. Forecasting systems must be tested against local datasets, evaluated under changing conditions and monitored after deployment.
\nTimestral's combination of relatively small model sizes, probabilistic predictions and private deployment options gives developers another approach to these problems.
\nFor now, the central question is how well the models perform outside public benchmarks, particularly in industries where data quality, unusual events and rapidly changing conditions can undermine historical patterns.
\nMistral says a detailed technical report will follow its initial release, providing further information about the models' training and evaluation.
","featured_image_id":385,"intro_image_id":385,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":6,"view_count":37,"ordering":0,"published_at":"2026-10-16 08:21:24","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-16 08:21:24","updated_at":"2026-10-16 08:21:24","key_takeaways":"[\"Mistral released Timestral on October 15, 2026, introducing four AI models designed specifically for time-series forecasting, ranging from 5 million to 209 million parameters.\",\"The company reports that Timestral-209M leads three major public forecasting benchmarks among pretrained, zero-shot models, while requiring fewer parameters than several competing systems.\",\"Timestral supports multivariate and probabilistic forecasting, can operate on standard CPUs and is available for private deployment using downloadable model weights under a Modified MIT licence.\",\"The models target practical applications including energy-demand prediction, financial forecasting, manufacturing operations and supply chain planning.\"]","faq":"[{\"q\":\"What is Mistral Timestral?\",\"a\":\"Timestral is a family of AI foundation models designed for time-series forecasting. It analyses historical numerical data to predict future values and provide estimates of uncertainty.\"},{\"q\":\"What makes Timestral different from ChatGPT or other language models?\",\"a\":\"Timestral is designed for forecasting numerical sequences rather than generating conversational responses, documents or code.\"},{\"q\":\"Is Timestral open source?\",\"a\":\"Mistral has released the inference code under Apache 2.0 and the downloadable model weights under a Modified MIT licence. The model-weight licence includes additional conditions.\"},{\"q\":\"Can Timestral run without a GPU?\",\"a\":\"Yes. Mistral provides CPU-compatible inference for all four models, although actual performance depends on the hardware and forecasting workload.\"},{\"q\":\"How accurate is Timestral?\",\"a\":\"Mistral reports leading results across GIFT-Eval, TIME and fev-bench among pretrained, zero-shot models. Accuracy on specific business datasets requires separate testing.\"},{\"q\":\"Where can developers download Timestral?\",\"a\":\"The model weights are available through Mistral's official Hugging Face collection, and the inference implementation is published on GitHub.\"}]","sources":"[{\"title\":\"Mistral official Timestral announcement, October 15, 2026\",\"url\":\"https://github.com/mistralai/timestral\"},{\"title\":\"Timestral model weights on Hugging Face\",\"url\":\"https://huggingface.co/collections/mistralai/timestral\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/9759ee6814b7438348be43e3b1fe3f8d.webp","featured_image_alt":"Timestral","featured_image_width":1200,"featured_image_height":716,"intro_image_path":"/api/uploads/2026/10/9759ee6814b7438348be43e3b1fe3f8d.webp","intro_image_width":1200,"intro_image_height":716,"seo_title":"Mistral Releases Timestral for AI Forecasting","meta_description":"Mistral launches Timestral, four lightweight AI forecasting models with open weights, CPU support and leading reported benchmark performance.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":168,"title":"Figma Expands Into AI-Powered Software Development","slug":"figma-expands-into-ai-powered-software-development","short_description":null,"content":"Figma is bringing working code, AI agents and backend infrastructure into its design platform, introducing a series of features that could change how teams move from product concepts to functioning applications.
\nThe company announced the updates on October 15 at Config India 2026 in Bengaluru, unveiling an expanded development workflow built around code layers, Figma Specs, integration with the Payload backend framework and tighter connections between design systems and AI coding agents.
\nThe announcements extend Figma's ambitions beyond interface design. Its platform is increasingly positioned as a place where designers, developers and product teams can collaborate on working software without moving through entirely separate design, prototyping and development environments.
\nThe timing reflects a broader shift in software development, as AI coding assistants make it easier to generate applications from written instructions and visual references.
\nOne of the most significant announcements is the expansion of code layers, a feature that allows users to bring functioning code directly onto the Figma canvas.
\nInstead of working exclusively with static screens or simulated interactions, designers can now create and modify interactive prototypes backed by executable code.
\nFigma supports several ways to begin. Users can convert existing design frames into code-backed experiences, import a connected GitHub repository, capture a live website through its Chrome extension or describe an application to an AI agent.
\nCode layers were introduced earlier in 2026 and are now entering an open beta across Figma's plans.
\nThe distinction matters because traditional design prototypes often approximate how software should behave without implementing its actual functionality. Code-backed prototypes allow teams to test interactions and make changes closer to the development process.
\nHowever, this does not mean Figma can automatically turn any design into a production-ready application. The company says support for increasingly complex codebases remains an area of future development.
\nThe company also introduced Figma Specs, a Markdown-based environment for capturing product requirements, technical decisions and design context.
\nSpecs can exist as standalone documents or alongside visual designs on the canvas. Teams can connect specifications to existing Figma elements and use AI agents to help develop or revise them.
\nThe feature addresses a growing challenge in AI-assisted software development: providing coding agents with reliable information about what a product is supposed to do.
\n![\"Figma](\"/api/uploads/2026/10/6ad044dbadde461b407c4b71793de5f1.webp\")
An AI agent may generate a functional interface from a prompt, but the result can diverge from established design systems, technical constraints or product requirements if those instructions are incomplete.
\nWith Specs, Figma wants written requirements, visual references and technical context to remain connected throughout development.
\nThe system also supports importing existing Markdown documents, allowing teams to bring documentation into the design environment rather than recreating it.
\nSpecs is entering open beta on eligible paid plans, with access rolling out gradually.
\nAnother major development is the integration of Payload into Figma Make.
\nPayload is an open-source content management system and application framework whose team joined Figma in 2025.
\nWith the integration, Figma's AI-assisted application builder can generate a backend that includes a database, content management, file storage, authentication, access controls and APIs.
\nThese are essential components of many real-world applications.
\nA business application, for example, may require user accounts, stored records, administrative tools and secure access to information. Visual design tools alone cannot provide those capabilities.
\nFigma says Payload can now supply that infrastructure directly within its development workflow, with hosted services available through Figma and self-hosting remaining an option for developers.
\nThe integration is rolling out across Figma Make plans.
\nWhile this moves Figma closer to full-stack application development, production deployment still requires consideration of security, reliability, maintenance and operational costs. Generating a backend is not equivalent to independently verifying that an application is ready for public use.
\nFigma is also expanding the information available to AI agents working with design systems.
\nNew capabilities allow organizations to attach guidelines to published design libraries and connect code sources from GitHub repositories or npm packages.
\nThe objective is to help agents generate components that follow existing patterns rather than repeatedly inventing new ones.
\nFigma has also announced improvements to its Model Context Protocol (MCP) integrations, extending how external AI tools interact with the platform.
\nThis could be particularly relevant for organizations already using coding agents alongside established design systems. The more accurately an agent can interpret existing components, implementation rules and technical context, the less manual correction may be required.
\nSome capabilities remain in limited testing. Code-source integration is entering closed beta, while centralized design-system management is currently a preview.
\nThe announcements arrive as the distinction between design applications, AI coding tools and low-code development platforms becomes less clear.
\nAI-assisted development has made it possible to generate interfaces and application prototypes with relatively little conventional programming. Tools built around this approach are increasingly competing for users who previously depended on separate design and development processes.
\nFigma is taking a different route from standalone prompt-to-app services.
\nRather than treating design as an input that is handed off to a coding system, it is attempting to keep the product's visual design, technical specifications, working code and collaboration history within a connected environment.
\nThat could help teams maintain consistency as applications evolve. It could also make Figma a more direct competitor to AI application builders and parts of the traditional software development toolchain.
\nThere are important limits to that comparison. Professional software engineering involves testing, infrastructure management, security reviews, integration work and ongoing maintenance. Figma's announcements do not remove those responsibilities.
\nNor has the company demonstrated that its new workflow is more reliable or economical than established alternatives across complex production projects.
\nFigma's latest releases suggest that the company sees AI-assisted software development as a central part of its future, rather than an additional feature for designers.
\nThe most consequential change may be the connection between information that has historically lived in different places: product documentation, interface designs, source code and backend systems.
\nFor development teams, the practical question is whether bringing these elements together produces better software with fewer handoffs and less rework.
\nSeveral of the newly announced capabilities are still in beta, and the company is introducing usage-based AI credits for code-layer tasks after a promotional period ending November 4, 2026.
\nThe next measure of Figma's progress will be how these tools perform in real projects, particularly when teams move beyond prototypes into software that must be maintained, secured and updated over time.
","featured_image_id":383,"intro_image_id":383,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":6,"view_count":20,"ordering":0,"published_at":"2026-10-16 08:06:51","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-16 08:06:51","updated_at":"2026-10-16 08:06:51","key_takeaways":"[\"Figma has introduced expanded AI-assisted development capabilities, bringing working code, technical specifications and backend infrastructure closer to its design environment.\",\"Code layers and Figma Specs are entering open beta, while Payload integration adds database, authentication, content management and API functionality to Figma Make.\",\"The announcements position Figma more directly against AI application builders, although complex production software still requires engineering, security and maintenance work.\"]","faq":"[{\"q\":\"What did Figma announce in October 2026?\",\"a\":\"Figma announced code layers entering open beta, the introduction of Figma Specs, native Payload backend integration, additional design-system capabilities and expanded AI-agent connectivity at Config India 2026.\"},{\"q\":\"Can Figma now build complete applications?\",\"a\":\"Figma Make can generate applications with working backend functionality through Payload. However, production readiness depends on the application's complexity, security, testing and deployment requirements.\"},{\"q\":\"What are Figma code layers?\",\"a\":\"Code layers allow developers and designers to create, import and modify interactive prototypes backed by real code directly on the Figma canvas.\"},{\"q\":\"What is Figma Specs?\",\"a\":\"Figma Specs is a Markdown-based environment that connects product requirements, technical documentation and visual design context for teams and AI agents.\"},{\"q\":\"Are the new Figma features free?\",\"a\":\"Availability varies. Code layers are rolling out across plans, with a temporary promotional period for generating and editing code on paid plans through November 4, 2026. Specs is available on eligible paid and institutional plans. Usage-based AI credits apply to certain features.\"},{\"q\":\"Does Figma replace traditional development tools?\",\"a\":\"No. The new capabilities reduce some handoffs between design and development, but they do not eliminate the need for programming expertise, software testing, infrastructure management or security reviews.\"}]","sources":"[{\"title\":\"Figma's full announcement by CEO Dylan Field\",\"url\":\"https://www.figma.com/blog/build-by-design/\"},{\"title\":\"Introducing Figma Specs\",\"url\":\"https://www.figma.com/blog/websites-with-payload-in-figma/\"},{\"title\":\"Payload integration and backend capabilities\",\"url\":\"https://www.figma.com/blog/websites-with-payload-in-figma/\"},{\"title\":\"Official availability, limitations and pricing details\",\"url\":\"https://help.figma.com/hc/en-us/articles/43663368947863-What-s-new-from-Config-India-2026\"},{\"title\":\"Figma's official release notes\",\"url\":\"https://www.figma.com/release-notes/\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/a03f2c4b575d120113ca3b050ebfc4da.webp","featured_image_alt":"Figma AI","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/a03f2c4b575d120113ca3b050ebfc4da.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Figma Expands Into AI-Powered Software Development","meta_description":"Figma introduces code layers, Specs and Payload integration, bringing AI agents and backend development into its design platform.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":167,"title":"Google Releases Gemini 4 to Challenge AI Rivals","slug":"google-releases-gemini-4-to-challenge-ai-rivals","short_description":null,"content":"Google has expanded access to Gemini 4 Argon, its newest frontier artificial intelligence model, bringing advanced reasoning, coding and professional task automation capabilities to developers, businesses and paying consumers.
\nThe wider release on October 13, 2026, follows an initial announcement on September 30, when Google limited access to selected cybersecurity professionals. Gemini 4 Argon is now available through Google's developer platforms, enterprise services, Gemini app and AI-powered Search features.
\nThe launch puts Google in more direct competition with OpenAI and Anthropic for customers building software, automating business processes and deploying AI systems capable of completing complex tasks.
\nAccording to Google DeepMind, the new model is designed to maintain reasoning across extended workflows rather than simply produce answers to individual questions. Its targeted applications include software engineering, financial analysis, legal research, cybersecurity and scientific work.
\nGoogle has published benchmark comparisons showing Gemini 4 Argon outperforming several competing models in selected professional tasks. The results are substantial, although they do not establish a universal lead.
\n![\"Gemini](\"/api/uploads/2026/10/055472bdf47319dc282c611cff672e3d.webp\")
On the Vals Index, which evaluates performance across knowledge-intensive professional work, Gemini 4 Argon achieved 68.9%, compared with 63.1% for GPT-6 Astra, 65.8% for Claude Fable 5.1 and 67.0% for Claude Opus 5.5.
\nThe model also performed strongly on DeepSWE v1.1, a software engineering benchmark, scoring 77.9%. GPT-6 Astra achieved 74.1%, while Claude Opus 5.5 scored 74.2%.
\nThe ranking changes on other tests.
\nOn FrontierSWE v2, GPT-6 Astra led with 65.5%, followed by Claude Opus 5.5 at 62.3%. Gemini 4 Argon scored 55.0%. Anthropic's Claude Opus 5.5 also retained an advantage on Terminal-bench 4.0, scoring 66.4% against Gemini's 57.4%.
\nScientific task evaluations produced similarly mixed results. GPT-6 Astra outperformed Gemini on Terminal-Bench Science 0.1, while Gemini led on LABBench 2 and RiemannBench.
\nThese figures come from Google's published model card and associated evaluation methodology. They provide useful comparisons under specific testing conditions, but should not be interpreted as independent confirmation that Gemini performs better across every application.
\nFor developers and enterprise customers, the practical question is which model performs reliably on their particular tasks, at an acceptable speed and cost.
\nGemini 4 Argon supports text, images, audio and video inputs within a combined context window of up to one million tokens.
\n![\"Gemini](\"/api/uploads/2026/10/f29c36231167869c99dd9faa37d5cddd.webp\")
That capacity allows applications to process extensive documents, software repositories and other large collections of information. It does not, by itself, guarantee that the model will correctly interpret every detail in a long input.
\nGoogle is emphasizing Argon's ability to handle multi-step activities, including code migration, software maintenance and technical investigation.
\nThe model uses a sparse mixture-of-experts architecture and was trained using Google's Tensor Processing Units. Its design is intended to improve performance on difficult reasoning tasks without requiring every component of the model to operate on each request.
\nGoogle is also integrating the model into its broader ecosystem, including its Antigravity development platform and Gemini's task-oriented features.
\nFor software developers, this expands the possibility of assigning larger engineering activities to AI systems rather than requesting isolated code snippets. For organizations, it creates additional options for automating research, document analysis and business processes.
\nHowever, increased autonomy introduces new operational risks. Systems that can execute multiple actions require careful permissions, monitoring and verification, particularly when handling sensitive information or modifying production software.
\nPricing is another important part of the launch.
\nGemini 4 Argon is available at an introductory API rate of $2 per million input tokens and $10 per million output tokens until November 30, 2026.
\nGoogle says standard pricing will increase to $4 per million input tokens and $20 per million output tokens after that date. Cached input tokens receive a 95% discount during the introductory period.
\nThe temporary reduction gives developers an incentive to experiment with Argon and compare its performance against systems already used in their applications.
\nThe eventual cost will depend on the complexity of individual tasks, the amount of generated output, repeated model calls and other operational requirements.
\nA lower advertised token price does not necessarily translate into a cheaper completed workflow. An AI system that requires more processing time, repeated attempts or extensive verification may cost more in practice.
\nGoogle's control over its own AI infrastructure, developer services and consumer distribution channels nevertheless gives the company considerable influence over how the model reaches customers.
\nGoogle says Gemini 4 Argon is available globally to developers and enterprise customers through the Gemini API, Google AI Studio and relevant business services.
\nConsumer access is more restricted. The model is available to Google AI Pro and Ultra subscribers through the Gemini app and AI Mode in Search.
\nGoogle AI Ultra subscribers can also use Argon through the Antigravity development platform.
\nDevelopers using the API require paid access, while availability and usage limits depend on the service and subscription tier.
\nThe release is therefore broader than the initial restricted rollout, but it does not amount to unrestricted free access for all users.
\nGoogle's model card acknowledges that Gemini 4 Argon can still produce incorrect information and may experience slow responses or timeouts. Its stated knowledge cutoff is March 2026.
\nThe company also identifies more serious risks associated with the model's advanced capabilities.
\nAccording to its published safety assessment, Google could not rule out that Argon had reached certain critical capability thresholds relating to biological information and offensive cybersecurity. It also identified concerns involving situational awareness and the potential to circumvent oversight.
\nGoogle says it has implemented safeguards intended to reduce those risks before wider deployment.
\nThat distinction matters. The existence of safety controls does not eliminate the underlying capabilities or guarantee that every possible misuse has been prevented.
\nGoogle's decision to initially provide restricted access to vetted cybersecurity defenders illustrates the difficulties AI developers face as increasingly powerful systems become commercially available.
\nGemini 4 Argon's release strengthens Google's position in a market where model providers increasingly compete on their ability to complete useful work.
\nOpenAI, Anthropic and Google are developing systems that can write and maintain software, interact with tools, analyze large volumes of information and carry out activities across multiple applications.
\nThe competition is no longer determined solely by which model provides the best response to a question.
\nCustomers must also consider reliability, integration, security controls, execution speed and the total cost of completing a task.
\nGoogle enters this stage with substantial advantages in distribution. Its AI models can be integrated into products already used by consumers, developers and enterprises, from Search and Gemini to cloud infrastructure and software development tools.
\nIts latest benchmark results suggest that Gemini 4 Argon is competitive with leading models in several demanding areas. They also show that OpenAI and Anthropic retain advantages on other evaluations.
\nThe next test will come from sustained real-world use, independent evaluations and the response of competing model providers.
\nFor now, Google's wider release gives developers and businesses another serious option in a market where leadership increasingly depends on practical results rather than a single benchmark ranking.
","featured_image_id":379,"intro_image_id":379,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":6,"view_count":29,"ordering":0,"published_at":"2026-10-14 16:57:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-14 16:57:49","updated_at":"2026-10-14 17:03:59","key_takeaways":"[\"Gemini 4 Argon became broadly available on October 13, 2026, following its initial restricted announcement on September 30.\",\"Google's benchmarks place Gemini 4 ahead of GPT-6 Astra and several Claude models in professional knowledge work and selected coding tasks. Competing models remain ahead in other evaluations.\",\"Developers pay $2 per million input tokens and $10 per million output tokens through November 30, before standard pricing doubles.\",\"Google's own assessment identifies risks involving cybersecurity, biological information and model oversight, alongside safeguards intended to reduce them.\"]","faq":"[{\"q\":\"When was Google Gemini 4 released?\",\"a\":\"Google first announced Gemini 4 Argon on September 30, 2026, with restricted access. Broader availability began on October 13, 2026.\"},{\"q\":\"Is Gemini 4 better than GPT-6 and Claude?\",\"a\":\"Gemini 4 Argon leads several benchmarks published by Google, including the Vals Index and DeepSWE v1.1. However, GPT-6 Astra and Claude Opus 5.5 outperform it on other tests. No single model leads across every category.\"},{\"q\":\"Is Gemini 4 available for free?\",\"a\":\"The initial rollout requires eligible paid Google AI subscriptions or paid API access. Google has not announced unrestricted free access to Argon for all users.\"},{\"q\":\"How much does Gemini 4 cost?\",\"a\":\"Introductory API pricing is $2 per million input tokens and $10 per million output tokens through November 30, 2026. Standard pricing then increases to $4 and $20 respectively.\"},{\"q\":\"What is Gemini 4's context window?\",\"a\":\"Gemini 4 Argon supports a combined input-output context window of up to one million tokens, including text, images, audio and video inputs.\"},{\"q\":\"Can Gemini 4 make mistakes?\",\"a\":\"Yes. Google's model card acknowledges hallucinations, occasional slow responses and timeouts. Its stated knowledge cutoff is March 2026.\"}]","sources":"[{\"title\":\"Google's official launch announcement\",\"url\":\"https://blog.google/intl/en-in/products/introducing-gemini-4-argon/\"},{\"title\":\"Google DeepMind Gemini 4 Argon model card\",\"url\":\"https://deepmind.google/models/model-cards/gemini-4-argon/\"},{\"title\":\"Google's Gemini 4 safety and security report\",\"url\":\"https://blog.google/innovation-and-ai/technology/safety-security/gemini-4-model-safety-security/\"},{\"title\":\"Original Gemini 4 announcement\",\"url\":\"https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/b5345c91845d4644e95cae0ed5c0d2ef.webp","featured_image_alt":"Google Argon","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/b5345c91845d4644e95cae0ed5c0d2ef.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Google Releases Gemini 4 to Challenge AI Rivals","meta_description":"Google releases Gemini 4 Argon with advanced coding, reasoning and agent capabilities. See its benchmarks against GPT-6 and Claude, pricing and access.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":166,"title":"AllWeb Invites AI Agents to Speak at Tirana Conference","slug":"allweb-invites-ai-agents-to-speak-at-tirana-conference","short_description":null,"content":"The regional digital conference is opening its stage to AI-powered agents, inviting developers and startups to propose presentations for its November event.
\nAllWeb Digital Conf is inviting artificial intelligence agents to take the stage as conference speakers for the first time, opening a call for technology companies, startups and creative businesses ahead of its 6 November 2026 event in Tirana, Albania.
\nThe initiative, announced on 12 October, will allow selected AI agents to deliver presentations to a live business audience. Rather than showcase products, the organisers want participating agents to share practical knowledge and examples that professionals can apply in their work.
\nIt is an unusual experiment for a conference traditionally built around human speakers. The question AllWeb wants to explore is whether an AI agent can deliver a session that is genuinely useful, engaging and relevant to an audience of business and technology professionals.
\nThe call is open to teams developing AI agents capable of presenting ideas and interacting with conference participants.
\nApplicants will be expected to explain their proposed presentation topic, provide practical case studies and describe how their agent would engage with the audience.
\nSuggested areas include digital marketing, customer experience, productivity, innovation and business growth.
\nThe organisers say proposals will be assessed based on their relevance and potential value to attendees, rather than the technical novelty of the AI system alone.
\n“For years, we’ve brought experts to the AllWeb stage to share what they know about business, marketing and technology,” said Gerti Boshnjaku, co-organiser of AllWeb Digital Conf and CEO of Business Magazine Albania.
\nBoshnjaku said the organisers now want to give AI agents the same challenge, focusing on useful knowledge and practical examples instead of technology demonstrations.
\nThe distinction matters. An AI agent capable of producing a presentation is not necessarily capable of responding reliably to questions, adapting to an audience or explaining complex subjects accurately.
\nAllowing agents onto the conference stage creates an opportunity to examine those capabilities in a setting where audiences expect more than a prepared script.
\nAllWeb Digital Conf has been part of the regional technology and business events landscape since its establishment in North Macedonia in 2012.
\nSince 2016, editions of the conference have also been organised in Albania by Business Magazine Albania, bringing together professionals working in digital marketing, communications, technology and entrepreneurship.
\nDarko Buldioski, Programme Director of AllWeb Digital Conf, described the decision to invite AI agents as speakers as a departure from the traditional conference format.
\n“Ten years from now, this might feel completely normal,” Buldioski said, reflecting on the possibility of AI agents becoming a regular part of professional events.
\nThe 2026 conference will take place under the theme “All for AI & AI for All”, reflecting the growing role of artificial intelligence in business and professional workflows.
\nAccording to the organisers, more than 350 business leaders, marketers, entrepreneurs and technology professionals are expected at the event.
\nThe conference is scheduled for 6 November 2026 at the Namazgjah Complex in Tirana.
\nFor developers and startups, the call offers an opportunity to demonstrate their AI agents beyond conventional product presentations, with an emphasis on delivering knowledge to a professional audience.
\nThe organisers have not yet announced the application deadline or direct submission link. Further details about the selection process and participation requirements are expected through the official conference website.
\nInterested companies and development teams can follow announcements at https://allweb.digital/.
\nThe practical test will come in November, when selected AI agents are expected to demonstrate whether they can meet the demands of a live conference audience.
","featured_image_id":378,"intro_image_id":378,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":9,"author_id":10,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":3,"view_count":21,"ordering":0,"published_at":"2026-10-14 08:54:06","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-14 08:54:06","updated_at":"2026-10-14 08:54:06","key_takeaways":"[\"AllWeb Digital Conf has opened a call for AI agents to deliver presentations at its 6 November 2026 conference in Tirana, Albania.\",\"The organisers are looking for agents capable of sharing useful business insights, case studies and interactive presentations rather than conventional product demonstrations.\",\"Technology companies, startups and creative businesses can propose AI agents. The official application deadline and submission link have not yet been announced.\"]","faq":"[{\"q\":\"What is the AllWeb Call for AI Agents?\",\"a\":\"It is an initiative inviting companies and developers to propose AI-powered agents as speakers at AllWeb Digital Conf 2026.\"},{\"q\":\"When and where will AllWeb 2026 take place?\",\"a\":\"The conference will take place on 6 November 2026 at the Namazgjah Complex in Tirana, Albania.\"},{\"q\":\"Who can apply to present an AI agent?\",\"a\":\"The call targets technology companies, startups and creative businesses developing AI agents capable of delivering presentations and interacting with an audience.\"},{\"q\":\"What kind of presentations are organisers looking for?\",\"a\":\"Presentations should provide practical insights and case studies related to marketing, customer experience, productivity, innovation or business growth.\"},{\"q\":\"How can interested teams apply?\",\"a\":\"The application link and deadline have not yet been announced. Interested teams should check allweb.digital for official updates.\"}]","sources":"[{\"title\":\"AllWeb Digital Conf — official conference website\",\"url\":\"https://allweb.digital/\"},{\"title\":\"Albania Tech — AllWeb Opens Its Stage to AI Agents for the First Time.\",\"url\":\"https://albaniatech.org/allweb-opens-its-stage-to-ai-agents-for-the-first-time/\"},{\"title\":\"Business Magazine Albania — AllWeb AI Agents announcemen\",\"url\":\"https://businessmag.al/allweb-hap-per-here-te-pare-skenen-per-ai-agents/\"}]","category_name":"Albania","category_slug":"albania","author_name":"Guest Contributor","author_slug":"guest-contributor","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/b458321579a93f02cc96e6777b3a2e99.webp","featured_image_alt":"AllWeb","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/b458321579a93f02cc96e6777b3a2e99.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"AllWeb Invites AI Agents to Speak at Tirana Conference","meta_description":"AllWeb invites AI agents to speak at its November 2026 conference in Tirana, opening a new opportunity for developers and startups.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":165,"title":"IMF Warns AI Boom Could Threaten Financial Stability","slug":"imf-warns-ai-boom-could-threaten-financial-stability","short_description":null,"content":"The International Monetary Fund has warned that a reversal in the artificial intelligence investment boom could trigger a stock market correction and expose weaknesses in the global financial system.
\nIn its latest Global Financial Stability Report, released on October 13 during the IMF and World Bank Annual Meetings in Bangkok, the institution identified disappointing AI investment returns as one of three major risks facing financial markets.
\n![\"IMF](\"/api/uploads/2026/10/b8464a78e10cfb523b87c69ea91ad93e.webp\")
The warning comes as technology companies commit enormous sums to data centres, computing infrastructure and advanced AI systems, with investors increasingly expecting those investments to generate substantial future profits.
The IMF is not forecasting a market crash. Its concern is that financial markets may be vulnerable if expectations surrounding AI growth prove too optimistic.
\nArtificial intelligence has helped sustain investor confidence despite economic uncertainty, geopolitical tensions and rising borrowing costs.
\nMuch of that optimism rests on the expectation that AI will deliver significant productivity gains and create profitable new markets. Technology companies are investing heavily in infrastructure to meet anticipated demand.
\nBut those expectations introduce financial risks.
\nIf companies fail to generate sufficient revenue from AI products and services, investors could reassess their valuations. A sudden loss of confidence could lead to falling share prices, tighter financing conditions and reduced investment.
\nThe IMF's concern extends beyond individual technology companies. AI-related businesses are increasingly interconnected through investments, financing arrangements and commercial partnerships, potentially allowing problems in one part of the industry to affect others.
\nThe institution has also warned that growing reliance on debt to finance AI infrastructure could make the consequences of disappointing returns more severe.
\nOne concern is the concentration of major stock market indices in a relatively small number of technology companies.
\nWhen these companies perform well, their gains can support broader market performance. The reverse is also true.
\nA substantial correction in highly valued AI-related stocks could therefore affect investors whose portfolios extend far beyond the technology sector.
\nThe IMF has previously identified elevated valuations and market concentration as vulnerabilities that could amplify financial shocks.
\nA decline in AI-related share prices would not necessarily become a financial crisis. The danger is that falling asset values could interact with existing pressures in debt markets, investment funds and financial institutions.
\nThe latest report places an AI-related correction alongside rising sovereign bond yields and disruptive cyber incidents as potential triggers of market stress.
\nThe IMF's October assessment finds that global financial markets have remained resilient despite repeated economic and geopolitical shocks.
\nHowever, that resilience does not mean the financial system is free from vulnerabilities.
\nRising government borrowing costs remain a particular concern, especially as sovereign debt issuance grows and investors demand higher returns.
\nAt the same time, financial institutions outside the traditional banking system are exposed to risks involving leverage, liquidity and less easily traded assets.
\nInsurance companies and pension funds hold investments that may become difficult to sell during periods of market stress. Private credit markets face pressure from borrowing and liquidity constraints, while highly leveraged hedge fund positions could magnify sudden price movements.
\nThese weaknesses matter because an AI-related market correction could occur alongside other financial shocks rather than in isolation.
\n![\"IMF\"](\"/api/uploads/2026/10/217bfe1477e5cec7ec93ed4628834f0e.webp\")
Despite its warnings, the IMF's assessment is not uniformly negative.
\nIts stress testing suggests that the banking sector as a whole remains resilient under a severe scenario involving AI, bond market and energy shocks.
\nNevertheless, some weaker banks could become undercapitalised, and vulnerabilities outside the banking system could intensify the effects of a downturn.
\nThe IMF is therefore urging policymakers to strengthen financial resilience before new shocks emerge.
\nIts priorities include sustainable public debt, effective financial supervision, central bank independence and stronger protection against cyber incidents.
\nArtificial intelligence continues to offer substantial opportunities for economic growth and productivity improvements.
\nThe IMF's warning does not challenge those possibilities. Instead, it draws attention to the difference between the long-term economic potential of AI and the financial expectations currently attached to it.
\nCompanies can develop valuable technologies while still struggling to justify the prices investors have assigned to their shares.
\nThe central question is whether the revenue and productivity gains generated by artificial intelligence will be sufficient to support the scale of investment already underway.
\nFor financial markets, that question is becoming increasingly important as spending on AI infrastructure expands and more companies depend on the industry's expected growth.
\nThe IMF's October report makes clear that the consequences of disappointing expectations could extend well beyond the companies building AI systems.
","featured_image_id":375,"intro_image_id":375,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":21,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":25,"ordering":0,"published_at":"2026-10-13 16:44:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-13 16:44:09","updated_at":"2026-10-13 16:46:26","key_takeaways":"[\"The IMF warns of an AI-related market correction. Disappointing investment returns or corporate profits could trigger falling technology stock prices and wider financial instability.\",\"AI investment is increasingly connected to financial risk. High market valuations, debt-financed infrastructure and interconnected technology companies could amplify a downturn.\",\"The financial system remains resilient but vulnerable. The IMF says most banks could withstand severe shocks, although weaker institutions and non-bank financial companies face greater risks.\"]","faq":"[{\"q\":\"Is the IMF predicting an AI stock market crash?\",\"a\":\"No. The IMF has identified a potential market correction as a financial stability risk if expectations surrounding AI investment or profitability are not met. It has not predicted that a crash is inevitable.\"},{\"q\":\"Why is the AI investment boom a financial concern?\",\"a\":\"Technology companies are investing heavily in AI infrastructure, sometimes using borrowed money. If future profits fail to justify those investments, valuations could fall and financial pressures could spread.\"},{\"q\":\"Could an AI market correction affect the wider economy?\",\"a\":\"Yes. Falling share prices can reduce investor wealth, restrict financing and discourage business investment. Financial interconnectedness could amplify these effects.\"},{\"q\":\"What does the IMF recommend?\",\"a\":\"The IMF recommends strengthening financial resilience, maintaining sustainable public debt, monitoring risks across banks and investment institutions, and improving cybersecurity protections.\"}]","sources":"[{\"title\":\"IMF Global Financial Stability Report announcement, October 13, 2026\",\"url\":\"https://www.imf.org/en/publications/gfsr/issues/2026/10/13/global-financial-stability-report-october-2026\"},{\"title\":\"IMF Global Financial Stability Report, October 2026\",\"url\":\"https://www.imf.org/en/publications/gfsr/issues/2026/10/13/global-financial-stability-report-october-2026\"},{\"title\":\"IMF Annual Report 2026, AI Deployment and Disruption\",\"url\":\"https://www.imf.org/en/publications/gfsr/issues/2026/10/13/global-financial-stability-report-october-2026\"}]","category_name":"Business & Economy","category_slug":"business-economy","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/9ab96bf686067ba017d05b13756e9af1.webp","featured_image_alt":"International Monetary Fund","featured_image_width":1600,"featured_image_height":1068,"intro_image_path":"/api/uploads/2026/10/9ab96bf686067ba017d05b13756e9af1.webp","intro_image_width":1600,"intro_image_height":1068,"seo_title":"IMF Warns AI Boom Could Threaten Financial Stability","meta_description":"The IMF warns that an AI investment correction could threaten global financial stability as technology valuations and borrowing risks increase.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":164,"title":"UK Uses AI to Find 1,352 Cybersecurity Flaws","slug":"uk-uses-ai-to-find-1-352-cybersecurity-flaws","short_description":null,"content":"The United Kingdom is expanding its use of artificial intelligence to protect government systems after two pilot programmes identified and addressed 1,352 cybersecurity vulnerabilities across 548 code repositories in 18 government departments.
\nThe findings were announced on October 12, 2026, as British authorities introduced further measures to strengthen national cyber defences and prepare for security threats involving increasingly autonomous AI systems.
\nThe initiative represents a practical application of advanced AI in government cybersecurity, demonstrating how AI models can help identify software weaknesses while raising new questions about the security of autonomous systems themselves.
\nThe UK's Government Cyber Unit has been testing frontier AI models to identify vulnerabilities in public-sector software and improve the detection of advanced cyber threats.
\nFollowing the results of two pilot programmes, the government announced plans to extend the use of AI-assisted vulnerability detection, particularly across systems considered critical to national security and public services.
\nAn earlier government case study provides insight into how the technology was deployed.
\nOver one month, technical teams worked across nine government organisations, using AI models to examine public code repositories. The exercise produced 407 findings, including serious vulnerabilities involving authentication bypass, potential data exposure and remote code execution.
\nAccording to the government, all critical vulnerabilities identified during that exercise were remediated, with no evidence that they had been exploited.
\nThe reported cost of AI model usage was approximately £13,000.
\nAmong the findings was a vulnerability involving an automated GitHub workflow used by a government digital service. A specially constructed comment on a pull request could have triggered unauthorised code execution, potentially exposing credentials and other sensitive information.
\nThe example demonstrates how AI-assisted analysis can uncover security weaknesses that arise from interactions between different software components.
\nHowever, the government's overall figure of 1,352 vulnerabilities does not come with a complete public breakdown of the severity of every finding or the validation methods used across both pilot programmes.
\n![\"Photo:](\"/api/uploads/2026/10/02945637b78e456d071f58b064e747fb.webp\")
\nThe government's announcement also addresses a different category of cybersecurity risk: autonomous AI agents.
\nUnlike conventional AI assistants that primarily respond to requests, autonomous agents can perform sequences of actions, access connected systems and execute tasks with limited human intervention.
\nThese capabilities introduce new operational risks.
\nAn AI agent with access to sensitive systems could act outside its intended instructions, mishandle confidential information or perform actions with unintended consequences.
\nBritish authorities are working with the National Cyber Security Centre and the AI Security Institute to develop an approach for responding to serious incidents involving such systems.
\nThe work will examine how organisations can identify, investigate and contain incidents where autonomous AI behaviour contributes to security failures.
\nExisting cybersecurity procedures are largely designed around malicious actors, compromised infrastructure and conventional software vulnerabilities.
\nAutonomous systems introduce additional considerations, including the permissions granted to agents, the instructions they follow and the mechanisms available to interrupt their actions.
\nThe government has not yet published a completed incident-response framework.
\nThe UK is also examining whether its Cyber Security and Resilience Bill should more explicitly address risks associated with advanced artificial intelligence.
\nPotential measures include stronger reporting requirements for serious AI incidents and near misses, improved information sharing and additional protections for essential services.
\nHealthcare, energy and water infrastructure are among the sectors where disruptions to digital systems could have significant consequences.
\nThese proposals remain under consideration and should not be confused with requirements already established in law.
\nSeparately, the UK's AI Security Institute continues to evaluate advanced AI models, including through testing arrangements that provide access to models before their public release.
\nIndependent evaluation has become an increasingly important part of the government's approach to AI safety, particularly as models gain capabilities that allow them to interact directly with software systems and external tools.
\nThe government's findings demonstrate the potential value of using advanced AI to inspect large software repositories and identify weaknesses that may otherwise remain undetected.
\nThey also illustrate an important distinction between discovering a vulnerability and improving the security of an entire organisation.
\nAI-generated findings still require technical validation. Security teams must determine whether identified weaknesses are exploitable, assess their severity and implement appropriate fixes.
\nA large number of reported vulnerabilities does not necessarily indicate a proportional improvement in cybersecurity. The significance depends on the quality of the findings, the risks involved and whether the problems are successfully resolved.
\nThe earlier UK pilot provides useful evidence of practical results, including documented vulnerabilities and remediation. The broader programme will require similarly transparent reporting to establish whether AI-assisted detection can consistently outperform or complement existing security methods.
\nThe British government is moving beyond limited experiments by extending AI-assisted vulnerability detection and developing new procedures for security incidents involving autonomous systems.
\nDetails of the wider programme's budget, deployment timetable and performance targets have not yet been fully disclosed.
\nThe next measure of success will be whether AI-assisted security assessments produce lasting improvements across government infrastructure, including faster remediation and reduced exposure to serious vulnerabilities.
\nFor now, the UK's findings offer a concrete example of AI being deployed to address existing cybersecurity problems, even as governments prepare for a new generation of risks created by increasingly capable autonomous systems.
","featured_image_id":371,"intro_image_id":371,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":52,"ordering":0,"published_at":"2026-10-12 20:45:19","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-12 20:45:19","updated_at":"2026-10-12 20:45:19","key_takeaways":"[\"1,352 vulnerabilities identified: Two UK government AI cybersecurity pilots identified and addressed vulnerabilities across 548 code repositories in 18 departments.\",\"AI agents introduce new security challenges: British authorities are developing procedures for responding to incidents involving autonomous AI systems.\",\"Cybersecurity rules could expand: The UK is considering additional safeguards under its Cyber Security and Resilience Bill, including improved incident reporting.\"]","faq":"[{\"q\":\"How many cybersecurity vulnerabilities did the UK identify using AI?\",\"a\":\"The UK government reported that two pilot programmes identified and addressed 1,352 vulnerabilities across 548 code repositories in 18 government departments.\"},{\"q\":\"How is artificial intelligence being used in UK cybersecurity?\",\"a\":\"Advanced AI models are being used to analyse software repositories, identify potential security weaknesses and support technical teams investigating vulnerabilities.\"},{\"q\":\"What are the cybersecurity risks of autonomous AI agents?\",\"a\":\"Autonomous agents can interact with external systems and perform tasks with limited human supervision. If improperly configured or controlled, they may access sensitive information, execute unintended actions or create new security vulnerabilities.\"},{\"q\":\"Is the UK introducing new laws to regulate AI cybersecurity risks?\",\"a\":\"The government is considering additional measures under its Cyber Security and Resilience Bill. Possible changes include improved reporting of serious AI incidents and stronger protections for essential services. These proposals are not yet enacted.\"}]","sources":"[{\"title\":\"UK government announcement, October 12, 2026\",\"url\":\"https://www.gov.uk/government/news/britain-steps-up-national-defences-against-growing-ai-risks\"},{\"title\":\"UK Parliament, ministerial statement\",\"url\":\"https://questions-statements.parliament.uk/written-statements/detail/2026-10-12/hlws365\"},{\"title\":\"Government case study on AI-powered cyber defence\",\"url\":\"https://www.gov.uk/government/case-studies/when-ai-leaves-the-lab-testing-frontier-models-in-government-cyber-defence\"},{\"title\":\"- Government guidance on AI and vulnerability risks\",\"url\":\"https://www.gov.uk/guidance/ai-open-code-and-vulnerability-risk-in-the-public-sector\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/7abc2ec47b0ba09b1fd2f479fd354b2e.webp","featured_image_alt":"National Cyber Security Centre","featured_image_width":1600,"featured_image_height":1067,"intro_image_path":"/api/uploads/2026/10/7abc2ec47b0ba09b1fd2f479fd354b2e.webp","intro_image_width":1600,"intro_image_height":1067,"seo_title":"UK Uses AI to Find 1,352 Cybersecurity Flaws","meta_description":"Britain expands AI-powered cyber defence after identifying 1,352 vulnerabilities across government systems and plans new protections against AI agents.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":163,"title":"ArtCraft Is Challenging the Software Giants","slug":"artcraft-is-challenging-the-software-giants","short_description":null,"content":"A developer has used artificial intelligence to build free alternatives to some of the world's most established software applications. After targeting Adobe's creative tools, the project has expanded to Microsoft Word, Excel and PowerPoint.
\nThe applications are far from finished. But the development raises a question that goes beyond software pricing: what happens when customers who have spent years depending on a company's products discover that building alternatives is becoming easier?
\nBrandon Thomas, an experienced software developer behind the ArtCraft project, has attracted attention for creating open-source alternatives to applications such as Adobe Photoshop, Illustrator and Premiere Pro.
\nUsing Anthropic's Claude to accelerate development, Thomas has helped create a growing collection of applications written in Rust, designed to reproduce familiar professional workflows without proprietary subscription requirements.
\nThe project has now expanded to Microsoft Office with three applications:
\nWordCraft, an alternative to Microsoft Word.
\nGridCraft, a spreadsheet application modelled on Excel.
\nDeckCraft, a presentation tool inspired by PowerPoint.
\nThe source code is publicly available on GitHub. ArtCraft's website currently lists 12 applications spanning image editing, video production, documents, spreadsheets, presentations, audio and computer-aided design.
\n![\"Seven](\"/api/uploads/2026/10/818531dd83710319d5146ecd7abf496f.webp\")
Thomas is not someone who discovered programming through a few AI prompts. Reporting by Tom's Hardware identifies him as a developer with approximately 15 years of software engineering experience. That distinction matters. AI has accelerated the development process, but the project should not be mistaken for evidence that engineering experience has become unnecessary.
\nThe applications are also designed to operate locally, with support for several operating systems. Their architecture includes command-line controls and Model Context Protocol (MCP) interfaces, allowing AI agents to interact directly with the software.
\nThat last feature deserves attention. ArtCraft is trying to reproduce familiar applications while making them accessible to a different kind of user: software agents capable of carrying out tasks on someone's behalf.
\nAI Balkans also tested a selection of the ArtCraft applications.
\nOur initial experience was remarkably positive. The applications we tried worked smoothly, demonstrating that these projects are more than attractive interface concepts or screenshots circulating on social media.
\nThis was an initial hands-on assessment, not an exhaustive technical evaluation. We did not establish full compatibility with Adobe or Microsoft products, nor did we perform a comprehensive assessment of security, accessibility, performance under heavy workloads or professional reliability.
\nThose distinctions are important.
\nArtCraft's own documentation acknowledges substantial limitations. WordCraft's maintainers estimate that the application is approximately 60% of the way towards fully replacing Word for real work. GridCraft is estimated at around 50%, while DeckCraft is approximately 55% ready by the project's own assessments.
\nThese are developer estimates rather than independently verified benchmarks.
\nA presentation application that can create slides is not necessarily ready to manage a complex corporate presentation. A spreadsheet that supports formulas still needs to preserve calculations, formatting and large workbooks reliably. Professional image editing demands accurate file handling, colour management and consistent results.
\nArtCraft is still alpha software, and its developers say so.
\nYet there is a substantial difference between an unfinished project that barely functions and one that already performs useful work while development continues.
\nOur experience suggests these applications deserve serious attention, even if recommending them as complete replacements would be premature.
\nFor years, companies such as Adobe and Microsoft have operated from positions of considerable market strength.
\nTheir products are deeply embedded in education, business, government and creative industries. Files, training programmes, established workflows and collaboration practices make switching difficult, even when users become dissatisfied.
\nThis creates a degree of protection that newer software developers rarely enjoy.
\nIt also creates a risk.
\nIn 2025, Adobe changed its Creative Cloud individual plans, introducing Creative Cloud Pro and revised pricing in affected markets. Microsoft subsequently announced changes to commercial Microsoft 365 subscription prices, effective July 2026, alongside additional capabilities.
\n![\"PDFCraft](\"/api/uploads/2026/10/2f439dc32896299dd165f2da4dca21ac.webp\")
\nBoth companies presented their changes as part of continuing investment in product development and customer value. There are legitimate costs associated with maintaining complex software, providing security updates, operating cloud infrastructure and supporting enterprise customers.
\nNevertheless, customers judge those decisions according to their own needs and budgets.
\nAdobe's subscription practices have also faced regulatory scrutiny. In March 2026, the company announced a settlement with the US Department of Justice concerning allegations about subscription disclosures and cancellation practices. Adobe denied wrongdoing while agreeing to a $75 million payment and $75 million in services for eligible customers.
\nNone of this establishes that ArtCraft users have abandoned Adobe or Microsoft in significant numbers. There is no reliable evidence yet of such a migration.
\nWhat it does establish is the commercial environment in which alternatives are emerging.
\nFor an established software company, changing a subscription policy may be relatively straightforward. Recovering a customer's confidence can be much harder.
\nAnd if dissatisfaction encourages developers to create alternatives, reversing the original decision does not make those alternatives disappear.
\nTheir source code remains available. Other developers can contribute, modify the applications or build competing projects. Users who have begun testing different software may discover that they no longer need every feature of their previous subscriptions.
\nThe competitive environment has already changed, even if the market leader retains most of its customers.
\nOpen-source alternatives to commercial software are not new.
\nLibreOffice has provided a mature alternative to Microsoft Office for years. GIMP, Inkscape, Blender and numerous other projects demonstrate that powerful software does not necessarily require the same commercial model as proprietary applications.
\nArtCraft did not invent this approach.
\nWhat makes the current development interesting is the speed at which AI-assisted software engineering can help experienced developers assemble, test and improve substantial applications.
\nTasks that previously required considerable manual programming can now be accelerated through coding agents. A developer can delegate parts of implementation, testing and documentation while concentrating on architecture, product direction and unresolved technical problems.
\nThis does not eliminate the difficult parts of software engineering. Generated code still needs inspection, security review, integration and maintenance. The larger an application becomes, the more consequential those responsibilities are.
\nIt does, however, change the economics of experimentation.
\nA small team can now attempt projects that might previously have required significantly more development resources. Even when most experiments fail, the number of potential competitors can increase.
\nFor established companies, the challenge is no longer limited to other large corporations with comparable engineering budgets.
\nIt also includes experienced independent developers, open-source communities and small teams equipped with increasingly capable AI development tools.
\nThe advantage of scale remains substantial. But the cost of attempting to challenge that scale is changing.
\nFor freelancers, creative studios, startups and smaller organisations across the Balkans, software subscriptions represent recurring operating costs.
\nA free alternative that handles everyday tasks reliably could make a practical financial difference. Locally running applications may also appeal to organisations seeking greater control over their files and working environments.
\nThe opportunity comes with important qualifications.
\nOrganisations exchanging complex Word documents, Excel workbooks, PowerPoint presentations or Adobe project files need dependable compatibility. Businesses handling confidential material need confidence in software security and maintenance. Local language support, including Macedonian and Albanian text handling, should be tested rather than assumed.
\nThe sensible approach is to experiment with alternatives on non-critical work before considering any wider migration.
\nFor the region's developers, the larger lesson may be more important than the applications themselves. AI-assisted engineering creates opportunities to develop specialised software for smaller markets, local languages and workflows that major international vendors have little commercial incentive to prioritise.
\nThe opportunity is not simply to reproduce another company's interface. It is to build useful software around needs that remain poorly served.
\nAdobe and Microsoft have advantages that ArtCraft cannot reproduce quickly: decades of development, established customer relationships, technical support, extensive integrations and mature enterprise infrastructure.
\nThere are also unresolved questions about long-term maintenance, code quality, file compatibility and the intellectual-property boundaries of software designed to resemble established commercial applications. ArtCraft describes its work as clean-room reimplementation, but that description alone does not settle every potential legal question.
\nThe project may grow into a sustainable alternative. Some applications may progress much further than others. Development could also slow as the difficulty of maintaining a dozen complex products becomes apparent.
\nThe next meaningful evidence will come from sustained development, independent testing, improvements in file compatibility and actual adoption, rather than GitHub attention alone.
\nStill, ArtCraft offers an important lesson to the software industry.
\nEstablished companies can change prices, revise policies and introduce new features. Those decisions remain within their control.
\nWhat they cannot fully control is what their customers and competitors learn to build in response.
\nAnd once those alternatives exist, winning users back may require more than restoring the conditions that persuaded them to look elsewhere.
","featured_image_id":368,"intro_image_id":368,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":8,"view_count":53,"ordering":0,"published_at":"2026-10-12 15:51:22","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-12 15:51:22","updated_at":"2026-10-12 15:51:22","key_takeaways":"[\"ArtCraft is challenging established software business models. Brandon Thomas and contributors have developed free, open-source alternatives to Adobe and Microsoft applications with substantial AI assistance.\",\"AI Balkans' initial testing was positive, but the apps are unfinished. The project shows functional promise while acknowledging major remaining compatibility and reliability work.\",\"Changing a business policy cannot automatically restore lost trust. As AI reduces some software development barriers, dissatisfied users and independent developers have more opportunities to explore alternatives.\"]","faq":"[{\"q\":\"Who created ArtCraft's Adobe and Microsoft alternatives?\",\"a\":\"Brandon Thomas is the developer most closely associated with the project. He has used Anthropic's Claude in developing the applications, while public repositories allow contributions from others.\"},{\"q\":\"Are WordCraft, GridCraft and DeckCraft free?\",\"a\":\"Yes. They are publicly available as open-source software through ArtCraft's GitHub repositories, with permissive licensing.\"},{\"q\":\"Can these applications replace Adobe Creative Cloud or Microsoft 365?\",\"a\":\"Not comprehensively at present. They remain early-stage products, with important limitations involving compatibility, feature depth and reliability. They may already be useful for selected tasks.\"},{\"q\":\"Does ArtCraft mean large software companies are losing customers?\",\"a\":\"Not necessarily. There is no verified evidence that ArtCraft has caused substantial customer losses for Adobe or Microsoft. Its significance lies in demonstrating how the economics of creating competing software may be changing.\"}]","sources":"[{\"title\":\"ArtCraft official application directory\",\"url\":\"https://getartcraft.com/apps\"},{\"title\":\"Tom's Hardware, 11 October 2026\",\"url\":\"https://www.tomshardware.com/software/microsoft-office/creator-of-free-adobe-clones-unveils-open-source-microsoft-office-excel-powerpoint-and-word-replications-autodesk-tools-and-other-subscription-based-services-may-come-next\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/1b29d4650b242a8dafafb0395b604b2d.webp","featured_image_alt":"Artcraft","featured_image_width":1200,"featured_image_height":630,"intro_image_path":"/api/uploads/2026/10/1b29d4650b242a8dafafb0395b604b2d.webp","intro_image_width":1200,"intro_image_height":630,"seo_title":"What Adobe and Microsoft Can’t Undo","meta_description":"ArtCraft's AI-built Adobe and Microsoft alternatives raise questions about software subscriptions, customer trust and the cost of competition.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":162,"title":"Europe’s AI Factories Face Access Limits","slug":"europe-s-ai-factories-face-access-limits","short_description":null,"content":"A review of European regulations and computing access programmes shows that all six Western Balkan economies have a route to Europe's AI infrastructure. High demand is already restricting access to some of its most powerful systems.
\nEurope's AI Factories promise startups, researchers and businesses access to supercomputing resources that would otherwise be prohibitively expensive. The Western Balkans are included in that ambition, but an examination of European Commission documents and EuroHPC access rules reveals an important qualification: eligibility does not guarantee available computing capacity.
\nAll six Western Balkan economies are associated with the EU's Horizon Europe research programme. Under Article 16 of the EuroHPC Regulation, users established in countries associated with Horizon Europe or the Digital Europe Programme are entitled to access the European Union's share of computing time on EuroHPC supercomputers acquired after 2020, subject to the applicable access conditions.
\nThat includes organisations based in Albania, Bosnia and Herzegovina, Kosovo, Montenegro, North Macedonia and Serbia.
\nThe opportunity is substantial. European programmes offer qualifying companies free access to GPU resources for AI development, testing and research.
\nBut EuroHPC's own notices show that demand is already affecting availability. Some computing systems are temporarily unavailable for new allocations, raising practical questions about how smaller companies can obtain the resources they need.
\nMuch of the regional attention has focused on North Macedonia and Serbia, which were selected in October 2025 to host AI Factory Antennas.
\nNorth Macedonia's VEZILKA initiative and Serbia's SAIFA are intended to connect local organisations with the European AI Factory network, including technical expertise, computing infrastructure and related services.
\nNeither country, however, has exclusive access to the network.
\nThe European Commission's list of Horizon Europe associated countries includes all six Western Balkan economies. Article 16(4) of Regulation (EU) 2021/1173 provides an access entitlement to the Union's share of computing time on EuroHPC systems acquired after 2020.
\nThe distinction is particularly relevant for Kosovo.
\nAlthough Kosovo is not named in one of EuroHPC's general AI Factory eligibility lists, it is officially associated with Horizon Europe. The broader regulation therefore provides a legal basis for access to qualifying computing resources.
\nThis does not remove the need to satisfy individual programme requirements. Applicants must still meet the conditions of the relevant call, including technical suitability and the proposed use of the infrastructure.
\nIt does mean that the absence of a national AI Factory or antenna should not automatically be understood as exclusion from European computing resources.
\nEuroHPC offers several access routes for industrial innovation, depending on the size and complexity of a project.
\nThe Playground programme is aimed at startups, small businesses and organisations beginning to work with advanced computing. It allows smaller AI experiments to receive allocations lasting one, two or three months.
\nApplications are processed on a first-come, first-served basis. EuroHPC targets access within two working days after successful eligibility and technical assessments.
\nThe Fast Lane programme serves companies with more demanding requirements, allowing applications for up to 50,000 GPU hours.
\nFor projects requiring greater capacity, the Large Scale programme offers allocations exceeding 50,000 GPU hours, with project durations of three, six or twelve months.
\nQualifying research and innovation activities can receive access without computing fees. Commercial access outside those conditions may be charged.
\nThese arrangements create a realistic opportunity for regional developers working on language models, scientific applications and other computationally demanding AI systems.
\nA company in Albania developing an Albanian-language model, for example, does not necessarily need to purchase its own high-performance hardware or wait for Albania to establish an AI Factory.
\nIt can explore the relevant European access programmes and apply for computing time.
\nWhether its application succeeds depends on the project requirements and available resources.
\nThe clearest limitation appears in EuroHPC's published access notices.
\nOn its Large Scale access page, EuroHPC states that the MareNostrum and LUMI-G resources are unavailable for new allocations because of exceptionally high demand.
\nIts Fast Lane programme also identifies the MareNostrum5 ACC partition as unavailable for new Fast Lane allocations.
\nApplicants are advised to select alternative systems when their preferred computing resources cannot be accessed.
\nThese restrictions do not mean that Europe's entire AI Factory network has reached capacity. Other systems remain available through the relevant programmes.
\nThey do demonstrate that access is constrained by actual hardware availability, even when an organisation meets the formal eligibility requirements.
\nFor smaller businesses, choosing an alternative system may require adjustments to software environments, project schedules or computing requirements.
\nEuroHPC also publishes limits on the maximum GPU hours that can be allocated on individual systems over specified periods. Large projects must take those constraints into account when planning their applications.
\nThe result is a system in which free access is possible but cannot be treated as unlimited or guaranteed.
\nThe AI Factory Antennas in North Macedonia and Serbia introduce another important dimension.
\nTheir purpose is to bring the services of established European AI Factories closer to national research and business ecosystems.
\nThis can include connecting companies with technical expertise and supporting their use of European AI infrastructure.
\nFor a startup with limited experience in high-performance computing, such assistance could be valuable.
\nHowever, the existence of these antennas does not mean that companies in North Macedonia and Serbia automatically receive preferential GPU allocations.
\nNor does it establish that organisations in Albania, Bosnia and Herzegovina, Kosovo or Montenegro receive less computing time.
\nThe public documents examined do not provide sufficient country-by-country allocation figures to support either conclusion.
\nWhat can be established is that the European network combines centralised computing infrastructure with a developing set of national and regional support structures.
\nThe ability of smaller companies to use that infrastructure will depend partly on how effectively those structures connect applicants with available resources.
\nThere is a difference between announcing access to infrastructure and demonstrating that organisations are using it.
\nEuroHPC publishes the application procedures, eligibility criteria and technical arrangements governing its programmes.
\nThe material reviewed for this article does not establish how many Western Balkan organisations have successfully obtained AI Factory computing allocations, or how GPU hours have been distributed among the six economies.
\nThose figures would provide a clearer picture of the programme's practical reach.
\nA country could be fully eligible under European rules while relatively few of its businesses make use of the available resources. Reasons might include limited awareness, technical capacity, project readiness or infrastructure availability.
\nWithout application and allocation data, none of those explanations can be established as the principal cause.
\nThe documented capacity restrictions are nevertheless relevant. They demonstrate that formal access and actual resource availability are different parts of the same system.
\nThe European Commission has presented AI Factories as an important part of Europe's effort to develop its own AI capabilities and support innovation.
\nFor the Western Balkans, the programme provides access to infrastructure on a scale that smaller national economies would struggle to finance independently.
\nThe legal framework offers a route for organisations across all six economies, while the antennas in North Macedonia and Serbia are intended to bring support services closer to local users.
\nThe next question is measurable: how many regional startups and research institutions will actually receive computing time, and what will they produce with it?
\nEuroHPC's published restrictions already show that demand can limit access to particular systems.
\nFor the Western Balkans, the real value of Europe's AI Factories will be determined by successful applications, usable computing capacity and completed projects, rather than the number of infrastructure partnerships announced.
","featured_image_id":367,"intro_image_id":367,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":7,"view_count":39,"ordering":0,"published_at":"2026-10-11 09:24:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-11 09:24:38","updated_at":"2026-10-11 09:25:21","key_takeaways":"[\"All six Western Balkan economies are associated with Horizon Europe, providing a legal route to qualifying EuroHPC computing resources.\",\"Eligible startups can apply for free computing time through programmes such as Playground, Fast Lane and Large Scale.\",\"EuroHPC has reported high-demand restrictions on certain computing systems, showing that formal access does not guarantee capacity.\"]","faq":"[{\"q\":\"Can companies from all six Western Balkan economies access AI Factories?\",\"a\":\"The EuroHPC legal framework provides an access route for users in Horizon Europe associated countries on systems acquired after 2020. Individual application requirements and restrictions still apply.\"},{\"q\":\"Is access completely free?\",\"a\":\"Qualifying research and SME innovation activities can use computing resources without access fees. Other commercial uses may be chargeable.\"},{\"q\":\"Does Kosovo qualify?\",\"a\":\"Kosovo is associated with Horizon Europe and falls within the country category covered by Article 16(4) for eligible EuroHPC systems. Specific calls may apply additional conditions.\"},{\"q\":\"Why are some supercomputers unavailable?\",\"a\":\"EuroHPC reports exceptionally high demand for selected resources, including LUMI-G and MareNostrum-related partitions.\"}]","sources":"[{\"title\":\"EuroHPC Regulation, Articles 16–18\",\"url\":\"https://eur-lex.europa.eu/eli/reg/2021/1173/2026-01-20/eng\"},{\"title\":\"European Commission, Horizon Europe associated countries\",\"url\":\"https://research-and-innovation.ec.europa.eu/news/all-research-and-innovation-news/updates-association-third-countries-horizon-europe-2021-12-21_en\"},{\"title\":\"EuroHPC, Playground Access\",\"url\":\"https://www.eurohpc-ju.europa.eu/playground-access-ai-factories_en\"},{\"title\":\"EuroHPC, Fast Lane Access\",\"url\":\"https://www.eurohpc-ju.europa.eu/fast-lane-access-ai-factories_en\"},{\"title\":\"EuroHPC, Large Scale Access\",\"url\":\"https://www.eurohpc-ju.europa.eu/large-scale-access-ai-factories_en\"},{\"title\":\"European Commission, AI Factories\",\"url\":\"https://digital-strategy.ec.europa.eu/en/policies/ai-factories\"}]","category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/31034a2a772706065f08a3397f88a8cb.webp","featured_image_alt":"AI Antenna","featured_image_width":1200,"featured_image_height":800,"intro_image_path":"/api/uploads/2026/10/31034a2a772706065f08a3397f88a8cb.webp","intro_image_width":1200,"intro_image_height":800,"seo_title":"Europe’s AI Factories Face Access Limits","meta_description":"All six Western Balkan economies have routes to European AI computing resources, but capacity restrictions are already limiting availability.","robots":"index","canonical_url":"eu-ai-factories-western-balkans-access-limits","og_title":"","og_description":""},{"id":161,"title":"Most Americans Say AI Is Developing Too Fast","slug":"most-americans-say-ai-is-developing-too-fast","short_description":null,"content":"Nearly two-thirds of Americans believe artificial intelligence is developing too quickly, according to a new national survey released on October 8 by the AP-NORC Center for Public Affairs Research.
\nThe findings show that 64% of U.S. adults consider the current pace of AI development too fast, compared with just 8% who believe it is moving too slowly. Approximately one-quarter say development is progressing at about the right pace.
\nThe survey, conducted between September 24 and 28, 2026, involved 2,140 adults across the United States. It comes as technology companies continue introducing increasingly capable AI systems into workplaces, education, information services and other areas of daily life.
\nBeyond concerns about speed, the research identifies two priorities shared by large majorities of respondents: keeping AI under human control and protecting workers from its potential consequences.
\nAlthough attitudes differ between political groups, concern about the pace of AI development is widespread.
\nAmong Democrats, 76% believe AI is advancing too quickly. The figure is lower among Republicans, at 54%, but still represents a majority.
\nThe findings suggest that unease about AI development extends across political affiliations, even when respondents disagree about the broader role of government in regulating emerging technologies.
\nThere is also considerable agreement about several policy objectives.
\nRoughly eight in ten respondents consider protecting American workers and keeping AI under human control important priorities for the U.S. government.
\nThe survey also found broad support across political groups for technological innovation, indicating that concern about AI does not necessarily translate into opposition to its development.
\nDifferences become more pronounced when respondents are asked about environmental protection and competition with other countries.
\nThe emphasis on employment protection reflects an increasingly important question in the public discussion about artificial intelligence: how technological progress should be managed when it affects existing jobs.
\nAI systems are becoming capable of performing tasks previously carried out by employees, particularly in areas involving information processing, customer communication, administrative work and software development.
\nThe AP-NORC findings do not measure actual job losses or predict how employment will change. They measure public priorities at a time when businesses and governments are deciding how to introduce increasingly autonomous technologies.
\nHuman oversight is another central concern.
\nAs AI applications move beyond generating content toward carrying out tasks and supporting decisions, questions about responsibility become more significant.
\nFor public institutions and companies, this includes determining when automated systems should require human approval and who remains accountable when mistakes occur.
\nThe survey suggests that many Americans want these considerations to remain central to AI policy.
\nThe research also examined public confidence in how political leaders are addressing artificial intelligence.
\nOnly 31% of respondents approve of President Donald Trump's handling of AI.
\nAsked which major political party they trust more to manage the technology, 23% selected the Democratic Party and 21% the Republican Party. Another 39% said neither party could be trusted to do a good job.
\nThese findings point to limited public confidence in the ability of political institutions to address the challenges associated with AI development.
\nThey also suggest that respondents' concerns extend beyond the companies building AI systems to the institutions responsible for establishing rules and safeguards.
\nThe survey examined whether Americans are using AI to obtain information ahead of the November 2026 midterm elections.
\nTwenty-one percent reported that they had used or planned to use AI to learn about a political candidate or an issue appearing on the ballot.
\nThe figures were lower for other election-related activities.
\nTwelve percent had used or planned to use AI to find information about voting procedures, while 10% reported the same for voter registration information. Another 10% said they had used or planned to use AI to help decide whom to vote for.
\nBecause these figures combine existing use with intended future use, they should not be interpreted as the proportion of Americans already relying on AI for electoral decisions.
\nNevertheless, the findings indicate that AI is becoming part of how some Americans approach political information, even as concerns about its reliability and oversight persist.
\nThe survey reflects American public opinion and cannot be treated as evidence of attitudes in Europe or the Western Balkans.
\nHowever, the policy questions it raises are relevant to countries considering how to regulate AI and introduce it into public services, education and business.
\nFor Western Balkan economies, where AI adoption is developing alongside efforts to strengthen digital infrastructure and public-sector capacity, questions about skills, employment and human accountability are likely to remain important.
\nThe findings also illustrate why measuring AI adoption alone provides an incomplete picture of public attitudes.
\nPeople may use AI tools while remaining concerned about how quickly the technology is advancing and whether sufficient safeguards are in place.
\nFor policymakers, the challenge is to distinguish public support for technological innovation from public confidence in the institutions overseeing its development.
\nThe AP-NORC survey provides evidence of that distinction in the United States. Whether similar attitudes exist across the Western Balkans would require comparable regional research.
","featured_image_id":366,"intro_image_id":366,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":29,"ordering":0,"published_at":"2026-10-09 08:40:58","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-09 08:40:58","updated_at":"2026-10-09 08:40:58","key_takeaways":"[\"64% of U.S. adults believe AI is developing too quickly, including 76% of Democrats and 54% of Republicans.\",\"Around eight in ten Americans want the government to prioritize keeping AI under human control and protecting workers.\",\"Confidence in political leadership on AI is limited, with 39% saying neither major U.S. political party can be trusted to handle the technology well.\"]","faq":"[{\"q\":\"How many Americans believe AI is developing too quickly?\",\"a\":\"According to AP-NORC, 64% of U.S. adults believe AI is advancing too fast, while 8% say it is advancing too slowly.\"},{\"q\":\"How many people participated in the survey?\",\"a\":\"The survey included 2,140 U.S. adults interviewed between September 24 and 28, 2026. Its overall margin of sampling error is ±2.9 percentage points.\"},{\"q\":\"What concerns do Americans have about AI?\",\"a\":\"The survey identifies strong public support for protecting workers and keeping AI under human control. It also reveals differences over environmental concerns and international competition.\"},{\"q\":\"Does the survey represent attitudes in the Western Balkans?\",\"a\":\"No. The findings represent U.S. public opinion. Comparable polling would be necessary to determine whether people in Western Balkan countries hold similar views.\"}]","sources":"[{\"title\":\"AP-NORC Center for Public Affairs Research, Most Adults Think AI Is Developing Too Fast, October 8, 2026\",\"url\":\"https://apnorc.org/projects/most-adults-think-ai-is-developing-too-fast/\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/e15424ff39580ada3b0b7f9718e618f3.webp","featured_image_alt":"AI Developing Too Fast","featured_image_width":1200,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/e15424ff39580ada3b0b7f9718e618f3.webp","intro_image_width":1200,"intro_image_height":900,"seo_title":"64% of Americans Say AI Is Developing Too Fast","meta_description":"A new AP-NORC survey finds 64% of Americans believe AI is moving too fast, with strong support for worker protection and human oversight.","robots":"index","canonical_url":"americans-ai-developing-too-fast-survey","og_title":"","og_description":""},{"id":160,"title":"Western Balkans Digital Summit 2026 to Focus on AI","slug":"western-balkans-digital-summit-2026-to-focus-on-ai","short_description":null,"content":"The Western Balkans Digital Summit will bring government officials, technology experts and business representatives to Herceg Novi, Montenegro, from October 12 to 14, with artificial intelligence, cybersecurity and regional digital integration among the central topics.
\nThe eighth edition will take place at One&Only Portonovi under Montenegro's 2026 presidency of the Berlin Process. More than 150 participants are expected, including representatives of the six Western Balkan economies, the European Commission, academic institutions, businesses and civil society organizations.
\n![\"WB](\"/api/uploads/2026/10/72375180aff282d64967a55e08d278d5.webp\")
The event is organized by Montenegro's Ministry of Public Administration in partnership with the Regional Cooperation Council (RCC), which has supported the summit since its introduction in 2018.
\nAccording to the Montenegrin government's announcement, this year's programme is intended to address practical challenges in developing a more connected, secure and competitive regional digital economy.
\nThe summit's programme is organized around five main areas: digital economy, digital skills, digital infrastructure, artificial intelligence and connectivity, and trust and security.
\nAI discussions will focus on the implementation of advanced technologies and their role in supporting regional connectivity. The wider programme will also examine how businesses, governments and educational institutions can develop the skills and infrastructure needed for digital transformation.
\nCybersecurity will be another central topic, with attention to strengthening the region's resilience against digital threats and improving cooperation between institutions.
\nThe Regional Cooperation Council has also identified digital public services and online child protection among the subjects expected to feature in the discussions.
\nFor the Western Balkans, these issues extend beyond technology adoption. Differences in institutional capacity, digital infrastructure and access to skilled professionals affect how quickly countries can introduce new digital services and benefit from AI applications.
\nIn an interview published by the Ministry of Public Administration on October 8, Montenegro's Minister of Public Administration, Marash Dukaj, outlined the country's expectations for the summit.
\nHe said regional cooperation should produce practical results in areas such as cross-border data exchange, digital identity, public services and cybersecurity.
\nOne priority is closer alignment with the European Union's Digital Single Market, including work toward more affordable mobile roaming between the Western Balkans and EU countries.
\nDukaj also pointed to the need for stronger cooperation on cybersecurity, arguing that increasingly interconnected public systems require coordinated responses to digital threats.
\nMontenegro has announced plans to expand its digital public services and develop an AI assistant for its e-government platform.
\nThe minister stressed that AI in public administration must remain subject to human oversight, particularly when systems are involved in legal or administrative decisions.
\nThese are national priorities presented by Montenegro ahead of the summit, rather than agreements already reached by participating governments.
\nThe Western Balkans Digital Summit has served as a meeting point for regional governments and technology stakeholders since 2018. The previous edition took place in Skopje in October 2025.
\nThis year's gathering comes as the region continues working to align its digital policies and infrastructure with European frameworks.
\nFor businesses, improved digital connectivity and interoperable public services could reduce administrative barriers when operating across borders.
\nFor governments, common approaches to cybersecurity, digital identity and AI governance could support more reliable public services and regional cooperation.
\nHowever, the extent to which these ambitions translate into practical results will depend on agreements, financing and implementation after the summit.
\nThe most relevant outcomes will be any concrete commitments involving AI governance, cybersecurity cooperation, digital public services and regional connectivity.
\nParticipants are also expected to discuss how to strengthen digital skills and prepare the workforce for changing technological requirements.
\nWith AI increasingly entering public administration and business operations, the challenge for Western Balkan economies is to move from policy discussions toward measurable improvements in services, institutional capacity and digital infrastructure.
\nThe October 12–14 summit provides another opportunity to establish shared priorities. Its longer-term significance will depend on what participating institutions agree to implement after leaving Herceg Novi.
","featured_image_id":364,"intro_image_id":364,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":13,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":87,"ordering":0,"published_at":"2026-10-09 08:34:38","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-09 08:34:38","updated_at":"2026-10-09 08:34:38","key_takeaways":"[\"The eighth Western Balkans Digital Summit takes place in Herceg Novi, Montenegro, from October 12 to 14, with over 150 participants expected.\",\"Artificial intelligence, cybersecurity, infrastructure, digital skills and economic integration form the programme's five central areas.\",\"Montenegro wants the summit to advance practical regional cooperation, including digital public services and closer integration with the EU digital market.\"]","faq":"[{\"q\":\"When and where is the Western Balkans Digital Summit 2026?\",\"a\":\"The summit is scheduled for October 12–14, 2026, at One&Only Portonovi in Herceg Novi, Montenegro.\"},{\"q\":\"What are the main topics?\",\"a\":\"AI, cybersecurity, digital economy, digital skills, infrastructure, connectivity and regional digital integration.\"},{\"q\":\"Who organizes the summit?\",\"a\":\"Montenegro's Ministry of Public Administration, under government patronage and in partnership with the Regional Cooperation Council.\"},{\"q\":\"Can participants register?\",\"a\":\"The government announced that registration is open through its official event registration link.\"}]","sources":"[{\"title\":\"Government of Montenegro, October 7 announcement\",\"url\":\"https://www.gov.me/clanak/otvorene-prijave-za-osmi-digitalni-samit-zapadnog-balkana-u-herceg-novom\"},{\"title\":\"Regional Cooperation Council, summit information\",\"url\":\"https://www.rcc.int/events/1670/rcc-co-organizes-8th-western-balkans-digital-summit\"},{\"title\":\"Montenegro Ministry of Public Administration, October 8 interview\",\"url\":\"https://www.gov.me/clanak/dukaj-crna-gora-zeli-da-kroz-samit-ubrza-povezivanje-regiona-sa-digitalnim-trzistem-eu\"},{\"title\":\"Western Balkans Digital Summit website\",\"url\":\"https://digitalsummitwb6.com/\"}]","category_name":"Montenegro","category_slug":"montenegro","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/cebed43670cecd766b9558cfff22f5b9.webp","featured_image_alt":"Montenegro","featured_image_width":1200,"featured_image_height":799,"intro_image_path":"/api/uploads/2026/10/cebed43670cecd766b9558cfff22f5b9.webp","intro_image_width":1200,"intro_image_height":799,"seo_title":"Western Balkans Digital Summit 2026 to Focus on AI","meta_description":"The Western Balkans Digital Summit opens in Montenegro on October 12, bringing AI, cybersecurity, digital skills and EU integration into focus.","robots":"index","canonical_url":"western-balkans-digital-summit-2026-ai","og_title":"","og_description":""},{"id":159,"title":"Skopje and Pristina Selected for UNDP AI Programme","slug":"skopje-and-pristina-selected-for-undp-ai-programme","short_description":null,"content":"Skopje and Pristina have been selected to participate in a United Nations Development Programme (UNDP) initiative that will help municipalities explore practical uses of artificial intelligence in public administration.
\nThe two Western Balkan capitals are among four municipalities chosen for the first phase of the CEF x AI programme, alongside Rivne in Ukraine and Yerevan in Armenia. UNDP announced the selection on October 8, 2026.
\nAccording to UNDP's official announcement, the programme will support cities in testing AI applications while developing the institutional capacity, governance arrangements and safeguards needed for responsible use.
\nThe selected municipalities will focus on different urban challenges, ranging from environmental management to energy efficiency and transport planning.
\nSkopje will examine how artificial intelligence could support the management of water resources and urban green spaces.
\nThe city's proposed work includes two initiatives, AI4Water and Green Cadastre, which will be developed through the programme's Demonstrator and BOOST tracks.
\nThe projects will explore how AI-generated evidence and analysis might help municipal teams and public enterprises make decisions about natural resources, water management and urban greenery.
\nUNDP describes the work as an opportunity to examine how new information affects professional judgment, coordination between municipal departments and practical resource-management decisions.
\nThe selection does not mean that AI systems have already been deployed or that operational improvements have been demonstrated.
\nThe next stage will involve developing specific activities and testing whether the proposed approaches can provide useful information for municipal decision-making.
\nPristina will focus on using generative AI to support municipal transport planning.
\nThrough the BOOST track, the city will work on a pilot intended to help planners organize and assess information about transport routes and changing neighbourhood needs.
\nThe initiative will examine whether AI can make relevant planning information more accessible and useful to municipal professionals.
\nUNDP emphasizes that professional judgment and responsibility should remain with municipal staff.
\nThe proposed system is therefore intended to support transport planning rather than make independent decisions about routes or infrastructure.
\nCEF x AI is part of UNDP's City Experiment Fund, an initiative supporting municipalities across Europe and Central Asia in addressing complex urban challenges.
\nThe programme has two complementary components.
\nThe CEF x AI Demonstrator focuses on building longer-term institutional capabilities. Participating municipalities will explore AI across administrative activities while strengthening governance, coordination and internal processes.
\nThe GenAI Challenge, delivered through UNDP's BOOST impact acceleration programme in partnership with IEEE Humanitarian Technologies, connects municipalities with external innovation teams to develop open-source generative AI solutions for specific local challenges.
\nThe programme places emphasis on defining the municipal problem before selecting the technology.
\nThis approach is especially relevant for public institutions, where data quality, departmental coordination and institutional accountability may determine whether an AI project delivers useful results.
\nAccording to UNDP, ten applications were reviewed by a selection panel involving representatives of the Slovak Republic's Ministry of Finance, the City of Bratislava and UNDP.
\nThe selection criteria included strategic relevance, institutional commitment, technical feasibility, responsible AI safeguards, implementation capacity and long-term sustainability.
\nThe final group includes four municipalities with different priorities. Rivne will explore AI-supported energy modernization and building efficiency, while Yerevan will focus on air quality and environmental governance.
\nSkopje and Pristina had previously participated in a wider municipal AI development process, including a UNDP bootcamp held in Podgorica, Montenegro, in June 2026.
\nTheir latest selection represents a further stage of that process, moving from initial exploration toward the design of practical municipal experiments.
\nDuring autumn 2026, the selected municipalities will work with UNDP to refine their proposals, establish implementation plans and identify their first activities.
\nThe programme is implemented through UNDP's City Experiment Fund with financial support from the Ministry of Finance of the Slovak Republic.
\nFor Skopje and Pristina, the immediate challenge will be translating the selected concepts into usable projects while ensuring that municipal employees retain responsibility for decisions.
\nThe results could also offer lessons for other Western Balkan municipalities considering AI for environmental monitoring, infrastructure management and public services.
\nThe next developments to watch are the specific implementation plans, the data and institutional resources allocated to each project, and the evidence produced during the initial pilots.
","featured_image_id":363,"intro_image_id":363,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":24,"ordering":0,"published_at":"2026-10-09 08:22:38","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-09 08:22:38","updated_at":"2026-10-09 08:22:38","key_takeaways":"[\"Skopje and Pristina are among four cities selected for the first phase of UNDP's CEF x AI programme.\",\"Skopje will explore AI applications for water and greenery management, while Pristina will examine AI-supported transport planning.\",\"The programme emphasizes practical experimentation, municipal responsibility and AI governance rather than immediate large-scale deployment.\"]","faq":"[{\"q\":\"Which cities were selected for the UNDP AI programme?\",\"a\":\"Skopje, Pristina, Rivne and Yerevan.\"},{\"q\":\"What will Skopje use AI for?\",\"a\":\"The city plans to explore AI-supported water management and urban greenery initiatives, including AI4Water and Green Cadastre.\"},{\"q\":\"What is Pristina's project about?\",\"a\":\"Pristina will investigate whether generative AI can help municipal teams evaluate transport routes and changing neighbourhood needs.\"},{\"q\":\"Are the AI systems already operational?\",\"a\":\"The announcement concerns programme selection and planned activities. It does not establish that the proposed systems have already been deployed.\"}]","sources":"[{\"title\":\"UNDP, selection of four municipalities, October 8, 2026\",\"url\":\"https://innovation.eurasia.undp.org/the-cef-x-ai-programme/\"},{\"title\":\"UNDP, original programme launch, June 12, 2026\",\"url\":\"https://innovation.eurasia.undp.org/undp-launches-cef-x-ai-programme-for-municipalities-across-europe-and-central-asia/\"},{\"title\":\"UNDP, municipal AI bootcamp follow-up, July 21, 2026\",\"url\":\"https://innovation.eurasia.undp.org/where-does-ai-actually-help-a-city-ten-municipalities-spent-three-days-finding-out/\"}]","category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/f3d03d10c71f37b55da58b7d71f4b6ce.webp","featured_image_alt":"UNDP","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/f3d03d10c71f37b55da58b7d71f4b6ce.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Skopje and Pristina Join UNDP AI Programme","meta_description":"Skopje and Pristina join UNDP's CEF x AI programme to explore AI applications in water management, urban greenery and transport planning.","robots":"index","canonical_url":"skopje-pristina-undp-ai-programme","og_title":"","og_description":""},{"id":158,"title":"Google Unveils Gemini Agent and New Medical AI Findings","slug":"google-unveils-gemini-agent-and-new-medical-ai-findings","short_description":null,"content":"Google announced a new AI agent designed to carry out tasks across workplace applications on October 8, alongside findings from a clinical study in which its medical AI system interacted with real patients before their appointments.
\nThe announcements cover two areas where AI is moving into practical use: everyday business operations and medical consultations. Although the technologies serve different purposes, both raise questions about how much responsibility AI systems should be given and how their work should be supervised.
\nAt its Gemini at Work 2026 event, Google Cloud introduced the Gemini agent, a system designed to complete tasks across business applications rather than simply respond to individual requests.
\nAccording to Google Cloud's announcement, the agent can access organizational information, plan activities, use connected tools and deliver completed work through applications employees already use.
\nThe system is designed to operate across Gmail, Google Drive, Docs, Sheets, Slides, Chat and Calendar. It can also connect to external services, including Microsoft 365, Slack, Salesforce and development environments.
\nFor example, an employee could ask the agent to research a market, prepare a financial analysis and produce a presentation. Instead of managing each application separately, the agent would coordinate the different steps.
\nGoogle also introduced the concept of coworker agents. These are AI agents with dedicated organizational identities, including email addresses, calendars and access permissions. Teams can assign them ongoing responsibilities and interact with them through shared documents or messaging channels.
\nThe company says organizations will be able to control what information these agents can access and review records of their actions.
\nGiving AI systems permission to execute workplace tasks introduces challenges beyond those associated with conventional chatbots.
\nAn agent that prepares a document has limited operational consequences. An agent connected to corporate databases, financial systems or internal communications may be able to take actions affecting an organization's operations.
\nGoogle says its architecture addresses these concerns through separate agent identities, role-based permissions, audit logs and administrative policies.
\nThe system can also select different AI models for different tasks, including models from Google's Gemini family and Anthropic's Claude family. This approach is intended to balance performance and operating costs.
\nFor businesses considering AI agents, the announcement shifts attention toward practical questions about access permissions, oversight and the reliability of automatically executed tasks.
\nOn the same day, Google announced results from a clinical study conducted with Beth Israel Deaconess Medical Center and published in The Lancet.
\nThe research involved 98 patients who interacted with Google's Articulate Medical Intelligence Explorer (AMIE), an experimental conversational AI system, before urgent primary care appointments.
\nUnlike earlier evaluations using simulated consultations, the study examined interactions involving actual patients in a clinical environment.
\nPhysicians supervised the AI conversations in real time. According to Google's summary of the research, none required interruption under the study's predefined safety criteria.
\nThe company reported that AMIE's lists of possible diagnoses included the physicians' final diagnoses in 90% of cases.
\nClinicians also said the AI-generated summaries helped them prepare for patient visits in 75% of cases and influenced their approach to care in more than half of the consultations.
\nThese results suggest that conversational AI may help physicians collect information before appointments and organize relevant details. However, they do not establish that the system can independently diagnose patients or safely replace medical professionals.
\nThe study represents an early investigation of patient-facing AI in a supervised clinical setting.
\nIts relatively small sample, the involvement of supervising physicians and the controlled research conditions limit how widely the findings can be applied.
\nA diagnostic match rate also does not establish that an AI system makes correct clinical decisions in every situation. Medical assessments involve more than generating possible diagnoses, including examination, testing, treatment decisions and follow-up care.
\nGoogle acknowledges that larger clinical trials are needed before the technology's safety and effectiveness can be assessed at scale.
\nThe potential value is therefore more specific than replacing a doctor. AI may be able to help gather patient histories, prepare clinical summaries and support communication between patients and healthcare professionals.
\nThe workplace and healthcare announcements show Google developing AI systems for environments where accuracy, permissions and accountability matter.
\nIn business settings, organizations will need to decide which tasks agents can complete independently and which actions require human approval.
\nHealthcare presents additional requirements involving patient safety, privacy, clinical validation and professional responsibility.
\nFor organizations in the Western Balkans, these developments provide a useful reference point for future AI adoption. The underlying questions are relevant to local businesses, hospitals and public institutions, even where resources and technical infrastructure differ substantially from those of Google's early enterprise customers and clinical research partners.
\nThe next tests will take place outside product demonstrations and limited research settings. For Gemini, organizations will need to establish whether autonomous workflows provide measurable benefits under real operating conditions. For AMIE, larger clinical studies must determine whether the early findings translate into safe and useful care.
","featured_image_id":362,"intro_image_id":362,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":30,"ordering":0,"published_at":"2026-10-09 08:15:56","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-09 08:15:56","updated_at":"2026-10-09 08:15:56","key_takeaways":"[\"Google introduced a Gemini workplace agent capable of coordinating tasks across business applications and connected systems.\",\"A clinical study involving 98 patients found promising results for Google's experimental AMIE system under physician supervision.\",\"Both developments raise important questions about human oversight, reliability and accountability in AI deployment.\"]","faq":"[{\"q\":\"What is Google's new Gemini agent?\",\"a\":\"An AI system designed to plan and execute multi-step workplace tasks using connected applications and organizational information.\"},{\"q\":\"Did Google's medical AI diagnose patients with 90% accuracy?\",\"a\":\"Not exactly. Google reported that AMIE's differential diagnoses included doctors' final diagnoses in 90% of cases. This is not equivalent to independent diagnostic accuracy.\"},{\"q\":\"Can AMIE replace doctors?\",\"a\":\"No. The system remains experimental, and the published study used physician supervision. Further clinical trials are needed.\"},{\"q\":\"Why do these announcements matter?\",\"a\":\"They demonstrate how AI development is moving toward systems that undertake work directly in business environments and assist with specific clinical activities.\"}]","sources":"[{\"title\":\"Google Cloud, Gemini at Work 2026, October 8\",\"url\":\"https://cloud.google.com/blog/products/ai-machine-learning/welcome-to-gemini-at-work-2026\"},{\"title\":\"Google, AMIE clinical study summary, October 8\",\"url\":\"https://blog.google/innovation-and-ai/technology/health/amie-clinical-study-lancet/\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/62faf2606a5020595a7d583e1a75d7c5.webp","featured_image_alt":"Google Gemini At Work","featured_image_width":1600,"featured_image_height":788,"intro_image_path":"/api/uploads/2026/10/62faf2606a5020595a7d583e1a75d7c5.webp","intro_image_width":1600,"intro_image_height":788,"seo_title":"Google Launches Gemini Agent and Reports Medical AI Study","meta_description":"Google introduces a workplace AI agent and reports results from a clinical study involving 98 patients and its experimental medical AI system.","robots":"index","canonical_url":"google-gemini-agent-medical-ai-study","og_title":"","og_description":""},{"id":157,"title":"Mistral Launches Large 4 as Europe Pushes Open AI","slug":"mistral-launches-large-4-as-europe-pushes-open-ai","short_description":null,"content":"French AI company Mistral has launched the public preview of Mistral Large 4, its largest and most capable model to date, as the European company continues to position itself as a major open-weight alternative to US and Chinese AI developers.
\nThe model was released on October 6 and is available through Mistral Studio and its API. Mistral says the open weights will be released later this month.
\nMistral Large 4 is a multimodal Mixture-of-Experts model with around 1 trillion total parameters and 49 billion active parameters. It supports a context window of up to one million tokens and can work across text, documents, images, coding and agentic workflows.
\nThe launch matters beyond model specifications.
\nMistral has increasingly tied its product strategy to a broader European argument around open models, infrastructure and technological sovereignty. Large 4 strengthens that position at a time when European governments and companies are looking for alternatives to relying entirely on AI systems developed outside the continent.
\nMistral says Large 4 is its strongest model so far for coding, multimodal understanding and agentic tasks.
\nThe company also claims the model is competitive with leading open models globally and performs particularly well in enterprise domains such as cybersecurity, finance and law. These performance claims come from Mistral’s own evaluations and will need broader independent testing as the model becomes more widely available.
\nThe open-weight release expected later this month will be particularly important.
\nOpen-weight models allow organisations to run and adapt models in their own environments rather than relying only on hosted APIs. For companies, governments and research institutions dealing with sensitive data, this can provide greater control over deployment, infrastructure and data handling.
\nThat is one reason open models have become a central part of Europe’s AI debate.
\nMistral has made open-weight development one of its defining positions, while also offering commercial cloud services and enterprise products.
\nLarge 4 arrives as Europe attempts to strengthen several parts of its AI ecosystem at the same time.
\nThe European Union is expanding AI Factory infrastructure around EuroHPC supercomputers, while national governments are increasing investment in sovereign computing capacity, research and AI services.
\nMistral itself has also been expanding its European footprint.
\nIn September, the company announced a new German hub in Munich and said it was increasing its work with European industrial customers. Earlier in the same month, Mistral announced a €3 billion Series D funding round that valued the company at more than €21 billion.
\nThe company has increasingly presented its strategy around three connected elements: open models, European infrastructure and enterprise deployment.
\nLarge 4 fits directly into that approach.
\nFor organisations in the Western Balkans, the significance is less about whether Mistral Large 4 wins a benchmark against another frontier model.
\nThe more practical question is whether increasingly capable European open-weight models can become realistic options for companies, universities and public institutions that want greater control over where their AI systems run and how data is processed.
\nThe availability of open weights can make local or regional deployment possible, but model size also matters.
\nA trillion-parameter model requires substantial computing resources even when only a fraction of its parameters are active during inference. That means Large 4 is unlikely to be something most smaller organisations simply download and operate on ordinary infrastructure.
\nInstead, access may depend on cloud providers, specialised data centres, European AI Factories or other shared computing infrastructure.
\nThis makes the development of models such as Large 4 closely linked to the infrastructure question Europe is already trying to solve.
\nMistral Large 4 is currently available as a public preview through Mistral’s hosted services.
\nAccording to the company, the model’s weights are expected to be released before the end of October.
\nThat release will provide a clearer picture of how developers and organisations can deploy, customise and evaluate the model outside Mistral’s own platform.
\nFor Europe, Large 4 is another test of whether the continent can build AI systems that are not only technically competitive, but also usable across its own research, business and public-sector ecosystem.
","featured_image_id":361,"intro_image_id":361,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":35,"ordering":0,"published_at":"2026-10-07 09:00:36","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-07 09:00:36","updated_at":"2026-10-07 09:00:36","key_takeaways":"[\"Mistral released the public preview of Mistral Large 4 on October 6.\",\"The model has around 1 trillion total parameters and 49 billion active parameters.\",\"It supports multimodal inputs, agentic workflows and a context window of up to one million tokens.\",\"Mistral says open weights will be released later in October.\",\"The launch strengthens Mistral’s position in Europe’s push for open and more sovereign AI infrastructure.\"]","faq":"[{\"q\":\"What is Mistral Large 4?\",\"a\":\"Mistral Large 4 is Mistral’s latest general-purpose multimodal AI model and its largest model released so far.\"},{\"q\":\"Is Mistral Large 4 open source?\",\"a\":\"The model is currently available in public preview. Mistral says the open weights will be released later in October 2026.\"},{\"q\":\"How large is the model?\",\"a\":\"Mistral describes Large 4 as having around 1 trillion total parameters, with 49 billion active parameters using a Mixture-of-Experts architecture.\"},{\"q\":\"What can Mistral Large 4 do?\",\"a\":\"It is designed for text generation, coding, multimodal understanding, document processing and agentic workflows.\"},{\"q\":\"Why does it matter for Europe?\",\"a\":\"Mistral is one of Europe’s leading AI model developers, and Large 4 adds another high-capability open-weight model to Europe’s effort to reduce dependence on external AI providers.\"}]","sources":"[{\"title\":\"Mistral Large 4 announcement\",\"url\":\"https://mistral.ai/news/mistral-large-4/\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/f5859a24aaf0e093b88e318587a04820.webp","featured_image_alt":"Mistral","featured_image_width":1600,"featured_image_height":914,"intro_image_path":"/api/uploads/2026/10/f5859a24aaf0e093b88e318587a04820.webp","intro_image_width":1600,"intro_image_height":914,"seo_title":"Mistral Launches Large 4 as Europe Pushes Open AI","meta_description":"Mistral launches Large 4, a trillion-parameter multimodal model, strengthening Europe’s push for open-weight and sovereign AI.","robots":"index","canonical_url":"mistral-launches-large-4-europe-open-ai","og_title":"","og_description":""},{"id":156,"title":"VEZILKA Deepens AI Factory Cooperation With Greece","slug":"vezilka-deepens-ai-factory-cooperation-with-greece","short_description":null,"content":"North Macedonia’s VEZILKA AI Factory is expanding cooperation with Greece’s PHAROS AI Factory, with the two initiatives discussing how expertise, computing resources and services could be shared more effectively across the European AI Factory network.
\nThe talks took place during a visit to Athens by Gjorgji Madjarov of the Faculty of Computer Science and Engineering at Ss. Cyril and Methodius University in Skopje. Madjarov met Stefanos Kollias, chairman of GRNET and coordinator of PHAROS, as well as GRNET project manager Ilias Hatzakis.
\nAccording to VEZILKA, the discussions covered cooperation in areas of common interest, exchange of technical knowledge and approaches to federated access to computing resources.
\nFor North Macedonia, that last point is particularly important.
\nVEZILKA operates as the country’s AI Factory Antenna, connecting local researchers, companies and institutions with infrastructure available through Europe’s wider network of AI Factories. Unlike a full AI Factory with its own large-scale supercomputing infrastructure, the antenna model is designed to provide national access to expertise, services and computing capacity available elsewhere in the network.
\nPHAROS is one of those connections.
\nThe Greek AI Factory is linked to the DAEDALUS supercomputer and focuses on AI applications in areas including health, language and culture, and sustainability.
\nVEZILKA already identifies PHAROS as a strategic partner and says the cooperation gives organisations in North Macedonia a pathway towards European AI computing infrastructure.
\nThe latest meeting focused not only on infrastructure, but also on how organisations can move from an initial AI idea towards testing and implementation.
\nThat includes questions around how computing resources are requested, how technical support is delivered and how AI Factory services can be made easier to use for researchers, businesses and public institutions.
\nThese issues are becoming increasingly relevant as the European Union expands its AI Factory network around high-performance computing infrastructure.
\nFor smaller countries, building independent large-scale AI computing capacity is expensive. Participation in a shared European network offers another model: local organisations can receive support through national structures while accessing specialised infrastructure elsewhere.
\nVEZILKA says experience from PHAROS could help shape its own services and support mechanisms as the North Macedonian initiative develops.
\nThe project officially started in April 2026 and runs for 36 months. Its total budget is approximately €6.2 million, including around €3.1 million in EU funding through the EuroHPC Joint Undertaking, with additional support from North Macedonia’s Ministry of Digital Transformation and FCSE at UKIM.
\nVEZILKA focuses on five main areas: language and culture, public administration, health, energy and agriculture.
\nThe cooperation also illustrates how AI infrastructure in Southeast Europe is beginning to develop across national borders rather than as isolated projects.
\nNorth Macedonia does not need to reproduce every component of a large AI Factory locally if researchers and companies can gain reliable access to computing resources and expertise through partners such as PHAROS.
\nThe practical test will be whether that model becomes easy enough for local organisations to use.
\nAccess to a European supercomputer matters only if companies, researchers and public institutions can move from application procedures to actual experiments, models and services.
\nVEZILKA and PHAROS are now discussing how that path can work more effectively, including through federated resource provisioning and closer exchange of expertise.
\nAs VEZILKA moves further into its implementation phase, the next measure of progress will be how many organisations in North Macedonia move beyond training and awareness and begin using the infrastructure and technical services available through the European AI Factory ecosystem.
","featured_image_id":360,"intro_image_id":360,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":3,"view_count":27,"ordering":0,"published_at":"2026-10-07 08:54:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-07 08:54:16","updated_at":"2026-10-07 08:55:14","key_takeaways":"[\"VEZILKA and Greece’s PHAROS AI Factory are expanding cooperation around expertise, computing resources and AI services.\",\"Discussions include federated access to resources available through the European AI Factory network.\",\"VEZILKA acts as North Macedonia’s national AI Factory Antenna rather than operating a standalone large-scale AI supercomputer.\",\"Its partnership with PHAROS provides a route towards European high-performance computing infrastructure, including the DAEDALUS supercomputer.\",\"The next challenge is turning infrastructure access into practical AI projects for companies, researchers and public institutions.\"]","faq":"[{\"q\":\"What is VEZILKA?\",\"a\":\"VEZILKA is North Macedonia’s AI Factory Antenna, established within the European AI Factories ecosystem to support companies, researchers, institutions and other organisations working with artificial intelligence.\"},{\"q\":\"What is PHAROS?\",\"a\":\"PHAROS is Greece’s AI Factory and part of the European network of AI Factories supported through EuroHPC.\"},{\"q\":\"Does North Macedonia have its own AI supercomputer?\",\"a\":\"VEZILKA operates as an AI Factory Antenna. Its model connects users in North Macedonia with computing resources, expertise and services available through partner AI Factories rather than replicating all large-scale infrastructure locally.\"},{\"q\":\"What is DAEDALUS?\",\"a\":\"DAEDALUS is Greece’s high-performance computing system and forms part of the infrastructure supporting PHAROS AI Factory.\"},{\"q\":\"Why does the VEZILKA-PHAROS partnership matter?\",\"a\":\"The partnership could make advanced computing infrastructure and AI expertise more accessible to researchers, businesses and public institutions in North Macedonia.\"}]","sources":"[{\"title\":\"VEZILKA Strengthens Cooperation with Greek AI Factory – PHAROS\",\"url\":\"https://vezilka.ai/vezilka-strengthens-cooperation-with-pharos-ai-factory/\"}]","category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/932fa0e513dea4f2895e1d61d4c5fa76.webp","featured_image_alt":"Vezilka Greece Meeting","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/932fa0e513dea4f2895e1d61d4c5fa76.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"VEZILKA Expands AI Factory Cooperation With Greece","meta_description":"VEZILKA and Greece’s PHAROS are expanding cooperation on AI expertise, computing resources and access to European AI infrastructure.","robots":"index","canonical_url":"vezilka-deepens-ai-factory-cooperation-with-greece","og_title":"","og_description":""}],"most_read":[{"id":123,"title":"When AI Betrays Law and Justice: Skopje Conference Brings Legal and AI Experts Together","slug":"when-ai-betrays-law-and-justice-skopje-conference-brings-legal-and-ai-experts-together","short_description":"","content":"Skopje will host an international scientific conference on 16 September bringing together legal scholars, researchers, digital policy experts and technology practitioners to examine how artificial intelligence is affecting law, human rights, constitutionalism and democratic accountability.
\n![\"Conference_WHEN](\"/api/uploads/2026/09/6e16c36b3fccfcb59293184a1583b896.webp\")
The conference, “When AI Betrays Law and Justice: Human Rights, Constitutionalism and Democratic Accountability in the Algorithmic Age,” will take place at Hotel Panoramika and is organised by the Faculty of Law “Iustinianus Primus” at Ss. Cyril and Methodius University in Skopje.
\nIt is being held within the research project “Alignment of the Macedonian Legal System with the European Union Artificial Intelligence Legislation,” funded by North Macedonia’s Ministry of Education and Science.
\nThe event brings together an international group of speakers alongside academics and researchers from North Macedonia, with the programme moving from broader questions of democracy and justice to new empirical findings on how AI is understood and used in the country.
\n![\"Conference_WHEN](\"/api/uploads/2026/09/930938b08df737aac4efa72c7e5a4790.webp\")
Among the international speakers is Paul Nemitz, former Principal Adviser for Strategies for Digital Transition at the European Commission and Visiting Professor of Law at the College of Europe.
\nNemitz will address the relationship between Big Tech, artificial intelligence, democracy and the rule of law.
\nJudit Varga, Founding Patron of the Central European Academy at the University of Miskolc, will examine algorithmic decision-making and its implications for constitutional government.
\nThe international programme also includes Laura Kask, CEO of Estonian digitalisation consultancy Proud Engineers, whose contribution will focus on trust in digital societies and the challenges of designing institutions that remain accountable as digital systems become more influential.
\nMax Hubner, Founder and CEO of The Change Lawyer in the Netherlands, will focus on the relationship between artificial intelligence, people and justice.
\nTheir contributions will be joined by perspectives from academics at the Faculty of Law “Iustinianus Primus”.
\n![\"Conference_WHEN](\"/api/uploads/2026/09/a5bd74f18cc977a74141b2fe7ad27a9e.webp\")
Professor Aleksandar Spasov will examine how AI affects fundamental concepts in law, while Professor Boban Misoski will address procedural justice in criminal trials and the possible role of artificial intelligence in judicial processes.
\nProfessor Lazar Jovevski will focus on AI, labour markets and the wider consequences of technological change for people and work.
\nTogether, these discussions place AI within a legal and institutional context rather than treating it only as a technological development.
\nThe central issue is increasingly one of responsibility. As AI-supported systems enter public administration, employment, business and legal processes, institutions need to determine where automated decision-making is appropriate, how those decisions can be challenged and who remains accountable when something goes wrong.
\n![\"Conference_WHEN](\"/api/uploads/2026/09/54abe6789af5d91677f55df88d19c18d.webp\")
A central part of the conference will be dedicated specifically to AI in North Macedonia, with the presentation of new empirical findings examining public perceptions, corporate practices and institutional readiness.
\nThe research team is coordinated by Professor Tanja Karakamisheva-Jovanovska from the Faculty of Law “Iustinianus Primus”.
\nThe team also includes Professor Hristina Runcheva Tasev, Professor Dragan Gocevski, Assistant Professor Konstantin Bitrakov and Assistant Katerina Zhateva, all from the Faculty of Law.
\nThe research section will also include expert contributions from Professor Vladimir Bozhinovski and Suad Seferi, President and Founder of AI NOW Society.
\nThe study is particularly relevant because much of the discussion around AI in North Macedonia and the wider Western Balkans has so far focused on strategies, regulation, digital transformation and adoption.
\nEmpirical research can provide a different picture.
\nIt can show how people actually perceive AI, how companies are approaching its use and whether institutions have the structures, knowledge and capacity required to manage it responsibly.
\nThat becomes increasingly important as North Macedonia works toward closer alignment with European Union legislation while AI systems are already entering workplaces, universities, companies and public institutions.
\nThe regional dimension of the discussion will also be addressed through “Beyond Adoption: AI Readiness in the Western Balkans,” presented by Suad Seferi.
\nThe contribution will look beyond the number of organisations using AI and toward a wider question: whether institutions across the Western Balkans are genuinely prepared to manage the technology.
\nAI adoption and AI readiness are not necessarily the same thing.
\nAn organisation may already use generative AI tools while lacking internal policies governing their use. A company may introduce automation without clearly assigning responsibility for its outputs. Public institutions may experiment with AI before developing procedures for transparency, human oversight or risk assessment.
\nReadiness therefore involves more than access to technology. It includes institutional capacity, governance, skills, legal understanding, infrastructure and the ability to manage risk.
\nFor Western Balkan countries, these issues are closely connected with European integration and the gradual alignment of domestic frameworks with EU digital and AI regulation.
\nThe conference is being led by academics from the Faculty of Law “Iustinianus Primus”, with Tanja Karakamisheva-Jovanovska serving as project coordinator. Opening remarks will also include Sasho Georgievski, Dean of the Faculty of Law.
\nThe wider group involved in the event reflects the different areas now being affected by AI: constitutional law, criminal justice, labour, public administration, digitalisation, institutional governance and technology policy.
\nThat multidisciplinary approach matters because many of the hardest questions surrounding AI cannot be answered by engineers alone.
\nWhen an automated system affects employment, legal rights, access to services or institutional decisions, technical performance is only one part of the problem. Law, accountability and human responsibility remain equally important.
\nFor North Macedonia, the conference also creates an opportunity to connect European discussions around AI regulation with evidence about what is actually happening inside the country.
\nThe findings presented in Skopje could help move the local debate away from general discussions about AI adoption and toward more specific questions about institutional preparedness, governance and responsibility.
\n“When AI Betrays Law and Justice” will take place on 16 September 2026 at Hotel Panoramika in Skopje.
\n![\"Conference_WHEN](\"/api/uploads/2026/09/e473d77baaa964fdb12ae45567b311c3.webp\")
Participation is free, with places limited. Registration is open until 12 September 2026.
","featured_image_id":269,"intro_image_id":269,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":5,"view_count":162,"ordering":0,"published_at":"2026-09-04 12:56:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-04 12:56:03","updated_at":"2026-09-04 12:56:53","key_takeaways":null,"faq":null,"sources":null,"category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/f4feaa8c83cf7349d9d43e70474b2e10.webp","featured_image_alt":"Poster for “When AI Betrays Law and Justice” with Lady Justice, an AI face, and Skopje’s riverfront.","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/09/f4feaa8c83cf7349d9d43e70474b2e10.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Skopje Conference Examines AI, Law and Accountability","meta_description":"International experts and Macedonian researchers meet in Skopje to discuss AI, law, human rights and new research on North Macedonia’s AI readiness.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":143,"title":"Serbian Teachers Test AI Assistant Built for Classrooms","slug":"serbian-teachers-test-ai-assistant-built-for-classrooms","short_description":"","content":"Serbian education nonprofit Petlja Foundation is testing an artificial intelligence assistant designed specifically for teachers, as schools experiment with ways to use AI for lesson preparation, classroom materials and student projects.
\nThe Petlja AI Assistant is built on Microsoft Azure and Azure OpenAI Service and is currently available to around 250 teachers in Serbia. Between 50 and 60 use it actively, according to Petlja director Milan Simić.
\nRather than giving students a general-purpose chatbot, the project focuses primarily on supporting teachers.
\nOne of those teachers is Sanja Rajić, who teaches mathematics and programming at the Palanka Gymnasium in Bačka Palanka. She began testing an early version of the assistant in 2023.
\nRajić uses the tool to prepare lesson structures, generate exercises for different levels of knowledge and create variations of tests. Formal lesson preparation that could previously take hours can be produced as a starting point by the assistant and then reviewed and adapted by the teacher.
\n![\"Petlja](\"/api/uploads/2026/09/91bac1aacac66549f697e079968bb523.webp\")
That distinction is central to how the system is being used. Teachers remain responsible for checking the generated material and deciding what is appropriate for their students.
\nThe tool has also found its way into practical student projects.
\nStudents at the Palanka Gymnasium have used the assistant while working on robotics, programming and computer vision projects. In one case, students building a mechanical finger used it to help troubleshoot a 3D-printed component when they could not find the required screw locally.
\nTwo students from the school also worked on BrAInUp, an educational well-being application created for Serbia's 2026 Student Companies Competition, with Rajić and the AI assistant supporting parts of the technical development.
\nPetlja was founded in 2017, the same year Serbia introduced programming as a mandatory subject from fifth grade.
\nThe foundation initially focused on creating programming and computer science materials that students could use directly through a web browser. Its platform has since expanded into teacher training, digital education and AI-supported learning.
\nPetlja reports more than 100,000 active users over the past three months, with more than 18,000 course participants during the current school year and more than 50 courses available in Serbian.
\nIts work with Serbia's education system includes teacher training, materials for the Digital World subject, resources for specialised IT classes and development of AI tools for education.
\nThese activities are carried out through projects involving Serbia's Ministry of Education, the United Nations Development Programme and other partners.
\nMicrosoft has supported Petlja through technology, expertise and financial assistance, while Azure provides the infrastructure used by the current AI Assistant.
\nRunning an AI assistant for a large number of teachers creates another problem: cost.
\nEvery interaction with a commercial AI model carries an operating expense. Petlja is therefore working through the UNICEF TLM4EL project on an open-source model adapted for Serbian education.
\nThe plan is to eventually run the model on donated hardware rather than relying entirely on commercial cloud AI services.
\nPetlja says this could lower operating costs and allow more of the system's data to remain under local control. There is not yet a confirmed timeline or funding arrangement for a large-scale rollout.
\nThat part of the project could prove more significant than the assistant itself.
\nFor education systems across the Western Balkans, introducing generative AI is only one part of the problem. Schools also have to decide who controls the infrastructure, where educational data is processed, how tools work in local languages and who is responsible for checking their output.
\nPetlja's experiment provides one model: start with teachers, keep human review in the process and build the technology around the education system rather than introducing a general AI tool and expecting schools to adapt to it.
\nThe next test will be whether that model can move beyond several dozen active users and operate sustainably across a much larger part of Serbia's school system.
","featured_image_id":314,"intro_image_id":314,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":11,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":149,"ordering":0,"published_at":"2026-09-26 09:41:32","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-26 09:41:32","updated_at":"2026-09-26 09:41:32","key_takeaways":"[\"Around 250 Serbian teachers currently have access to Petlja's AI Assistant, with 50 to 60 using it actively.\",\"Teachers are using the system for lesson preparation, exercises, tests and student technology projects, while reviewing and adapting its output.\",\"Petlja is also working on an open-source model for Serbian education that could eventually run locally and reduce dependence on commercial AI services.\"]","faq":"[{\"q\":\"What is the Petlja AI Assistant?\",\"a\":\"It is an AI tool developed by Serbia's Petlja Foundation to support teachers with lesson planning, educational materials and other classroom tasks.\"},{\"q\":\"How many teachers are using it?\",\"a\":\"Around 250 teachers currently have access. Petlja says approximately 50 to 60 are active users.\"},{\"q\":\"Does the AI replace teachers?\",\"a\":\"No. The current model is designed as a support tool. Teachers are expected to review, modify and approve generated materials before using them in class.\"},{\"q\":\"Is Petlja developing its own AI model?\",\"a\":\"Petlja is working through the UNICEF TLM4EL project on adapting an open-source model for Serbian education, with plans for local deployment on donated hardware.\"}]","sources":"[{\"title\":\"Microsoft published the detailed Petlja case study on 24 September 2026\",\"url\":\"https://news.microsoft.com/source/emea/features/petlja-microsoft-ai-serbian-classrooms/\"}]","category_name":"Serbia","category_slug":"serbia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/da3c6e043da6db9d73c50cda0c5179c2.webp","featured_image_alt":"A teacher speaks to students beside a laptop and document camera in a bright classroom technology lab.","featured_image_width":1600,"featured_image_height":1069,"intro_image_path":"/api/uploads/2026/09/da3c6e043da6db9d73c50cda0c5179c2.webp","intro_image_width":1600,"intro_image_height":1069,"seo_title":"Serbian Teachers Test AI Assistant Built for Classrooms","meta_description":"Serbia's Petlja is testing an AI assistant for teachers and developing a locally hosted model designed for the country's education system.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":151,"title":"Startup Revolution Begins in Skopje: What to Expect","slug":"startup-revolution-begins-in-skopje-what-to-expect","short_description":null,"content":"Startup Revolution AI Summit begins in Skopje on October 1, bringing four days of artificial intelligence, startup building, investment, policy discussions, mentoring, pitching and networking across several locations in the city.
\nThe programme opens on Thursday evening before moving into two full conference days at the Macedonian Philharmonic on October 2 and 3. The summit closes on Sunday alongside the Wizz Air Skopje Marathon.
\nMore than 100 speakers and hundreds of startups are expected to participate, according to the organisers. Rather than following a single conference track, Startup Revolution is structured around different audiences, from first-time founders and scaleups to investors, policymakers, corporate leaders and students.
\n![\"A](\"/api/uploads/2026/09/442d17c5e6debf3d6b592232a1f3e899.webp\")
October 1 functions primarily as the summit’s arrival and networking day.
\nPremium-pass participants have an afternoon side event outside Skopje, including a visit and wine tasting at Stobi.
\nThe wider summit programme begins in the evening at Srekja, home of Startup Club Skopje, with the official welcome reception and opening party. This is the first networking opportunity before the conference moves to the Macedonian Philharmonic the following morning.
\nFriday, October 2, is the first full conference day.
\nRegistration, coffee and networking begin at 09:00, followed by the official opening ceremony at 10:00.
\nThe programme then splits across several themes, including AI and emerging technologies, ecosystem and data, energy and climate tech, fintech, healthtech, the founder journey and early-stage startup development.
\nFor people building their first company, Friday has a particularly practical focus.
\nSessions cover how founders find their first customers, build sales processes, use mentors effectively, position an early-stage company and avoid common legal mistakes. Other parts of the programme deal with bootstrapping, going from an idea to an operating startup and preparing a startup pitch.
\nA practical masterclass will walk founders through the process of getting their first customer, while another session examines legal mistakes startup founders should avoid.
\nThere will also be a discussion with early-stage founders about bootstrapping and the decisions companies face before they have significant external funding.
\nStartup Revolution is not limited to stage presentations.
\nFriday includes hands-on sessions and workshops, including a founder resilience workshop focused on dealing with rejection and continuing towards a first customer or investment.
\nOne-to-one mentoring sessions are also scheduled with mentors connected to the Swiss Entrepreneurship Program and other summit partners.
\nLater in the day, founders will work directly on pitching, including a practical session on how startup pitches are structured and delivered.
\nInvestment and informal networking continue after the main sessions.
\nAn Investor Kafana is scheduled in the evening for Premium participants, followed by additional side events, including a programme centred on discussing startup failure and a late-night Game Night.
\nThe format reflects a wider part of Startup Revolution’s structure: meetings between founders and investors are spread throughout the summit instead of being confined to one investment panel.
\nSaturday, October 3, shifts the emphasis.
\nThe main tracks include government and policy, AI transformation and scaleups, trust and cybersecurity, fintech, marketing and the creator economy, business and investment.
\nThe morning begins with a policy-focused programme involving representatives from North Macedonia and the wider region.
\nSessions include discussions about building competitive economies in the age of AI, AI policy, public support for startups and the role of innovation institutions.
\nA regional panel will bring together representatives connected to innovation organisations in North Macedonia, Serbia and Montenegro to discuss policies, innovation funding and public-sector support for startup growth.
\nThere will also be a session examining cultural diplomacy and the startup ecosystem, with participants from organisations including the Serbian Chamber of Commerce and the Macedonian E-commerce Association.
\nOne of Saturday morning’s specific announcements is the special launch of the Skopje Innovation District, scheduled for 10:45.
\nThe programme lists professors Vladimir Trajkovikj and Radmil Polenakovikj for the launch.
\nThe initiative adds a local ecosystem component to a programme otherwise heavily focused on regional and international startup development.
\nTechnology inside government also has its own place in the programme.
\nNorth Macedonia’s Minister of Digital Transformation Stefan Andonovski is scheduled for a session on digital identity, digital trust and the next generation of public services.
\nGovernment-supported startups will also be presented during the policy track.
\nAnother session turns towards sovereign AI for businesses, adding questions around control of AI systems and organisational data to the programme.
\nOne of Saturday’s strongest AI-focused sections is the AI Transformation & Scaleup Track.
\nRather than concentrating only on new models or AI products, the sessions deal with what happens when companies try to integrate AI into actual operations.
\nA case study will examine how AI was introduced into business operations, followed by a masterclass on moving from isolated AI pilots towards what the programme describes as AI-native operations.
\nAnother panel asks a question increasingly relevant to startups: if AI tools make it easier for almost anyone to build a product, what separates the companies that successfully reach the market?
\nThe scaleup programme also covers leadership, marketplace growth, access to European funding and the operational changes founders face as their companies grow.
\nSaturday afternoon turns further towards financing and scale.
\nSessions examine what founders should protect during negotiations, how companies can access European funding without immediately giving up equity and what happens during legal and investment due diligence before money changes hands.
\nThe programme also brings founders, investors and policymakers together during a dedicated brunch and networking period.
\nFor more established companies, these sessions move beyond the early-stage question of how to launch and towards how to finance and structure the next stage of growth.
\nTrust and cybersecurity form another part of Saturday’s agenda.
\nThe track brings security and privacy into the broader conversation about AI adoption, alongside sessions involving academic and industry participants.
\nFINKI is also represented in the programme, linking the summit with North Macedonia’s academic and technical community.
\nSaturday also brings one of the summit’s main pitching competitions.
\nThe Startup World Cup programme begins in the afternoon, followed by a pitching final involving selected startups.
\nCompanies listed in the competition include Intrascope.app, VNYX B.V., Telentir, Developico/InSight, Memoria Global, Safe Watch App, Athena Chat, Factum AI, SupraOS and Impala Research.
\nEach company receives a short pitching window followed by questions before the jury deliberates and announces the winner.
\nThe competition connects the Skopje event with Startup World Cup’s international startup competition.
\nStartup Revolution also runs a separate programme away from the main Philharmonic stages.
\nAt INNOX, student teams participating in the I Have an Idea 4 incubation programme will spend Saturday working with mentors, preparing presentations and pitching startups they have developed and validated.
\nThe programme begins with one-to-one mentoring before moving into pitch preparation.
\nThirty teams are scheduled across two pitching blocks in the evening.
\nThis creates a second startup pipeline within the summit, aimed at students and new founders rather than more mature companies competing in the Startup World Cup.
\nSeveral smaller formats run alongside the main stages.
\nThe Founder Games Live Challenge takes place on Saturday morning, while a Premium workshop focuses on turning personal expertise and professional visibility into revenue.
\nCombined with mentoring, pitching and founder-focused sessions, these activities give the summit a practical layer alongside the main conference programme.
\nSaturday’s formal programme does not end with the final pitch.
\nAn invite-only Women’s Mixer brings together founders, investors and government representatives in the evening.
\nThe final night then moves back to Srekja for the Unicorn Party, which the organisers position as the summit’s closing social event before Sunday’s final activities.
\nStartup Revolution closes differently from a traditional technology conference.
\nOn Sunday morning, the programme connects with the Wizz Air Skopje Marathon, which begins at 08:00 near the Assembly of North Macedonia.
\nParticipants can run, walk or attend the race before Startup Revolution formally closes in the afternoon.
\nThe summit therefore moves across three main locations during its four days: Srekja and Startup Club Skopje for the opening and evening events, the Macedonian Philharmonic for the main conference on Friday and Saturday, and central Skopje for Sunday’s marathon programme.
\nFor visitors trying to navigate the schedule, Friday and Saturday contain most of the summit’s professional programme.
\nFriday is particularly relevant for early-stage founders, people developing their first startup, mentors and those interested in specialised sectors such as fintech, healthtech and climate tech.
\nSaturday broadens the conversation towards AI transformation, scaleups, investors, public policy, cybersecurity, government digitalisation and startup financing, while also hosting the two major pitching programmes.
\nThursday and Sunday provide the opening and closing around those two conference days.
\nWith simultaneous stages, workshops, mentoring sessions, investor meetings and side events, attending Startup Revolution will require some choices. The full agenda is available through the summit’s official website, where participants can plan the tracks most relevant to them.
","featured_image_id":374,"intro_image_id":374,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":8,"view_count":147,"ordering":0,"published_at":"2026-10-01 19:29:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-01 19:29:45","updated_at":"2026-10-12 21:28:05","key_takeaways":"[\"Startup Revolution runs from October 1 to 4 in Skopje, with the main conference programme concentrated on October 2 and 3.\",\"Friday focuses heavily on early-stage startups, founders, AI and emerging technology, fintech, healthtech, climate tech, mentoring and finding first customers.\",\"Saturday moves towards AI transformation, scaleups, investment, government policy, cybersecurity and startup financing.\",\"The programme includes the launch of the Skopje Innovation District, Startup World Cup pitching and pitches from 30 student teams in the I Have an Idea 4 programme.\",\"Workshops, one-to-one mentoring, investor meetings, networking and side events run alongside the main conference stages.\"]","faq":"[{\"q\":\"When does Startup Revolution AI Summit 2026 take place?\",\"a\":\"Startup Revolution runs from October 1 to October 4, 2026, in Skopje, North Macedonia.\"},{\"q\":\"Where is the main Startup Revolution programme held?\",\"a\":\"The main conference programme on October 2 and 3 takes place at the Macedonian Philharmonic. Opening events are held at Srekja and Startup Club Skopje, while the final day connects with the Wizz Air Skopje Marathon.\"},{\"q\":\"What can startup founders expect?\",\"a\":\"The programme includes founder sessions, sales and go-to-market workshops, legal guidance, mentoring, investor networking, pitch preparation, scaleup sessions and startup competitions.\"},{\"q\":\"Is Startup Revolution only focused on artificial intelligence?\",\"a\":\"No. AI is a major part of the programme, but the summit also covers startups, investment, fintech, healthtech, energy and climate technology, cybersecurity, public policy, digital government, marketing and company growth.\"}]","sources":"[{\"title\":\"Startup Revolution AI Summit — Official Agenda\",\"url\":\"https://startuprevolution.ai/pages/agenda\"},{\"title\":\"Startup Revolution AI Summit — Venue Information\",\"url\":\"https://startuprevolution.ai/pages/venue\"}]","category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/fcb5754a231cf2daafe44e37f1a61c00.webp","featured_image_alt":"Startup revolution","featured_image_width":1600,"featured_image_height":1066,"intro_image_path":"/api/uploads/2026/10/fcb5754a231cf2daafe44e37f1a61c00.webp","intro_image_width":1600,"intro_image_height":1066,"seo_title":"Startup Revolution Begins in Skopje: What to Expect","meta_description":"Startup Revolution opens in Skopje with four days of AI, startups, investment, policy, pitching and business-focused sessions.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":98,"title":"Monika on Responsible AI, Women in Tech and the Balkans","slug":"monika-on-responsible-ai-women-in-tech-and-the-balkans","short_description":"","content":"The Women in Tech leader speaks with AI Balkans about leadership, responsible artificial intelligence, regional collaboration and the importance of keeping people at the centre of technological progress.
\nArtificial intelligence is transforming how people work, learn and build organizations, but its development also raises deeper questions about responsibility, trust and the role of human connection.
\n![\"Monika.jpg\"](\"/api/uploads/2026/08/80da13b7d01f3842c2a55721b82f45d4.webp\")
Monika has dedicated much of her professional journey to empowering women, building communities and supporting the practical adoption of artificial intelligence. Her work brings together technology, leadership, mentorship and a belief that progress should ultimately improve people’s lives.
\nIn this interview with AI Balkans, she reflects on the lessons that shaped her career, the people who inspire her and the barriers women continue to face in technology. She also discusses the opportunities artificial intelligence creates for the Balkans, the importance of AI literacy and why organizations must begin with real problems rather than simply adopting the latest tools.
\nLooking back, what would you tell the younger Monika who was just starting her career in technology? Is there one piece of advice you wish someone had given you earlier?
\nLooking back, I would tell the younger Monika to stop waiting until she felt ready. Confidence comes after action, not before.
\nEarly in my career, I believed I needed another certificate, another degree, success or many years of experience before speaking up. Today, I know that opportunities belong to people who raise their hand, who are willing to take action, experiment, and seize the moment when it appears.
\nI would also tell her to build a strong personal board of advisors. Surround yourself with people who challenge you, support you, and inspire you to think bigger. The people you spend the most time with shape your mindset and your future.
\nRelationships matter as much as professional success. Your network is your net worth. A strong reputation, genuine curiosity, and a willingness to help others will open more doors than any job title or diploma ever will.
\nYou spend a lot of your time empowering other women. But who empowers you? Who has been the person, or the moment, that changed your own life the most?
\nI am inspired by people who choose action over excuses. Throughout my journey, I have been fortunate to meet mentors, founders, and community leaders from around the world who challenged me to think bigger.
\nBecoming a mother changed me the most. My children completely shifted my perspective on leadership, success, and impact. They remind me every day why I want to build a future where technology serves people, not the other way around.
\nBeyond technology and leadership, what keeps you grounded? What do you do when you need to disconnect, reflect, or simply remind yourself why you started this journey?
\nMy family keeps me grounded. Spending time with my children reminds me that the most meaningful impact starts at home.
\nI also enjoy traveling, long conversations with inspiring people, painting, reading about innovation, and spending time by the sea. Some of my best ideas come when I step away from my laptop and give myself space to think.
\nAI Balkans was recently launched as the first regional AI hub dedicated to artificial intelligence in the Western Balkans. How important is it for the region to finally have a platform that tells its own AI story instead of relying only on global narratives?
\nI believe this is an important milestone for our region.
\nOur Balkan has talented researchers, entrepreneurs, startups, educators, and engineers. What we have often lacked is visibility and collaboration. AI Balkans creates a platform where knowledge, success stories, and innovation from our region become visible to the world.
\n![\"Monika](\"/api/uploads/2026/08/208602ca935579627549ffafb122b7d6.webp\")
Instead of asking how to copy Silicon Valley, we should ask how to solve Balkan challenges with AI. Our healthcare systems, public administration, education, agriculture, and SMEs have unique needs. Regional collaboration gives us a stronger voice and creates opportunities that no single country would achieve alone.
Artificial intelligence is advancing faster than regulation and public understanding. In your opinion, what is the biggest ethical challenge we should be discussing today, not five years from now, but right now?
\nThe biggest challenge is not whether AI becomes more intelligent. The biggest challenge is whether humans remain responsible for the decisions AI influences.
\nWe need transparency. People deserve to know when AI is making recommendations, how data is used, and who is accountable when something goes wrong. We also need to invest in AI literacy because people cannot question systems they do not understand.
\nTechnology moves fast. Trust takes much longer to build.
\nMany organizations are rushing to adopt AI simply because it is trendy. Do you think we are using AI strategically in the Balkans, or are we still scratching the surface of what it can truly achieve?
\nWe are still scratching the surface.
\nMany organizations start with AI tools instead of business problems. They ask, \"Which AI platform should we buy?\" instead of asking, \"Which process wastes the most time?\"
\nThe organizations that succeed treat AI as a business transformation initiative, not an IT project. The greatest opportunity for the Balkans is helping SMEs become more productive, competitive, and innovative through practical AI adoption. And this is what I am focussed on in this last year with Davleo Consulting as an AI transformation advisor.
\nDespite decades of progress, many women still encounter invisible barriers in leadership. In your experience, has it become easier for women to build careers in technology, or have the obstacles simply become more subtle?
\nThere are more opportunities today than ever before, but many barriers still exist. Women are often expected to prove their expertise repeatedly, while confidence is rewarded differently. Women continue to carry a larger share of family responsibilities, which affects career progression.
\nThe good news is that communities like Women in Tech and AI Collective are changing this. When women have access to mentors, networks, and role models, they stop feeling like they have to succeed alone.
\nIf someone asked you what your greatest professional achievement is, would you name an award, an organization, or a person whose life changed because of your work?
\nWithout hesitation, I would choose the people.
\n![\"Monika](\"/api/uploads/2026/08/b6370d46b69b77ad4dfe5868d7e43ee0.webp\")
Awards are meaningful because they recognize hard work, but the moments that stay with me are seeing positive outcome from my work, hearing feedback such as: \"Your workshop gave me the confidence to change careers,\" or, \"Because of your mentorship, I applied for a leadership role and got the job.\"
Knowing that I have helped thousands of women build AI skills, grow their confidence, and create new opportunities for themselves is the achievement I value most. Every career change, promotion, business launched, or dream pursued because someone believed in their own potential is a reminder of why I do this work.
\nImpact is not measured by the awards on our shelf. It is measured by the lives we help change. That will always mean more to me than any trophy.
\nWe often celebrate success but rarely talk about failure. What has been the most difficult lesson of your career, and how did it shape the leader you are today?
\nOne of the hardest lessons was learning that passion alone does not build sustainable organizations.
\nEarly on, I said yes to too many projects because I wanted to help everyone. Over time, I learned that leadership is also about setting priorities, building strong teams, trusting others, and saying no when necessary.
\nToday, I focus less on doing everything myself and more on creating systems that help other people succeed.
\nImagine it's 2035. Artificial intelligence has transformed the way we work, learn, and live. What is the one thing you hope technology will never replace?
\nHuman connection.
\nAI will transform productivity, education, healthcare, and business. I welcome that future.
\nBut empathy, curiosity, meaningful conversations, ethical judgment, and the ability to inspire another person should always remain uniquely human. Technology should help us become more human, not less. That is the future I want to help build.
\nAbout Monika
\nMonika is a Women in Tech leader, mentor and AI transformation advisor whose work focuses on empowering women, strengthening technology communities and helping organizations approach artificial intelligence as a practical business transformation opportunity. Through her work with Davleo Consulting, Women in Tech and AI Collective, she supports skills development, leadership and responsible AI adoption.
","featured_image_id":230,"intro_image_id":230,"featured_image_credit":"","featured_image_overlay":"Monika Rizovska Tanevska","featured_image_overlay_position":"bottom-left","category_id":27,"author_id":7,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":7,"view_count":110,"ordering":0,"published_at":"2026-08-04 09:26:00","publish_down":null,"created_by":5,"updated_by":5,"created_at":"2026-08-04 09:33:48","updated_at":"2026-08-04 09:38:46","key_takeaways":null,"faq":null,"sources":null,"category_name":"Interviews","category_slug":"interviews","author_name":"Aleksandra Todorovikj","author_slug":"aleksandra-todorovikj","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/79098ff2e7c8d1312618c7472b89cae8.webp","featured_image_alt":"A woman in a white blazer looks toward a sunlit window with blinds.","featured_image_width":583,"featured_image_height":371,"intro_image_path":"/api/uploads/2026/08/79098ff2e7c8d1312618c7472b89cae8.webp","intro_image_width":583,"intro_image_height":371,"seo_title":"Monika on AI, Leadership and Women in Technology","meta_description":"Women in Tech leader Monika discusses responsible AI, leadership, regional collaboration and opportunities for women in technology across the Balkans.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":102,"title":"Slovenia Launches GaMS AI Model","slug":"slovenia-launches-gams-ai-model","short_description":"","content":"Slovenia has developed an open-source large language model designed specifically for the Slovene language, giving researchers, companies and public institutions a locally deployable alternative to relying entirely on global AI platforms.
\nThe University of Ljubljana's Faculty of Computer and Information Science presented GaMS, a family of generative AI models adapted for Slovene and intended for use in areas including healthcare, industry, cultural heritage and software development. The models can be downloaded and installed locally, while a public chatbot is also available through the project's website.
\nFor Slovenia, the project addresses a problem familiar to many smaller European countries. The largest AI systems are developed using enormous datasets dominated by English and other widely spoken languages. Languages with smaller digital footprints often receive less representation in those datasets.
\nGaMS shows one way that smaller language communities can respond without attempting to build a frontier AI model entirely from scratch.
\nGaMS was developed through the PoVeJMo project at the University of Ljubljana, coordinated by researchers working on language resources and technologies for Slovene.
\nOne of the project's central challenges was collecting enough high-quality material.
\nSlovene has around two million speakers, making the available pool of digital text much smaller than that available for English. Researchers therefore invited Slovene speakers to contribute their own writing through the povejmo.si platform.
\nTraining material was also contributed by organisations including Slovenia's National and University Library, the newspaper Dnevnik and the Slovenian Press Agency.
\nRather than creating the underlying technology from zero, researchers adapted existing open models.
\nA research paper published in March describes GaMS3-12B, a 12-billion-parameter model developed through continued pre-training of Google's Gemma 3 model followed by supervised fine-tuning.
\nThe training mixture contained around 140 billion Slovene, English, Bosnian, Serbian and Croatian tokens, followed by more than 200,000 Slovene and English instruction examples.
\nThe inclusion of neighbouring South Slavic languages is particularly relevant for Southeast Europe. It shows how linguistic resources from related languages can contribute to the development of AI for a language with a smaller digital footprint.
\nAccording to the research team, GaMS3-12B performed better than the original 12-billion-parameter Gemma 3 model across the Slovene evaluation scenarios used in the study.
\nResearchers tested the model on Slovene benchmarks, English-to-Slovene translation and a Slovene LLM arena where users compare responses produced by different models.
\nThe paper reports that GaMS3-12B achieved a win rate above 60 percent in the Slovene arena and performed comparably to the much larger commercial GPT-4o model in that particular evaluation.
\nThat comparison should be interpreted carefully. It does not mean GaMS performs at the same level as frontier commercial systems across every task or language.
\nThe more useful result is that focused training significantly improved an existing open model's performance for Slovene. That is an important distinction for countries that do not have the computing infrastructure or financial resources required to train the world's largest AI systems.
\nThe project has also moved beyond academic evaluation.
\nFour Slovenian companies have adapted GaMS for different applications.
\nHealthcare technology company Better has worked with the model in medical and clinical environments. Špica has explored industrial speech recognition, including situations involving background noise, dialects and multilingual workers.
\nSemantika has worked on applications for museums and cultural heritage, while XLAB has explored the model's use in generating descriptions and documentation related to computer infrastructure.
\nLocal deployment is another important part of the project.
\nThe GaMS models are openly available for organisations, researchers and developers to install in their own environments and adapt for specialised applications. The project's open-access catalogue currently includes several versions, including GaMS3-12B, GaMS3-12B-Instruct and other instruction-tuned variants.
\nRunning a language model locally can also allow organisations to keep sensitive information within their own infrastructure rather than automatically sending data to external AI services. The University of Ljubljana has highlighted this data-control aspect as one of the reasons for developing domestic AI infrastructure.
\nSlovenia's experience raises a larger question for the region: who will build AI that properly understands Europe's smaller languages?
\nMacedonian, Albanian, Bosnian, Serbian, Croatian and Montenegrin operate in the same global AI market, where the quantity of available training material and investment is heavily concentrated around larger languages.
\nCommercial AI platforms increasingly work across many of these languages, but language support alone does not guarantee the same quality of understanding, specialised vocabulary or locally relevant knowledge available in English.
\nGaMS offers a different approach.
\nInstead of trying to compete directly with global technology companies by training enormous foundation models, universities and public institutions can adapt capable open models using carefully prepared local datasets.
\nFor parts of the Western Balkans, regional cooperation could make that approach more practical.
\nThe GaMS researchers have already used Bosnian, Croatian and Serbian material alongside Slovene during training. Shared language resources, evaluation frameworks and computing infrastructure could similarly help countries with limited datasets build stronger local AI systems.
\nThere are still difficult questions around copyright, data quality, funding and long-term maintenance. Building a language model is also only the beginning. Models need continuous evaluation, updated datasets and applications that justify the investment.
\nSlovenia is already extending that work. University researchers are developing further Slovene-language AI tools, including an English-Slovene translator based on GaMS, while the country's AI Factory is providing support for organisations that want to deploy and adapt local models.
\nFor the Western Balkans, that may be the most important part of the Slovenian example.
\nThe question is no longer whether smaller European languages can have their own capable generative AI models. Slovenia has shown that they can.
\nThe next question is which countries in the region will decide that their languages deserve the same investment.
\nReaders who want to explore the project, try the available tools or learn more about GaMS can visit povejmo.si.
","featured_image_id":238,"intro_image_id":238,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":5,"view_count":96,"ordering":0,"published_at":"2026-08-10 09:17:15","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-10 09:17:15","updated_at":"2026-08-10 09:17:15","key_takeaways":null,"faq":null,"sources":null,"category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/0515ede3140f009e30ed51a369700a59.webp","featured_image_alt":"GaMS3-12B-Instruct chat interface with a prompt field and ARIS, NACP and NextGenerationEU logos.","featured_image_width":1600,"featured_image_height":824,"intro_image_path":"/api/uploads/2026/08/0515ede3140f009e30ed51a369700a59.webp","intro_image_width":1600,"intro_image_height":824,"seo_title":"Slovenia Launches GaMS AI Model for the Slovene Language","meta_description":"Slovenia has launched GaMS, an open-source AI language model built for Slovene, offering lessons for smaller languages across the Balkans.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":152,"title":"Hack for Humanity Comes to Skopje","slug":"hack-for-humanity-comes-to-skopje","short_description":null,"content":"Ten teams developed AI prototypes for wildfire response, parking, air pollution and other local challenges during Hack for Humanity: Skopje, held on October 3 as part of Startup Revolution AI Summit 2026.
\n![\"Hack](\"/api/uploads/2026/10/88ff12215c97cc2e0cdf727cf18f9491.webp\")
The event brought developers, students, founders, designers and other participants together for a short build session centred on problems affecting their own communities.
Hack for Humanity is a global initiative by The AI Collective that encourages local communities to identify challenges around them and use artificial intelligence and software to explore practical solutions. The wider initiative connects 130 AI Collective chapters across 50 countries.
\nThe Skopje edition was organised through AI Collective Skopje, with Monika Rizovska Tanevska involved in leading the local event. INOVA and Women in Tech Macedonia also supported the initiative.
\nThe idea behind the format is straightforward: people who understand a problem closely should have an opportunity to participate in building its solution.
\n![\"Atmosphere](\"/api/uploads/2026/10/26f1c16af9422d6ad8935adc86103d29.webp\")
![\"Working](\"/api/uploads/2026/10/819d61d003b04b575a6de28ca5f2c236.webp\")
![\"Monika](\"/api/uploads/2026/10/89e6633f2a6e32a0eb992e79ca73c645.webp\")
![\"5.jpeg\"](\"/api/uploads/2026/10/c1cb5d01a0a85485a78c08d6101f9812.webp\")
![\"4.jpeg\"](\"/api/uploads/2026/10/bb9514326473be440a886f3501afd205.webp\")
![\"3.jpeg\"](\"/api/uploads/2026/10/c6bff5f89081c439f4ad57de2c00e77b.webp\")
![\"2.jpeg\"](\"/api/uploads/2026/10/54db9144f7db593533398bf2c967bf42.webp\")
![\"1.jpeg\"](\"/api/uploads/2026/10/23dbb2ba879daaff32b75c47eb31b41e.webp\")
The programme started at 16:00 with a short briefing on the challenges.
\nParticipants then had around three hours to work in teams, develop their ideas and create prototypes before presenting them to the judging panel.
\nThe hackathon was designed to be open beyond experienced programmers. Developers, designers, entrepreneurs, students, researchers, professionals and participants with ideas for local problems could take part.
\nNo previous hackathon experience or finished project was required.
\nBy the end of the session, 10 teams had presented their work.
\nThe projects were evaluated by Prof. Dr. Vladimir Trajkovikj and Suad Seferi.
\n![\"Suad](\"/api/uploads/2026/10/e1f81e619d1cf4e0c06c50081576b242.webp\")
![\"Vladimir](\"/api/uploads/2026/10/3322886b43dc3dbd4645803d833bdd70.webp\")
![\"\"](\"/api/uploads/2026/10/46cc2590eb0d9c3f2001bb9b8e0637cc.webp\")
The winning team focused on one of North Macedonia's recurring environmental challenges: wildfires.
\nThe team developed an interactive mapping concept designed to locate fires across the country while also identifying nearby water supplies that could potentially support firefighting operations.
\nThe idea connected two important pieces of information in one interface: where a fire is located and where available water resources can be found nearby.
\nThe project took first place at the Skopje hackathon.
\nThe second-place team focused on a very different problem, but one familiar to drivers across the capital.
\nTheir project was designed to help users identify available parking spaces in Skopje and understand where parking is free or requires payment.
\nThe concept aimed to bring parking availability and payment information together, giving drivers a clearer picture of their options before reaching a destination.
\nTwo teams independently arrived at similar ideas around air pollution and outdoor activity.
\nTheir projects explored how environmental and air-quality information could help people understand current pollution conditions and decide when it may be more suitable to go outside, exercise or run.
\nRather than selecting one of the two similar concepts, the judges encouraged the teams to collaborate and combine their work.
\nThe two teams were jointly recognised in third place.
\nHack for Humanity also overlapped with DevDay Exchange Community: Skopje, a hands-on Codex workshop held during Startup Revolution AI Summit.
\n![\"1791042161592640.jpg\"](\"/api/uploads/2026/10/91b3f1e7c8bbcb813c089c1dba343712.webp\")
![\"Berat](\"/api/uploads/2026/10/7cba3fa3e39163ae5be3e63d7a80d17a.webp\")
The session took place from 16:30 to 17:30 at INOVA and was hosted by Berat Kjamili, an OpenAI Codex Ambassador, together with Ana Sekulovska and the AI Collective community.
\nIt brought together developers, students, founders and other builders for a live Codex demonstration, discussion around ideas and announcements from OpenAI's DevDay, and hands-on work with the coding tool.
\nMonika Rizovska Tanevska, organiser of Hack for Humanity Skopje, and Darko Janevski, an AI-focused entrepreneur and consultant, were listed among the supporters of the DevDay Exchange session.
\nHackathon teams were also provided with $150 in Codex credits per team, giving them access to AI-assisted development tools during the build.
\nFor teams working under a tight deadline, that meant Codex could be used to help move from an initial idea toward a demonstrable prototype within a few hours.
\nThe DevDay Exchange was a community-led event and not an official OpenAI event.
\nThe judging panel combined academic experience in computer science with work around the practical adoption of artificial intelligence.
\n![\"1000080113.jpg\"](\"/api/uploads/2026/10/4412e1ea9367413837fe63ffeed5cda6.webp\")
Prof. Dr. Vladimir Trajkovikj is a professor at the Faculty of Computer Science and Engineering at Ss. Cyril and Methodius University in Skopje. His academic and research work covers computer science, digital technologies and applied research, including participation in numerous international projects.
\nSuad Seferi is an AI strategist and educator, and founder and president of AINOW Society, a North Macedonia-based organisation working on AI literacy, education, research and responsible adoption. His work also covers AI strategy, organisational readiness, digital transformation and the practical implementation of AI.
\nThe teams were assessed after presenting what they had managed to develop during the limited hackathon period and explaining the problem their projects were designed to address.
\nHack for Humanity followed Humans in AI Week, an earlier AI Collective initiative in which communities discussed what it means to be human as artificial intelligence becomes more widely used.
\nThe hackathon shifted the focus from discussion to building.
\nThat was visible in Skopje. The leading projects did not begin with a general AI product and then search for a use case. They started with specific problems: wildfires, access to water during emergencies, parking and polluted air.
\nThe event also showed how quickly current AI development tools can change the format of a hackathon. With access to tools such as Codex, teams could spend more of their limited time testing an idea and assembling a prototype rather than working entirely from scratch.
\nThe harder stage begins after the event.
\nA prototype created in several hours can demonstrate an idea, but turning it into a useful service requires reliable local data, testing with real users and cooperation with the institutions or organisations responsible for the underlying problem.
\nFor the teams behind the wildfire, parking and environmental projects, Hack for Humanity Skopje provided the first version of that work.
","featured_image_id":352,"intro_image_id":352,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":10,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":95,"ordering":0,"published_at":"2026-10-03 20:41:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-03 20:41:08","updated_at":"2026-10-17 18:36:47","key_takeaways":"[\"Ten teams developed AI prototypes around local challenges during Hack for Humanity: Skopje.\",\"The winning project focused on mapping wildfires and nearby water supplies, while the second-place team worked on parking availability in Skopje.\",\"The hackathon ran alongside a DevDay Exchange Community session where participants worked hands-on with OpenAI Codex.\"]","faq":"[{\"q\":\"What is Hack for Humanity?\",\"a\":\"Hack for Humanity is a global initiative by The AI Collective that encourages local communities to identify problems around them and build AI-supported solutions.\"},{\"q\":\"How many teams participated in Skopje?\",\"a\":\"Ten teams developed and presented projects during the Skopje edition.\"},{\"q\":\"Who organised Hack for Humanity Skopje?\",\"a\":\"The event was organised through AI Collective Skopje, with Monika Rizovska Tanevska involved in leading the local edition and support from INOVA and Women in Tech Macedonia.\"},{\"q\":\"Who judged the projects?\",\"a\":\"The projects were judged by Prof. Dr. Vladimir Trajkovikj and Suad Seferi.\"},{\"q\":\"Was OpenAI involved in the Skopje event?\",\"a\":\"The Codex workshop was hosted by Berat Kjamili, who is officially listed by OpenAI as a Codex Ambassador. The session was part of OpenAI’s wider developer community programme, but the Skopje event itself was community-led rather than an official OpenAI event.\"}]","sources":null,"category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Guest Contributor","author_slug":"guest-contributor","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/19fd32d80806cd6e46c69cd2484e8488.webp","featured_image_alt":"Hack for Humanity","featured_image_width":1448,"featured_image_height":1086,"intro_image_path":"/api/uploads/2026/10/19fd32d80806cd6e46c69cd2484e8488.webp","intro_image_width":1448,"intro_image_height":1086,"seo_title":"Hack for Humanity Comes to Skopje","meta_description":"Ten teams built AI projects for wildfires, parking and pollution at Hack for Humanity Skopje, alongside a hands-on Codex workshop.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":160,"title":"Western Balkans Digital Summit 2026 to Focus on AI","slug":"western-balkans-digital-summit-2026-to-focus-on-ai","short_description":null,"content":"The Western Balkans Digital Summit will bring government officials, technology experts and business representatives to Herceg Novi, Montenegro, from October 12 to 14, with artificial intelligence, cybersecurity and regional digital integration among the central topics.
\nThe eighth edition will take place at One&Only Portonovi under Montenegro's 2026 presidency of the Berlin Process. More than 150 participants are expected, including representatives of the six Western Balkan economies, the European Commission, academic institutions, businesses and civil society organizations.
\n
The event is organized by Montenegro's Ministry of Public Administration in partnership with the Regional Cooperation Council (RCC), which has supported the summit since its introduction in 2018.
\nAccording to the Montenegrin government's announcement, this year's programme is intended to address practical challenges in developing a more connected, secure and competitive regional digital economy.
\nThe summit's programme is organized around five main areas: digital economy, digital skills, digital infrastructure, artificial intelligence and connectivity, and trust and security.
\nAI discussions will focus on the implementation of advanced technologies and their role in supporting regional connectivity. The wider programme will also examine how businesses, governments and educational institutions can develop the skills and infrastructure needed for digital transformation.
\nCybersecurity will be another central topic, with attention to strengthening the region's resilience against digital threats and improving cooperation between institutions.
\nThe Regional Cooperation Council has also identified digital public services and online child protection among the subjects expected to feature in the discussions.
\nFor the Western Balkans, these issues extend beyond technology adoption. Differences in institutional capacity, digital infrastructure and access to skilled professionals affect how quickly countries can introduce new digital services and benefit from AI applications.
\nIn an interview published by the Ministry of Public Administration on October 8, Montenegro's Minister of Public Administration, Marash Dukaj, outlined the country's expectations for the summit.
\nHe said regional cooperation should produce practical results in areas such as cross-border data exchange, digital identity, public services and cybersecurity.
\nOne priority is closer alignment with the European Union's Digital Single Market, including work toward more affordable mobile roaming between the Western Balkans and EU countries.
\nDukaj also pointed to the need for stronger cooperation on cybersecurity, arguing that increasingly interconnected public systems require coordinated responses to digital threats.
\nMontenegro has announced plans to expand its digital public services and develop an AI assistant for its e-government platform.
\nThe minister stressed that AI in public administration must remain subject to human oversight, particularly when systems are involved in legal or administrative decisions.
\nThese are national priorities presented by Montenegro ahead of the summit, rather than agreements already reached by participating governments.
\nThe Western Balkans Digital Summit has served as a meeting point for regional governments and technology stakeholders since 2018. The previous edition took place in Skopje in October 2025.
\nThis year's gathering comes as the region continues working to align its digital policies and infrastructure with European frameworks.
\nFor businesses, improved digital connectivity and interoperable public services could reduce administrative barriers when operating across borders.
\nFor governments, common approaches to cybersecurity, digital identity and AI governance could support more reliable public services and regional cooperation.
\nHowever, the extent to which these ambitions translate into practical results will depend on agreements, financing and implementation after the summit.
\nThe most relevant outcomes will be any concrete commitments involving AI governance, cybersecurity cooperation, digital public services and regional connectivity.
\nParticipants are also expected to discuss how to strengthen digital skills and prepare the workforce for changing technological requirements.
\nWith AI increasingly entering public administration and business operations, the challenge for Western Balkan economies is to move from policy discussions toward measurable improvements in services, institutional capacity and digital infrastructure.
\nThe October 12–14 summit provides another opportunity to establish shared priorities. Its longer-term significance will depend on what participating institutions agree to implement after leaving Herceg Novi.
","featured_image_id":364,"intro_image_id":364,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":13,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":87,"ordering":0,"published_at":"2026-10-09 08:34:38","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-09 08:34:38","updated_at":"2026-10-09 08:34:38","key_takeaways":"[\"The eighth Western Balkans Digital Summit takes place in Herceg Novi, Montenegro, from October 12 to 14, with over 150 participants expected.\",\"Artificial intelligence, cybersecurity, infrastructure, digital skills and economic integration form the programme's five central areas.\",\"Montenegro wants the summit to advance practical regional cooperation, including digital public services and closer integration with the EU digital market.\"]","faq":"[{\"q\":\"When and where is the Western Balkans Digital Summit 2026?\",\"a\":\"The summit is scheduled for October 12–14, 2026, at One&Only Portonovi in Herceg Novi, Montenegro.\"},{\"q\":\"What are the main topics?\",\"a\":\"AI, cybersecurity, digital economy, digital skills, infrastructure, connectivity and regional digital integration.\"},{\"q\":\"Who organizes the summit?\",\"a\":\"Montenegro's Ministry of Public Administration, under government patronage and in partnership with the Regional Cooperation Council.\"},{\"q\":\"Can participants register?\",\"a\":\"The government announced that registration is open through its official event registration link.\"}]","sources":"[{\"title\":\"Government of Montenegro, October 7 announcement\",\"url\":\"https://www.gov.me/clanak/otvorene-prijave-za-osmi-digitalni-samit-zapadnog-balkana-u-herceg-novom\"},{\"title\":\"Regional Cooperation Council, summit information\",\"url\":\"https://www.rcc.int/events/1670/rcc-co-organizes-8th-western-balkans-digital-summit\"},{\"title\":\"Montenegro Ministry of Public Administration, October 8 interview\",\"url\":\"https://www.gov.me/clanak/dukaj-crna-gora-zeli-da-kroz-samit-ubrza-povezivanje-regiona-sa-digitalnim-trzistem-eu\"},{\"title\":\"Western Balkans Digital Summit website\",\"url\":\"https://digitalsummitwb6.com/\"}]","category_name":"Montenegro","category_slug":"montenegro","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/cebed43670cecd766b9558cfff22f5b9.webp","featured_image_alt":"Montenegro","featured_image_width":1200,"featured_image_height":799,"intro_image_path":"/api/uploads/2026/10/cebed43670cecd766b9558cfff22f5b9.webp","intro_image_width":1200,"intro_image_height":799,"seo_title":"Western Balkans Digital Summit 2026 to Focus on AI","meta_description":"The Western Balkans Digital Summit opens in Montenegro on October 12, bringing AI, cybersecurity, digital skills and EU integration into focus.","robots":"index","canonical_url":"western-balkans-digital-summit-2026-ai","og_title":"","og_description":""},{"id":18,"title":"Who controls the machines that will control the world?","slug":"who-controls-the-machines-that-will-control-the-world","short_description":"Artificial General Intelligence (AGI) has become one of those terms that seems to mean everything and nothing at the same time. In Silicon Valley, it is…","content":"Artificial General Intelligence (AGI) has become one of those terms that seems to mean everything and nothing at the same time. In Silicon Valley, it is marketed as the inevitable next step in technological evolution. In policy circles, it is framed as a looming national security issue. For ethicists, it raises difficult questions about consciousness, rights, and responsibility.
\nBut the more immediate question is much simpler: who gets to control it?
\nThat question runs through the reflections of Ben Goertzel, the AI researcher who popularized AGI decades before it entered mainstream discourse and now leads SingularityNET and the Artificial Superintelligence Alliance. His argument is not about a single technological milestone. It is that AGI could reshape the architecture of power itself.
\nThe debate is often reduced to whether machines will become conscious or smarter than humans. Goertzel’s concern is that, long before that point, societies may already have handed over critical layers of knowledge production and governance to a small group of actors.
\nHistory offers familiar warnings. The internet was expected to decentralize information, social media to flatten hierarchies. Instead, both consolidated power inside a handful of corporations. Now, there is no guarantee AI will behave differently.
\nYet not everyone agrees that technology should be the center of the debate. Dr. Seth Dobrin, a geneticist-turned-AI strategist, argues that public discourse begins with a category error: attributing agency to systems that merely generate fluent language.
\n“We’re applying human moral vocabulary to mathematics. A model is software: weights, matrices, and optimization targets,” he argues in an interview via email for AGI Ethics News. In his view, AI should be treated like any engineered system: verified, predictable, and governed through accountability, not anthropomorphism.
\nPower and the illusion of inevitability
\nThat disagreement exposes a deeper divide. Goertzel explores whether future systems might exhibit some form of subjective experience. Dobrin insists the real issue is not machine being, but human responsibility.
\n“You test a tool. You empathize with a being. Confuse the two, and you end up doing neither well”
\nDespite their differences, both converge on a core point: the danger is not machine autonomy, but human concentration of control.
\nGoertzel describes a “sovereign tail” risk in which advanced AI becomes a strategic asset controlled by a few states or hyperscale corporations, forming the infrastructure of economic and informational life.
\n“The deeper danger then inverts: not concentration, but uncoordinated proliferation of forkable systems. The answer is the same: build intelligence as a decentralized network whose power lives in the collective whole.” Goerzel noted in an interview via email for AGI Ethics News.
\nDobrin is more direct. “AGI doesn’t hold power. Companies and states hold power. The technology is a lever; the hands on the lever are human,” he says. The real question is whether a small number of corporations will end up governing global information infrastructure.
\nBoth reject AGI as an autonomous historical force. Instead, it becomes an amplifier of existing power structures shaped by design, ownership, and deployment choices.
\nDecentralization, knowledge, and control
\nGoertzel challenges the assumption that AGI must be built on centralized compute. He instead envisions intelligence emerging across distributed networks rather than being locked inside corporate infrastructure.
\n“The geopolitically safest AGI is one nobody can own,” he argues, because intelligence should be “an event the network performs, not an object at rest anyone can capture.”
\nDobrin reaches a similar conclusion through governance logic. He advocates smaller, specialized models operating on sovereign data and embedded in local systems. The geopolitics of AI, he says, “will be decided by procurement choices, not prophecy.”
\nAt stake is a broader question: whether intelligence can be owned at all, or whether it behaves more like language or culture… something that exists only through distribution. As AI systems increasingly mediate knowledge itself, control over their architecture becomes control over perception. The machine is no longer just carrying knowledge, but it is also shaping it.
\nThe philosophical stakes, though, deepen with consciousness.
\nGoertzel, influenced by Alfred North Whitehead, leans toward pan-experientialism — the idea that experience may be a fundamental feature of organized matter. The question is not whether AI is conscious, but what kinds of internal states its architecture enables.
\nDobrin rejects this framing. “You test a tool. You empathize with a being. Confuse the two, and you end up doing neither well,” he says. AI systems, regardless of complexity, cannot hold rights because they cannot hold responsibility. Only humans and institutions can.
\nDespite this disagreement, both converge on a key concern: opacity.
\nGoertzel argues that ethical debate requires visibility into how systems operate. Black-box models encourage projection, while interpretable systems allow partial understanding of internal reasoning.
\nDobrin echoes this in engineering terms: “You’re entitled to systems that show their work mathematically. Proof, not prediction.”
\nAs systems scale, however, they often become less interpretable, intensifying the governance challenge.
\nA future shaped by choices, not inevitability
\nAI is no longer just distributing information; it is mediating and synthesizing it before humans encounter it. It is actively shaping knowledge itself.
\nGoertzel proposes governance through decentralized, constitutional-style AI ecosystems where transparency and accountability are embedded at the architectural level.
\nDobrin reduces governance to a simpler principle: responsibility never transfers to the machine. “A system can mediate a decision, but a human or institution must own it,” he says. Once “the algorithm decided” becomes acceptable, accountability collapses.
\nIn a statement for AGI Ethics News, innovation strategist Dan Herman points out that over 90 percent of AI startup funding is concentrated in the US and China. For smaller economies, the challenge is not dominance but coordination – building ecosystems capable of collective participation in AI infrastructure.
\nSeen together, Goertzel and Dobrin offer complementary diagnoses. One focuses on the architecture of intelligence, the other on institutional accountability, and both reject technological inevitability.
\nThe central question of AGI may never be whether machines become human-like. It is whether humans allow the systems they build to reproduce the oldest pattern in political history: the concentration of power in fewer and fewer hands.
\nEthical concerns addressed in this article:
\nNorth Macedonia-based AINOW Society has released version 0.95 of its free AI Literacy Platform, introducing a new development-focused learning environment alongside improvements to its existing Education and Business applications.
\nThe update expands the platform beyond AI fundamentals and professional use into the technical aspects of building AI-powered applications.
\n
The new environment, called AINOW DEV, introduces learning materials covering AI skills, Model Context Protocol (MCP) connectors and practical workflows. It is designed for students, developers and professionals who want to understand how AI systems interact with external tools and services.
\nAINOW Society, which also publishes AI Balkans, developed the platform as an open-source educational resource available in Macedonian, Albanian and English.
\nThe AINOW AI Literacy Platform originally focused on helping teachers and educational institutions understand artificial intelligence and use it responsibly.
\nIts Education environment currently includes 19 learning chapters, 1,106 educational prompts, 57 tools and a glossary of 60 AI-related terms. Teachers can use these resources to prepare lessons, develop classroom activities and create assessments.
\n
The platform expanded into professional applications with version 0.94, which introduced AINOW Business.
\nThat environment now provides 3,270 prompts across 33 professional fields, 2,190 AI agent definitions and 453 practical scenarios. Its resources cover areas including management, administration, finance, communication and software development.
\n
With version 0.95, users can access a third environment focused on understanding and developing AI applications.
\nThe new DEV environment includes a 13-chapter learning guide, seven AI tool profiles and three reference libraries containing:
\nMCP, or Model Context Protocol, is a technical standard that allows AI applications to communicate with external tools and data sources.
\n
As AI assistants become capable of performing tasks beyond generating text, understanding these connections is increasingly relevant to developers and organisations exploring automation.
\nDEV also provides practical exercises, tools for examining original AI skill files and a calculator for estimating token and context requirements.
\nThe reference libraries are intended for education and exploration. They do not represent automatically installed connectors or immediately executable integrations.
\nVersion 0.95 also introduces a unified interface across Education, Business and DEV.
\nThe update includes improved navigation, more consistent language controls, updated prompt libraries and better reading options on mobile devices.
\nUsers can now work with longer prompts through improved reading views, while mobile users benefit from a full-screen interface that makes reviewing and editing content easier.
\nThe release also improves local note-taking, settings and backup management. Business and DEV now provide separate local workspaces for saving notes and preferences.
\nTechnical changes address problems involving multilingual search, browser navigation and data synchronisation. The project's published release notes also report fixes for three previously identified GitHub CodeQL findings.
\nThe downloadable version passed automated verification checks, although additional physical-device and screen-reader testing was still pending at publication.
\nOne of the platform's distinguishing features is its approach to accessibility.
\nAll three learning environments are free to use and do not require an AINOW account. Users can access the online version or download the application for local use without installing a separate server.
\nDownloaded educational materials can also be used offline, although external AI services may still require internet access and separate accounts.
\nFor users in North Macedonia and the wider Western Balkans, the availability of learning materials in Macedonian and Albanian provides an alternative to relying entirely on English-language technical resources.
\nThe platform also offers practical material for schools, smaller organisations and individuals who may not have the budget for commercial training platforms.
\nIts open-source model allows developers and institutions to inspect the application code, adapt it to their needs and contribute improvements, subject to the project's licensing requirements.
\nThe new DEV environment extends AINOW's original educational focus toward technical AI literacy, giving learners a way to explore how AI-powered applications are assembled, connected and maintained.
\nThe platform is available at edu.ainow.mk. Further information about the update is available in the official AINOW Society announcement, while the downloadable application and technical release notes can be found on GitHub.
","featured_image_id":425,"intro_image_id":425,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":7,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":4,"ordering":0,"published_at":"2026-10-26 22:51:57","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-26 22:51:57","updated_at":"2026-10-26 22:51:57","key_takeaways":"[\"AINOW Society has released version 0.95 of its free AI Literacy Platform, introducing DEV as a third learning environment alongside Education and Business.\",\"AINOW DEV includes 13 learning chapters, 1,000 AI skills, 1,000 MCP references and 200 workflows designed to help users understand AI development and integrations.\",\"The platform is open source, supports Macedonian, Albanian and English, and offers offline access to downloaded learning resources without mandatory registration.\"]","faq":"[{\"q\":\"What is new in AINOW AI Literacy v0.95?\",\"a\":\"The update introduces AINOW DEV, a learning environment focused on AI development, MCP connectors, reusable skills and workflows. It also improves navigation, prompt reading, mobile usability and security across the platform.\"},{\"q\":\"Who can use the AINOW AI Literacy Platform?\",\"a\":\"The platform serves educators, students, professionals, organisations and developers through three dedicated environments: Education, Business and DEV.\"},{\"q\":\"Is AINOW AI Literacy free and available offline?\",\"a\":\"Yes. The platform is free, open source and accessible without registration. Downloaded learning materials can be used offline, although external AI services may require internet access.\"}]","sources":"[{\"title\":\"AINOW AI Literacy Platform v0.95 Introduces DEV\",\"url\":\"https://ainow.mk/blog/ainow-ai-literacy-platform-v0-95-introduces-dev\"}]","category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Aleksandra Todorovikj","author_slug":"aleksandra-todorovikj","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/ffc8dd720f66d531239cbc548799965c.webp","featured_image_alt":"Ainow literacy","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/ffc8dd720f66d531239cbc548799965c.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"AINOW v0.95 Adds DEV to Free AI Learning Platform","meta_description":"AINOW v0.95 launches DEV with 1,000 AI skills, 1,000 MCP references and 200 workflows. Free AI education in three languages.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":156,"title":"VEZILKA Deepens AI Factory Cooperation With Greece","slug":"vezilka-deepens-ai-factory-cooperation-with-greece","short_description":null,"content":"North Macedonia’s VEZILKA AI Factory is expanding cooperation with Greece’s PHAROS AI Factory, with the two initiatives discussing how expertise, computing resources and services could be shared more effectively across the European AI Factory network.
\nThe talks took place during a visit to Athens by Gjorgji Madjarov of the Faculty of Computer Science and Engineering at Ss. Cyril and Methodius University in Skopje. Madjarov met Stefanos Kollias, chairman of GRNET and coordinator of PHAROS, as well as GRNET project manager Ilias Hatzakis.
\nAccording to VEZILKA, the discussions covered cooperation in areas of common interest, exchange of technical knowledge and approaches to federated access to computing resources.
\nFor North Macedonia, that last point is particularly important.
\nVEZILKA operates as the country’s AI Factory Antenna, connecting local researchers, companies and institutions with infrastructure available through Europe’s wider network of AI Factories. Unlike a full AI Factory with its own large-scale supercomputing infrastructure, the antenna model is designed to provide national access to expertise, services and computing capacity available elsewhere in the network.
\nPHAROS is one of those connections.
\nThe Greek AI Factory is linked to the DAEDALUS supercomputer and focuses on AI applications in areas including health, language and culture, and sustainability.
\nVEZILKA already identifies PHAROS as a strategic partner and says the cooperation gives organisations in North Macedonia a pathway towards European AI computing infrastructure.
\nThe latest meeting focused not only on infrastructure, but also on how organisations can move from an initial AI idea towards testing and implementation.
\nThat includes questions around how computing resources are requested, how technical support is delivered and how AI Factory services can be made easier to use for researchers, businesses and public institutions.
\nThese issues are becoming increasingly relevant as the European Union expands its AI Factory network around high-performance computing infrastructure.
\nFor smaller countries, building independent large-scale AI computing capacity is expensive. Participation in a shared European network offers another model: local organisations can receive support through national structures while accessing specialised infrastructure elsewhere.
\nVEZILKA says experience from PHAROS could help shape its own services and support mechanisms as the North Macedonian initiative develops.
\nThe project officially started in April 2026 and runs for 36 months. Its total budget is approximately €6.2 million, including around €3.1 million in EU funding through the EuroHPC Joint Undertaking, with additional support from North Macedonia’s Ministry of Digital Transformation and FCSE at UKIM.
\nVEZILKA focuses on five main areas: language and culture, public administration, health, energy and agriculture.
\nThe cooperation also illustrates how AI infrastructure in Southeast Europe is beginning to develop across national borders rather than as isolated projects.
\nNorth Macedonia does not need to reproduce every component of a large AI Factory locally if researchers and companies can gain reliable access to computing resources and expertise through partners such as PHAROS.
\nThe practical test will be whether that model becomes easy enough for local organisations to use.
\nAccess to a European supercomputer matters only if companies, researchers and public institutions can move from application procedures to actual experiments, models and services.
\nVEZILKA and PHAROS are now discussing how that path can work more effectively, including through federated resource provisioning and closer exchange of expertise.
\nAs VEZILKA moves further into its implementation phase, the next measure of progress will be how many organisations in North Macedonia move beyond training and awareness and begin using the infrastructure and technical services available through the European AI Factory ecosystem.
","featured_image_id":360,"intro_image_id":360,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":3,"view_count":27,"ordering":0,"published_at":"2026-10-07 08:54:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-07 08:54:16","updated_at":"2026-10-07 08:55:14","key_takeaways":"[\"VEZILKA and Greece’s PHAROS AI Factory are expanding cooperation around expertise, computing resources and AI services.\",\"Discussions include federated access to resources available through the European AI Factory network.\",\"VEZILKA acts as North Macedonia’s national AI Factory Antenna rather than operating a standalone large-scale AI supercomputer.\",\"Its partnership with PHAROS provides a route towards European high-performance computing infrastructure, including the DAEDALUS supercomputer.\",\"The next challenge is turning infrastructure access into practical AI projects for companies, researchers and public institutions.\"]","faq":"[{\"q\":\"What is VEZILKA?\",\"a\":\"VEZILKA is North Macedonia’s AI Factory Antenna, established within the European AI Factories ecosystem to support companies, researchers, institutions and other organisations working with artificial intelligence.\"},{\"q\":\"What is PHAROS?\",\"a\":\"PHAROS is Greece’s AI Factory and part of the European network of AI Factories supported through EuroHPC.\"},{\"q\":\"Does North Macedonia have its own AI supercomputer?\",\"a\":\"VEZILKA operates as an AI Factory Antenna. Its model connects users in North Macedonia with computing resources, expertise and services available through partner AI Factories rather than replicating all large-scale infrastructure locally.\"},{\"q\":\"What is DAEDALUS?\",\"a\":\"DAEDALUS is Greece’s high-performance computing system and forms part of the infrastructure supporting PHAROS AI Factory.\"},{\"q\":\"Why does the VEZILKA-PHAROS partnership matter?\",\"a\":\"The partnership could make advanced computing infrastructure and AI expertise more accessible to researchers, businesses and public institutions in North Macedonia.\"}]","sources":"[{\"title\":\"VEZILKA Strengthens Cooperation with Greek AI Factory – PHAROS\",\"url\":\"https://vezilka.ai/vezilka-strengthens-cooperation-with-pharos-ai-factory/\"}]","category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/932fa0e513dea4f2895e1d61d4c5fa76.webp","featured_image_alt":"Vezilka Greece Meeting","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/932fa0e513dea4f2895e1d61d4c5fa76.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"VEZILKA Expands AI Factory Cooperation With Greece","meta_description":"VEZILKA and Greece’s PHAROS are expanding cooperation on AI expertise, computing resources and access to European AI infrastructure.","robots":"index","canonical_url":"vezilka-deepens-ai-factory-cooperation-with-greece","og_title":"","og_description":""},{"id":154,"title":"Meet Lumi: Inside AI Balkans 2.0","slug":"meet-lumi","short_description":null,"content":"AI Balkans has undergone its biggest platform update since launch, adding its own AI assistant, personalised reading, smarter search and a growing set of reference tools built around the publication's reporting.
\nAt the centre of the update is Lumi, the new AI Balkans assistant.
\nRather than starting with the open web, Lumi looks first at what AI Balkans already knows: more than 130 published articles, country and topic archives, the AI glossary, reviewed tools and hundreds of curated answers.
\nThat distinction matters for a news platform.
\nAsk Lumi about Montenegro, for example, and it does not generate a generic summary of AI in Montenegro. It searches the stories AI Balkans has actually published, prioritises articles filed under the country, tells the reader how much coverage is available and can also surface stories where Montenegro is mentioned.
\nFor news questions, Lumi works directly from published AI Balkans stories. When the site's own material does not contain an answer, an external AI system can be used as a fallback.
\nEvery answer also links back to the stories behind it.
\n![\"\"](\"/api/uploads/2026/10/43b0d1e3b3915cdfca3b4b9f8750cef3.webp\")
The previous \"Ask AI\" feature has now become Ask Lumi across AI Balkans.
\nLumi appears in the top navigation, search, article tools, case studies, the glossary and the AI tools catalogue. On mobile, the assistant is available through a floating button, while a full workspace is available at aibalkans.com/ai.
\nThe character also has its own small visual states while searching, thinking and writing.
\nIts name comes from lumen, the Latin word for light.
\nAsk \"Who are you?\" and Lumi answers:
\n\"I'm Lumi, the AI Balkans assistant. My name comes from lumen, the Latin word for light, because I help shed light on AI news, policy, tools and people across the Balkans.\"
\nThe assistant has also been tuned around how AI Balkans itself is organised. Country questions search news coverage rather than unrelated site content, stories filed directly under that country are prioritised, and follow-up questions maintain the original country context.
\nCountry and topic queries currently return in roughly a third of a second.
\nReaders who want to continue exploring elsewhere can send an answer directly to Google AI, ChatGPT or Perplexity with the original question and Lumi's answer already included.
\nLumi is also integrated directly into individual stories.
\nEvery article includes a menu for questions such as:
\nSummarize, Related articles, More from this author, Explain the key terms, What are the sources?, Who wrote this? and When was it published?
\nReaders can also highlight a sentence or phrase inside an article and select Ask Lumi to ask about that specific part of the story.
\nRelated-story and author searches can be refreshed through \"Different results\", instead of repeatedly showing the same recommendations.
\nThe idea is simple: reading a story should not have to end at the last paragraph.
\nAI Balkans is also introducing personalisation, but without creating a central profile of what each reader does.
\nReading preferences are calculated on the reader's own device.
\nPeople can follow subjects such as Policy and Regulation, AI Education or Cybersecurity, save stories, keep their reading progress and return to unfinished articles.
\nA For you section can recommend three stories based on previous reading or followed topics.
\nReturning visitors may also see Continue reading and Welcome back, which highlights material published since their previous visit.
\nFirst-time visitors see none of these features until there is enough activity to make them useful.
\nThe last eight Lumi conversations are similarly stored on the device rather than as an AI Balkans account history.
\nFinding an article has changed as well.
\nSearch now provides suggestions while typing, spelling corrections through \"did you mean\", and a short answer above conventional search results when the site has enough information to answer the query directly.
\nTopic pages have developed into small coverage hubs. They include an explanation of the subject, related topics and a chronological view of AI Balkans reporting grouped by month.
\nIndividual stories can contain a The story so far timeline, while Mentioned in this story connects readers with organisations, people and tools appearing in the article.
\nGlossary terms can be explained without leaving the page.
\nArticles also show estimated reading time remaining, reading progress, update information and short summaries, while photo galleries preserve the proportions of the original images.
\nTraditional numbered pagination has largely been replaced by Load more.
\nAnother addition is the Weekly Balkan AI Briefing.
\nAvailable at aibalkans.com/briefing, with its own RSS feed, the page collects the week's coverage and shows which countries and sections received the most attention, alongside the most-read story.
\nIt provides a different way of following the region for readers who do not visit every day.
\nThe larger change is that AI Balkans is gradually connecting reporting with structured reference material.
\nThe EU AI Act Tracker follows six Western Balkan countries and includes a timeline, upcoming deadline information and a visual overview of progress.
\nCountry pages now include a State of AI briefing, refreshed daily, together with facts drawn from the tracker.
\nAlongside the reporting sit an AI glossary, a catalogue of 25 reviewed AI tools, comparisons such as ChatGPT versus Claude, a tool finder and a regional AI directory.
\nAI Balkans has also started documenting practical implementations through case studies. Two are currently published: the AINOW Society AI literacy platform and Lumi itself.
\nOn desktop, the glossary, tools, case studies, directory and AI Act tracker can be opened from the Lumi workspace without leaving it.
\nAI Balkans started as a publication covering artificial intelligence across the Balkans.
\nThat remains its core job.
\nBut after more than 130 published stories, simply placing articles one after another was no longer enough. The reporting had started to become a body of regional knowledge: countries, organisations, policies, people, tools and developments that were connected to one another but difficult to explore through a conventional news archive.
\nThe new AI Balkans is an attempt to make those connections usable.
\nA reader can still arrive, read today's story and leave.
\nOr they can ask what has happened in a country, follow a subject, understand an unfamiliar term, trace earlier developments, compare tools, check the AI Act tracker or ask Lumi where the reporting leads next.
\nThe news remains the source.
\nThe difference is that readers can now do more with it.
","featured_image_id":357,"intro_image_id":357,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":6,"view_count":50,"ordering":0,"published_at":"2026-10-05 21:53:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-05 21:53:44","updated_at":"2026-10-05 22:32:15","key_takeaways":"[\"Lumi now connects directly to AI Balkans reporting, searching articles, topics, countries, glossary entries, tools and curated answers before relying on external AI.\",\"AI Balkans has added private on-device personalisation, including followed topics, saved stories, reading progress, recommendations and recent Lumi conversations.\",\"News is being connected with reference material, including the EU AI Act Tracker, country AI briefs, glossary, reviewed tools, directory, case studies and weekly briefing.\"]","faq":"[{\"q\":\"What is Lumi?\",\"a\":\"Lumi is the AI Balkans assistant. It answers questions using AI Balkans articles and structured site content first, with external AI available when the site does not contain an answer.\"},{\"q\":\"Does AI Balkans send my reading history to its servers for recommendations?\",\"a\":\"The new personalisation features are calculated on the reader's device. This includes followed topics, saved stories, reading progress and personalised story recommendations.\"},{\"q\":\"Can Lumi answer questions about individual AI Balkans articles?\",\"a\":\"Yes. Readers can request summaries, related articles, author information, key-term explanations, sources and publication information. Text inside an article can also be selected and sent directly to Lumi.\"},{\"q\":\"What else is available besides news?\",\"a\":\"AI Balkans now includes an EU AI Act Tracker covering six Western Balkan countries, country AI briefs, an AI glossary, 25 reviewed tools, a regional directory, case studies and the Weekly Balkan AI Briefing.\"}]","sources":"[{\"title\":\"Ask Lumi\",\"url\":\"https://aibalkans.com/ai/\"},{\"title\":\"Lumi Case Study\",\"url\":\"https://aibalkans.com/case-studies/lumi-ai-assistant\"},{\"title\":\"Weekly Balkan AI briefing\",\"url\":\"https://aibalkans.com/briefing\"}]","category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/79f2c9823a045398859dc3d4f848448d.webp","featured_image_alt":"AskLumi","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/79f2c9823a045398859dc3d4f848448d.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Meet Lumi: Inside AI Balkans 2.0 | AI Balkans","meta_description":"AI Balkans has undergone its biggest platform update since launch, adding its own AI assistant, personalised reading, smarter search and a growing set…","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":152,"title":"Hack for Humanity Comes to Skopje","slug":"hack-for-humanity-comes-to-skopje","short_description":null,"content":"Ten teams developed AI prototypes for wildfire response, parking, air pollution and other local challenges during Hack for Humanity: Skopje, held on October 3 as part of Startup Revolution AI Summit 2026.
\nThe event brought developers, students, founders, designers and other participants together for a short build session centred on problems affecting their own communities.
Hack for Humanity is a global initiative by The AI Collective that encourages local communities to identify challenges around them and use artificial intelligence and software to explore practical solutions. The wider initiative connects 130 AI Collective chapters across 50 countries.
\nThe Skopje edition was organised through AI Collective Skopje, with Monika Rizovska Tanevska involved in leading the local event. INOVA and Women in Tech Macedonia also supported the initiative.
\nThe idea behind the format is straightforward: people who understand a problem closely should have an opportunity to participate in building its solution.
\nThe programme started at 16:00 with a short briefing on the challenges.
\nParticipants then had around three hours to work in teams, develop their ideas and create prototypes before presenting them to the judging panel.
\nThe hackathon was designed to be open beyond experienced programmers. Developers, designers, entrepreneurs, students, researchers, professionals and participants with ideas for local problems could take part.
\nNo previous hackathon experience or finished project was required.
\nBy the end of the session, 10 teams had presented their work.
\nThe projects were evaluated by Prof. Dr. Vladimir Trajkovikj and Suad Seferi.
\n
The winning team focused on one of North Macedonia's recurring environmental challenges: wildfires.
\nThe team developed an interactive mapping concept designed to locate fires across the country while also identifying nearby water supplies that could potentially support firefighting operations.
\nThe idea connected two important pieces of information in one interface: where a fire is located and where available water resources can be found nearby.
\nThe project took first place at the Skopje hackathon.
\nThe second-place team focused on a very different problem, but one familiar to drivers across the capital.
\nTheir project was designed to help users identify available parking spaces in Skopje and understand where parking is free or requires payment.
\nThe concept aimed to bring parking availability and payment information together, giving drivers a clearer picture of their options before reaching a destination.
\nTwo teams independently arrived at similar ideas around air pollution and outdoor activity.
\nTheir projects explored how environmental and air-quality information could help people understand current pollution conditions and decide when it may be more suitable to go outside, exercise or run.
\nRather than selecting one of the two similar concepts, the judges encouraged the teams to collaborate and combine their work.
\nThe two teams were jointly recognised in third place.
\nHack for Humanity also overlapped with DevDay Exchange Community: Skopje, a hands-on Codex workshop held during Startup Revolution AI Summit.
\nThe session took place from 16:30 to 17:30 at INOVA and was hosted by Berat Kjamili, an OpenAI Codex Ambassador, together with Ana Sekulovska and the AI Collective community.
\nIt brought together developers, students, founders and other builders for a live Codex demonstration, discussion around ideas and announcements from OpenAI's DevDay, and hands-on work with the coding tool.
\nMonika Rizovska Tanevska, organiser of Hack for Humanity Skopje, and Darko Janevski, an AI-focused entrepreneur and consultant, were listed among the supporters of the DevDay Exchange session.
\nHackathon teams were also provided with $150 in Codex credits per team, giving them access to AI-assisted development tools during the build.
\nFor teams working under a tight deadline, that meant Codex could be used to help move from an initial idea toward a demonstrable prototype within a few hours.
\nThe DevDay Exchange was a community-led event and not an official OpenAI event.
\nThe judging panel combined academic experience in computer science with work around the practical adoption of artificial intelligence.
\n
Prof. Dr. Vladimir Trajkovikj is a professor at the Faculty of Computer Science and Engineering at Ss. Cyril and Methodius University in Skopje. His academic and research work covers computer science, digital technologies and applied research, including participation in numerous international projects.
\nSuad Seferi is an AI strategist and educator, and founder and president of AINOW Society, a North Macedonia-based organisation working on AI literacy, education, research and responsible adoption. His work also covers AI strategy, organisational readiness, digital transformation and the practical implementation of AI.
\nThe teams were assessed after presenting what they had managed to develop during the limited hackathon period and explaining the problem their projects were designed to address.
\nHack for Humanity followed Humans in AI Week, an earlier AI Collective initiative in which communities discussed what it means to be human as artificial intelligence becomes more widely used.
\nThe hackathon shifted the focus from discussion to building.
\nThat was visible in Skopje. The leading projects did not begin with a general AI product and then search for a use case. They started with specific problems: wildfires, access to water during emergencies, parking and polluted air.
\nThe event also showed how quickly current AI development tools can change the format of a hackathon. With access to tools such as Codex, teams could spend more of their limited time testing an idea and assembling a prototype rather than working entirely from scratch.
\nThe harder stage begins after the event.
\nA prototype created in several hours can demonstrate an idea, but turning it into a useful service requires reliable local data, testing with real users and cooperation with the institutions or organisations responsible for the underlying problem.
\nFor the teams behind the wildfire, parking and environmental projects, Hack for Humanity Skopje provided the first version of that work.
","featured_image_id":352,"intro_image_id":352,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":10,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":95,"ordering":0,"published_at":"2026-10-03 20:41:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-03 20:41:08","updated_at":"2026-10-17 18:36:47","key_takeaways":"[\"Ten teams developed AI prototypes around local challenges during Hack for Humanity: Skopje.\",\"The winning project focused on mapping wildfires and nearby water supplies, while the second-place team worked on parking availability in Skopje.\",\"The hackathon ran alongside a DevDay Exchange Community session where participants worked hands-on with OpenAI Codex.\"]","faq":"[{\"q\":\"What is Hack for Humanity?\",\"a\":\"Hack for Humanity is a global initiative by The AI Collective that encourages local communities to identify problems around them and build AI-supported solutions.\"},{\"q\":\"How many teams participated in Skopje?\",\"a\":\"Ten teams developed and presented projects during the Skopje edition.\"},{\"q\":\"Who organised Hack for Humanity Skopje?\",\"a\":\"The event was organised through AI Collective Skopje, with Monika Rizovska Tanevska involved in leading the local edition and support from INOVA and Women in Tech Macedonia.\"},{\"q\":\"Who judged the projects?\",\"a\":\"The projects were judged by Prof. Dr. Vladimir Trajkovikj and Suad Seferi.\"},{\"q\":\"Was OpenAI involved in the Skopje event?\",\"a\":\"The Codex workshop was hosted by Berat Kjamili, who is officially listed by OpenAI as a Codex Ambassador. The session was part of OpenAI’s wider developer community programme, but the Skopje event itself was community-led rather than an official OpenAI event.\"}]","sources":null,"category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Guest Contributor","author_slug":"guest-contributor","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/19fd32d80806cd6e46c69cd2484e8488.webp","featured_image_alt":"Hack for Humanity","featured_image_width":1448,"featured_image_height":1086,"intro_image_path":"/api/uploads/2026/10/19fd32d80806cd6e46c69cd2484e8488.webp","intro_image_width":1448,"intro_image_height":1086,"seo_title":"Hack for Humanity Comes to Skopje","meta_description":"Ten teams built AI projects for wildfires, parking and pollution at Hack for Humanity Skopje, alongside a hands-on Codex workshop.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":151,"title":"Startup Revolution Begins in Skopje: What to Expect","slug":"startup-revolution-begins-in-skopje-what-to-expect","short_description":null,"content":"Startup Revolution AI Summit begins in Skopje on October 1, bringing four days of artificial intelligence, startup building, investment, policy discussions, mentoring, pitching and networking across several locations in the city.
\nThe programme opens on Thursday evening before moving into two full conference days at the Macedonian Philharmonic on October 2 and 3. The summit closes on Sunday alongside the Wizz Air Skopje Marathon.
\nMore than 100 speakers and hundreds of startups are expected to participate, according to the organisers. Rather than following a single conference track, Startup Revolution is structured around different audiences, from first-time founders and scaleups to investors, policymakers, corporate leaders and students.
\n
October 1 functions primarily as the summit’s arrival and networking day.
\nPremium-pass participants have an afternoon side event outside Skopje, including a visit and wine tasting at Stobi.
\nThe wider summit programme begins in the evening at Srekja, home of Startup Club Skopje, with the official welcome reception and opening party. This is the first networking opportunity before the conference moves to the Macedonian Philharmonic the following morning.
\nFriday, October 2, is the first full conference day.
\nRegistration, coffee and networking begin at 09:00, followed by the official opening ceremony at 10:00.
\nThe programme then splits across several themes, including AI and emerging technologies, ecosystem and data, energy and climate tech, fintech, healthtech, the founder journey and early-stage startup development.
\nFor people building their first company, Friday has a particularly practical focus.
\nSessions cover how founders find their first customers, build sales processes, use mentors effectively, position an early-stage company and avoid common legal mistakes. Other parts of the programme deal with bootstrapping, going from an idea to an operating startup and preparing a startup pitch.
\nA practical masterclass will walk founders through the process of getting their first customer, while another session examines legal mistakes startup founders should avoid.
\nThere will also be a discussion with early-stage founders about bootstrapping and the decisions companies face before they have significant external funding.
\nStartup Revolution is not limited to stage presentations.
\nFriday includes hands-on sessions and workshops, including a founder resilience workshop focused on dealing with rejection and continuing towards a first customer or investment.
\nOne-to-one mentoring sessions are also scheduled with mentors connected to the Swiss Entrepreneurship Program and other summit partners.
\nLater in the day, founders will work directly on pitching, including a practical session on how startup pitches are structured and delivered.
\nInvestment and informal networking continue after the main sessions.
\nAn Investor Kafana is scheduled in the evening for Premium participants, followed by additional side events, including a programme centred on discussing startup failure and a late-night Game Night.
\nThe format reflects a wider part of Startup Revolution’s structure: meetings between founders and investors are spread throughout the summit instead of being confined to one investment panel.
\nSaturday, October 3, shifts the emphasis.
\nThe main tracks include government and policy, AI transformation and scaleups, trust and cybersecurity, fintech, marketing and the creator economy, business and investment.
\nThe morning begins with a policy-focused programme involving representatives from North Macedonia and the wider region.
\nSessions include discussions about building competitive economies in the age of AI, AI policy, public support for startups and the role of innovation institutions.
\nA regional panel will bring together representatives connected to innovation organisations in North Macedonia, Serbia and Montenegro to discuss policies, innovation funding and public-sector support for startup growth.
\nThere will also be a session examining cultural diplomacy and the startup ecosystem, with participants from organisations including the Serbian Chamber of Commerce and the Macedonian E-commerce Association.
\nOne of Saturday morning’s specific announcements is the special launch of the Skopje Innovation District, scheduled for 10:45.
\nThe programme lists professors Vladimir Trajkovikj and Radmil Polenakovikj for the launch.
\nThe initiative adds a local ecosystem component to a programme otherwise heavily focused on regional and international startup development.
\nTechnology inside government also has its own place in the programme.
\nNorth Macedonia’s Minister of Digital Transformation Stefan Andonovski is scheduled for a session on digital identity, digital trust and the next generation of public services.
\nGovernment-supported startups will also be presented during the policy track.
\nAnother session turns towards sovereign AI for businesses, adding questions around control of AI systems and organisational data to the programme.
\nOne of Saturday’s strongest AI-focused sections is the AI Transformation & Scaleup Track.
\nRather than concentrating only on new models or AI products, the sessions deal with what happens when companies try to integrate AI into actual operations.
\nA case study will examine how AI was introduced into business operations, followed by a masterclass on moving from isolated AI pilots towards what the programme describes as AI-native operations.
\nAnother panel asks a question increasingly relevant to startups: if AI tools make it easier for almost anyone to build a product, what separates the companies that successfully reach the market?
\nThe scaleup programme also covers leadership, marketplace growth, access to European funding and the operational changes founders face as their companies grow.
\nSaturday afternoon turns further towards financing and scale.
\nSessions examine what founders should protect during negotiations, how companies can access European funding without immediately giving up equity and what happens during legal and investment due diligence before money changes hands.
\nThe programme also brings founders, investors and policymakers together during a dedicated brunch and networking period.
\nFor more established companies, these sessions move beyond the early-stage question of how to launch and towards how to finance and structure the next stage of growth.
\nTrust and cybersecurity form another part of Saturday’s agenda.
\nThe track brings security and privacy into the broader conversation about AI adoption, alongside sessions involving academic and industry participants.
\nFINKI is also represented in the programme, linking the summit with North Macedonia’s academic and technical community.
\nSaturday also brings one of the summit’s main pitching competitions.
\nThe Startup World Cup programme begins in the afternoon, followed by a pitching final involving selected startups.
\nCompanies listed in the competition include Intrascope.app, VNYX B.V., Telentir, Developico/InSight, Memoria Global, Safe Watch App, Athena Chat, Factum AI, SupraOS and Impala Research.
\nEach company receives a short pitching window followed by questions before the jury deliberates and announces the winner.
\nThe competition connects the Skopje event with Startup World Cup’s international startup competition.
\nStartup Revolution also runs a separate programme away from the main Philharmonic stages.
\nAt INNOX, student teams participating in the I Have an Idea 4 incubation programme will spend Saturday working with mentors, preparing presentations and pitching startups they have developed and validated.
\nThe programme begins with one-to-one mentoring before moving into pitch preparation.
\nThirty teams are scheduled across two pitching blocks in the evening.
\nThis creates a second startup pipeline within the summit, aimed at students and new founders rather than more mature companies competing in the Startup World Cup.
\nSeveral smaller formats run alongside the main stages.
\nThe Founder Games Live Challenge takes place on Saturday morning, while a Premium workshop focuses on turning personal expertise and professional visibility into revenue.
\nCombined with mentoring, pitching and founder-focused sessions, these activities give the summit a practical layer alongside the main conference programme.
\nSaturday’s formal programme does not end with the final pitch.
\nAn invite-only Women’s Mixer brings together founders, investors and government representatives in the evening.
\nThe final night then moves back to Srekja for the Unicorn Party, which the organisers position as the summit’s closing social event before Sunday’s final activities.
\nStartup Revolution closes differently from a traditional technology conference.
\nOn Sunday morning, the programme connects with the Wizz Air Skopje Marathon, which begins at 08:00 near the Assembly of North Macedonia.
\nParticipants can run, walk or attend the race before Startup Revolution formally closes in the afternoon.
\nThe summit therefore moves across three main locations during its four days: Srekja and Startup Club Skopje for the opening and evening events, the Macedonian Philharmonic for the main conference on Friday and Saturday, and central Skopje for Sunday’s marathon programme.
\nFor visitors trying to navigate the schedule, Friday and Saturday contain most of the summit’s professional programme.
\nFriday is particularly relevant for early-stage founders, people developing their first startup, mentors and those interested in specialised sectors such as fintech, healthtech and climate tech.
\nSaturday broadens the conversation towards AI transformation, scaleups, investors, public policy, cybersecurity, government digitalisation and startup financing, while also hosting the two major pitching programmes.
\nThursday and Sunday provide the opening and closing around those two conference days.
\nWith simultaneous stages, workshops, mentoring sessions, investor meetings and side events, attending Startup Revolution will require some choices. The full agenda is available through the summit’s official website, where participants can plan the tracks most relevant to them.
","featured_image_id":374,"intro_image_id":374,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":8,"view_count":147,"ordering":0,"published_at":"2026-10-01 19:29:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-01 19:29:45","updated_at":"2026-10-12 21:28:05","key_takeaways":"[\"Startup Revolution runs from October 1 to 4 in Skopje, with the main conference programme concentrated on October 2 and 3.\",\"Friday focuses heavily on early-stage startups, founders, AI and emerging technology, fintech, healthtech, climate tech, mentoring and finding first customers.\",\"Saturday moves towards AI transformation, scaleups, investment, government policy, cybersecurity and startup financing.\",\"The programme includes the launch of the Skopje Innovation District, Startup World Cup pitching and pitches from 30 student teams in the I Have an Idea 4 programme.\",\"Workshops, one-to-one mentoring, investor meetings, networking and side events run alongside the main conference stages.\"]","faq":"[{\"q\":\"When does Startup Revolution AI Summit 2026 take place?\",\"a\":\"Startup Revolution runs from October 1 to October 4, 2026, in Skopje, North Macedonia.\"},{\"q\":\"Where is the main Startup Revolution programme held?\",\"a\":\"The main conference programme on October 2 and 3 takes place at the Macedonian Philharmonic. Opening events are held at Srekja and Startup Club Skopje, while the final day connects with the Wizz Air Skopje Marathon.\"},{\"q\":\"What can startup founders expect?\",\"a\":\"The programme includes founder sessions, sales and go-to-market workshops, legal guidance, mentoring, investor networking, pitch preparation, scaleup sessions and startup competitions.\"},{\"q\":\"Is Startup Revolution only focused on artificial intelligence?\",\"a\":\"No. AI is a major part of the programme, but the summit also covers startups, investment, fintech, healthtech, energy and climate technology, cybersecurity, public policy, digital government, marketing and company growth.\"}]","sources":"[{\"title\":\"Startup Revolution AI Summit — Official Agenda\",\"url\":\"https://startuprevolution.ai/pages/agenda\"},{\"title\":\"Startup Revolution AI Summit — Venue Information\",\"url\":\"https://startuprevolution.ai/pages/venue\"}]","category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/fcb5754a231cf2daafe44e37f1a61c00.webp","featured_image_alt":"Startup revolution","featured_image_width":1600,"featured_image_height":1066,"intro_image_path":"/api/uploads/2026/10/fcb5754a231cf2daafe44e37f1a61c00.webp","intro_image_width":1600,"intro_image_height":1066,"seo_title":"Startup Revolution Begins in Skopje: What to Expect","meta_description":"Startup Revolution opens in Skopje with four days of AI, startups, investment, policy, pitching and business-focused sessions.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":150,"title":"More Than 650 Join VEZILKA’s First Public AI Course","slug":"more-than-650-join-vezilka-s-first-public-ai-course","short_description":"","content":"More than 650 people joined the first session of VEZILKA’s new public AI-literacy programme in Skopje, offering an early indication of the level of public interest in practical artificial intelligence education in North Macedonia.
\nThe first session of “AI Literacy for Everyone” was held on September 29 at the Faculty of Computer Science and Engineering (FINKI), UKIM, with 580 participants joining online and 75 attending in person.
\nThe course is part of the wider training programme being developed by VEZILKA, North Macedonia’s AI Factory Antenna initiative.
\nIts opening session, “Introduction to AI,” was led by Prof. Andrea Kulakov and Elena Atanasoska and focused on the fundamentals of artificial intelligence, including how AI systems learn from data, where they can be useful, their limitations and why human verification remains necessary.
\nThe programme is aimed at people without a technical background rather than AI specialists.
\nThat distinction matters.
\nPublic discussion around AI in North Macedonia has expanded quickly during the past year, but access to structured training for ordinary citizens remains more limited than the availability of tools such as ChatGPT, Gemini and other AI assistants.
\nVEZILKA’s programme attempts to address that gap by moving beyond awareness campaigns and introducing a structured AI-literacy curriculum.
\n![\"Vezilka\"](\"/api/uploads/2026/10/ccf67523217c959cbdf79915552d6487.webp\")
“AI Literacy for Everyone” consists of three three-hour webinars, delivered in a hybrid format at FINKI and online.
\nThe training is free and forms part of VEZILKA’s 2026–2027 education programme.
\nThe programme is designed to help participants understand what artificial intelligence is, how it differs from conventional software and how AI tools can be used critically and safely.
\nVEZILKA’s broader training portfolio goes considerably further.
\nPlanned tracks include programmes for public administration, small and medium-sized businesses, teachers, university students, researchers and engineers, as well as specialised training for professionals working in health, energy, agriculture and government.
\nThat makes the public course one part of a wider attempt to build AI capacity alongside the computing infrastructure being developed through the project.
\nAI Balkans has previously reported on VEZILKA primarily through North Macedonia’s efforts to connect with Europe’s AI supercomputing network, develop Macedonian-language AI resources and introduce artificial intelligence into formal secondary education.
\nThe latest development addresses another part of the same challenge: whether enough people outside the technology sector understand how to use AI effectively and where its limits are.
\nNorth Macedonia can expand access to computing infrastructure, datasets and advanced AI systems, but those investments have limited value without people who know how to use them.
\nThe turnout at the first VEZILKA session therefore provides more than an attendance figure.
\nIt suggests that demand for basic AI education already exists beyond researchers, developers and technology companies.
\nThat demand is likely to become more important as AI tools move further into workplaces, schools and public institutions.
\nVEZILKA is also developing training specifically for civil servants and small businesses, two areas where the practical effects of AI adoption could be significant.
\nThe programme for public administration is expected to focus on the use of AI within government, while a separate course will target SMEs. Dates for those programmes have not yet been announced.
\nOther planned programmes include advanced training using GPU infrastructure and sector-specific courses built around Macedonian datasets and use cases.
\nThe project says its training activities will be delivered in Macedonian and English.
\nThe question now is whether the level of interest seen in the first public session can be maintained as the programme expands into more specialised areas.
\nWith 655 participants at its opening session, VEZILKA has at least demonstrated that AI literacy is no longer a niche subject in North Macedonia.
","featured_image_id":330,"intro_image_id":330,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":62,"ordering":0,"published_at":"2026-10-01 08:30:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-01 08:30:58","updated_at":"2026-10-01 08:33:59","key_takeaways":"[\"655 participants joined VEZILKA’s first “AI Literacy for Everyone” session on September 29.\",\"The programme targets the general public and requires no technical background.\",\"VEZILKA is also preparing AI training for government, SMEs, educators, researchers and sector professionals.\"]","faq":"[{\"q\":\"What is “AI Literacy for Everyone”?\",\"a\":\"It is a free introductory AI course organised through VEZILKA for members of the general public without a technical background.\"},{\"q\":\"How many people attended the first session?\",\"a\":\"VEZILKA reported 580 online participants and 75 attending in person at FINKI, for a total of 655 participants.\"},{\"q\":\"How long is the programme?\",\"a\":\"The current course consists of three webinars lasting three hours each.\"},{\"q\":\"What other AI training is VEZILKA planning?\",\"a\":\"Future programmes are planned for public administration, SMEs, educators, researchers, engineers and professionals in areas including health, agriculture and energy.\"}]","sources":"[{\"title\":\"VEZILKA, “More Than 650 Participants Join VEZILKA’s First AI Course”\",\"url\":\"https://vezilka.ai/more-than-650-participants-join-vezilka%E2%80%99s-first-ai-course/\"},{\"title\":\"VEZILKA, “Training Courses 2026–2027”\",\"url\":\"https://vezilka.ai/training/training-courses/\"}]","category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/7a1d720f40ed91f3b1aee05102a7c36b.webp","featured_image_alt":"Vezilka Lecture","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/7a1d720f40ed91f3b1aee05102a7c36b.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"650+ Join VEZILKA’s First Public AI Course","meta_description":"More than 650 people joined VEZILKA’s first public AI-literacy session in North Macedonia, with 580 online and 75 at FINKI.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":147,"title":"North Macedonia Tests AI System to Detect Landfill Fires","slug":"north-macedonia-tests-ai-system-to-detect-landfill-fires","short_description":"","content":"North Macedonia is preparing a pilot system that will use cameras, intelligent video analytics and artificial intelligence to detect smoke, fires and other hazards at landfills.
\nThe Ministry of Environment and Spatial Planning, the Crisis Management Center and Skopje-based technology company Modirum Platforms signed a memorandum on September 26 establishing the framework for the project.
\nUnder the pilot, landfills will be monitored through cameras connected to video-analysis technology capable of identifying signs of smoke, fire and other potential risks.
\nThe Crisis Management Center will provide 24-hour monitoring and alert the relevant institutions when the system identifies a potential danger, according to the ministry statement reported by KOHA.
\nThe aim is to shorten the time between the first signs of a fire and the response from emergency and environmental authorities.
\nModirum Platforms will provide the equipment for temporary use for a period of one year as part of the pilot.
\nThe government has not yet announced which landfill sites will be included in the first phase, nor how many cameras or monitoring points will be deployed.
\nResults from the trial will be used to assess whether the system should later be expanded across the country.
\nThe project comes at a time when fires at landfills and other open areas remain a recurring problem in North Macedonia.
\nOn September 25 alone, the Crisis Management Center reported fires involving waste at several locations, including the Krasta landfill near Kumanovo and an area near Veles.
\nTwo days earlier, landfill fires were also reported among a wider series of open-space fires across the country.
\nThe technology being tested is a relatively practical use of AI.
\nInstead of relying only on a person watching multiple camera feeds, computer-vision systems can analyse video continuously and flag patterns that may indicate smoke or fire.
\nThe final decision and response would still sit with the Crisis Management Center and other responsible institutions.
\nThat distinction is important because the pilot is not an autonomous emergency-response system. AI is being introduced as an additional detection and monitoring layer.
\nIf it works as intended, the main value would be earlier identification of incidents, particularly at locations where fires can spread before authorities receive a report.
\nNorth Macedonia has faced significant pressure from wildfires and open-area fires in recent years.
\nA recent State Audit Office assessment found weaknesses in the country's disaster protection and rescue system, including limited investment in equipment and modernization. The audit also highlighted forest fires as one of the country's major disaster risks.
\nThe new landfill-monitoring pilot addresses only one part of that broader problem.
\nIt does, however, provide a concrete example of AI being introduced into a public-safety workflow in North Macedonia rather than being discussed only through conferences, strategies or experimental applications.
\nThe next test will be practical: whether the system can reliably detect incidents early enough to improve response times, and whether the results justify extending the technology beyond the initial pilot.
","featured_image_id":327,"intro_image_id":327,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":36,"ordering":0,"published_at":"2026-09-27 16:07:33","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-27 16:07:33","updated_at":"2026-09-27 16:07:33","key_takeaways":"[\"North Macedonia is launching a one-year pilot using cameras, intelligent video analytics and AI to detect landfill fires and other hazards.\",\"The Crisis Management Center will monitor the system 24/7 and notify relevant institutions when risks are detected.\",\"Results from the pilot will be used to decide whether the system should be expanded nationally.\"]","faq":"[{\"q\":\"What is North Macedonia testing?\",\"a\":\"A video-monitoring system that combines cameras, intelligent video analytics and artificial intelligence to detect smoke, fires and other hazards at landfills.\"},{\"q\":\"Who is involved in the project?\",\"a\":\"The Ministry of Environment and Spatial Planning, the Crisis Management Center and Modirum Platforms.\"},{\"q\":\"How long will the pilot run?\",\"a\":\"The equipment will be provided for temporary use for one year.\"},{\"q\":\"Will AI automatically respond to fires?\",\"a\":\"No. The system is designed to help detect potential hazards. The Crisis Management Center will monitor alerts and notify the relevant institutions.\"}]","sources":null,"category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/dbca512e30aa1f1f72ea4e477b6e9dc4.webp","featured_image_alt":"Six people in business attire meet around a conference table with water bottles and phones.","featured_image_width":1536,"featured_image_height":1025,"intro_image_path":"/api/uploads/2026/09/dbca512e30aa1f1f72ea4e477b6e9dc4.webp","intro_image_width":1536,"intro_image_height":1025,"seo_title":"North Macedonia Tests AI to Detect Landfill Fires","meta_description":"North Macedonia will pilot AI-powered camera monitoring to detect smoke and fires at landfills and improve early emergency response.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":137,"title":"AI Tech Summit Day One Keeps Humans in the Loop","slug":"ai-tech-summit-day-one-keeps-humans-in-the-loop","short_description":"","content":"Human involvement remained one of the clearest themes running through the opening sessions of AI Tech Summit “Filip Avramchev” in Skopje on Tuesday, as speakers from Microsoft, NVIDIA and the autonomous-driving field discussed very different parts of the AI stack but repeatedly returned to the role of people in designing, supervising and using these systems.
\nThe summit opened on September 22 at the National Opera and Ballet, with the first sessions moving from enterprise AI and infrastructure to autonomous systems and real-world applications.
\nFatos Ismali, Chief AI Architect at Microsoft UK, opened the programme with a discussion focused on enterprise AI and the challenge of moving artificial intelligence from experiments into real organisational use.
\nA recurring point in that discussion was that introducing AI into a company is not simply a technical exercise. Organisations still need people who understand the process, evaluate the output and decide where AI should and should not be used.
\nThe same principle appeared from a different direction in the presentation by Koen van den Berg, AI Factory and Data Center Readiness Lead for EMEA at NVIDIA.
\nVan den Berg focused on the infrastructure behind modern AI, including the growing importance of computing capacity, data centres and the systems needed to support increasingly demanding models.
\nBut even in a discussion centred on infrastructure, the wider message remained tied to people. AI systems may become more capable and more autonomous, but their usefulness still depends on how they are designed, deployed and governed by the organisations using them.
\nThat question became even more concrete when the programme moved into autonomous driving.
\nAtanasko Boris Mitrev, a self-driving car engineer and AI and robotics researcher associated with the Autoware Foundation, discussed how autonomous vehicles perceive and respond to the environment around them.
\nAutonomous driving is often presented as one of the clearest examples of machines taking over tasks previously performed entirely by people. Yet the discussion again returned to human responsibility, especially around system design, testing, safety and the limits placed on automated decision-making.
\nThat overlap between the first sessions was notable.
\nMicrosoft approached AI from the perspective of organisations and production systems. NVIDIA looked at the infrastructure required to run increasingly capable models. Autonomous driving brought AI into the physical world, where decisions can have immediate consequences.
\nDespite those differences, the underlying message was similar: more automation does not automatically mean less human responsibility.
\nIf anything, the expansion of AI into more complex processes appears to be increasing the importance of human judgement, particularly when systems move from assisting people to taking actions on their behalf.
\nThat is also a relevant discussion for the Western Balkans, where companies and institutions are beginning to move beyond basic experimentation with generative AI and toward questions of integration, governance and accountability.
\nThe challenge is no longer only whether organisations can access powerful AI tools. It is whether they have the people, processes and oversight needed to use them responsibly and effectively.
\n![\"818062370_17922127380432408_3227289845579453813_n.jpg\"](\"/api/uploads/2026/09/9fc494199e75938866fa79a95968694e.webp\")
![\"815267678_17922107223432408_3700574278079193273_n.jpg\"](\"/api/uploads/2026/09/b07f7a0cb639aee0107ecefdda6db1fd.webp\")
![\"792089532_17922121620432408_6279880774790567305_n.jpg\"](\"/api/uploads/2026/09/a1363f4edafacce47c261d1f39aeb455.webp\")
![\"791922811_17922121212432408_1016001326841381599_n.jpg\"](\"/api/uploads/2026/09/dccd9ab63d0d960be50bf64acf2b4c34.webp\")
![\"791143531_17922107850432408_5628362771352023028_n.jpg\"](\"/api/uploads/2026/09/fa3829382bb7d0f8e739cfb1d7620349.webp\")
AI Tech Summit “Filip Avramchev” continues on September 23, with further sessions covering AI agents, software engineering, robotics, enterprise applications and other areas of applied artificial intelligence. AINOW Society is supporting this year’s edition as a community partner.
","featured_image_id":298,"intro_image_id":298,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":48,"ordering":0,"published_at":"2026-09-22 18:54:58","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-22 18:54:58","updated_at":"2026-09-22 18:54:58","key_takeaways":null,"faq":null,"sources":null,"category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/c36ad774c5a6d75e6e7a119095819f5d.webp","featured_image_alt":"Empty AI Tech Summit stage with four panel chairs, a branded screen and magenta-lit lettering.","featured_image_width":1440,"featured_image_height":958,"intro_image_path":"/api/uploads/2026/09/c36ad774c5a6d75e6e7a119095819f5d.webp","intro_image_width":1440,"intro_image_height":958,"seo_title":"AI Tech Summit Day One Keeps Humans in the Loop","meta_description":"Day one of AI Tech Summit Skopje 2026 focused on human oversight, enterprise AI, infrastructure and autonomous systems.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":129,"title":"Two AI Summits Put Skopje at the Centre of the Regional Tech Conversation","slug":"two-ai-summits-put-skopje-at-the-centre-of-the-regional-tech-conversation","short_description":"","content":"Skopje will host two major technology gatherings within less than two weeks this autumn, bringing AI professionals, founders, investors, researchers and technology companies to North Macedonia for two events with distinctly different formats.
\nAI Tech Summit “Filip Avramchev” will take place on 22 and 23 September at the National Opera and Ballet, followed by Startup Revolution AI Summit from 1 to 4 October.
\n![\"Ai](\"/api/uploads/2026/09/038a2881f3d1b2c5c7526fb8dd3d0d91.webp\")
Together, the events reflect a technology community in North Macedonia that is becoming increasingly active around artificial intelligence, entrepreneurship and the infrastructure needed to turn new technology into working products and businesses.
\nThey are also approaching that conversation from different directions.
\nThe 2026 edition of AI Tech Summit “Filip Avramchev” is structured around four parts: a Main Stage, AI Labs, an Expo and a student hackathon.
\nIts programme is particularly focused on applied artificial intelligence, enterprise adoption, data infrastructure, robotics and the technical systems behind AI deployment. The announced speaker lineup includes professionals working across Microsoft, NVIDIA and a number of AI and technology companies.
\nAmong them is Fatos Ismali, Chief AI Architect at Microsoft, who is expected to address the challenge of taking AI systems from experimentation into production.
\nKoen van den Berg, AI Factory and Data Centre Readiness Lead for EMEA at NVIDIA, will bring another part of the AI discussion to Skopje: computing infrastructure, AI-ready data centres and the growing concept of sovereign AI.
\nThe programme is also moving beyond generative AI. Recent additions include researchers and companies working in autonomous vehicles, humanoid robotics and physical AI, including systems designed to operate and make decisions outside the traditional computer interface.
\nThat gives the summit a strongly technical character, while its Labs, Expo and hackathon are intended to extend the programme beyond presentations and panel discussions.
\nNine days later, Startup Revolution AI Summit will open in Skopje with a considerably broader startup and business focus.
\n![\"startuprevolution.jpg\"](\"/api/uploads/2026/09/a18a6c3064b9bcdcccf934a4d0eab12f.webp\")
Running from 1 to 4 October, the event combines artificial intelligence with entrepreneurship, investment, cybersecurity, health technology, fintech, energy and climate technology.
\nIts programme includes several stages, startup presentations, investor meetings, networking activities and sessions focused on the development of technology companies in Central and Southeast Europe.
\nStartup Revolution is organised by Startup Club Skopje and positions AI as part of a wider conversation about how companies are created, financed and scaled.
\nThat distinction matters.
\nWhile many technology conferences now include AI somewhere in their programmes, the rapid adoption of generative AI has increasingly blurred the line between an “AI company” and a company simply using AI as part of its product or operations.
\nStartup Revolution appears set to explore that intersection directly, bringing founders and investors into the same programme as people working on AI and emerging technologies.
\nIts October agenda includes dedicated tracks covering AI and emerging technologies alongside ecosystem development, data, healthtech, business and investment.
\nThe proximity of the two events gives Skopje an unusual concentration of AI and technology activity over a short period.
\nMore importantly, their different formats offer two views of where the regional technology sector is heading.
\nOne is centred more closely on AI technology itself: how systems are built, deployed and connected with infrastructure, robotics and enterprise operations.
\nThe other places AI inside the broader machinery of startups, investment, company building and regional entrepreneurship.
\nFor North Macedonia, that combination may be more relevant than simply counting the number of AI events taking place.
\nThe larger question is what happens after the conferences finish: whether conversations turn into products, research collaborations, investments, new companies or stronger connections between North Macedonia and the wider European technology ecosystem.
\nAI Balkans will follow both summits as part of its coverage of artificial intelligence and the technology ecosystem across North Macedonia and the wider Balkan region.
","featured_image_id":281,"intro_image_id":281,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":10,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":76,"ordering":0,"published_at":"2026-09-09 09:53:56","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-09 09:53:56","updated_at":"2026-09-09 09:53:56","key_takeaways":null,"faq":null,"sources":null,"category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Guest Contributor","author_slug":"guest-contributor","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/d8fe0944d9f244a62629a6c3d79d8e65.webp","featured_image_alt":"Speakers and panelists occupy an AI-themed stage before a seated audience, with city landmarks and a digital face onscreen.","featured_image_width":1448,"featured_image_height":1086,"intro_image_path":"/api/uploads/2026/09/d8fe0944d9f244a62629a6c3d79d8e65.webp","intro_image_width":1448,"intro_image_height":1086,"seo_title":"Two AI Summits Put Skopje on the Regional Tech Map","meta_description":"Skopje will host AI Tech Summit and Startup Revolution AI Summit in autumn 2026, bringing AI, startups, investment and technology into focus.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":128,"title":"North Macedonia Introduces AI as an Elective Subject in Secondary Schools","slug":"north-macedonia-introduces-ai-as-an-elective-subject-in-secondary-schools","short_description":"","content":"Secondary-school students across North Macedonia will be able to study artificial intelligence as a new elective subject from the 2026/27 school year, as the country begins bringing formal AI education into classrooms.
\nThe new subject, Introduction to Artificial Intelligence, consists of a 36-hour curriculum developed through the VEZILKA project in cooperation with the Bureau for Development of Education.
\nThe programme has now been officially included in the country’s secondary-education curriculum.
\nStudents will be introduced to the basic concepts behind artificial intelligence, but the subject goes beyond explaining how AI works.
\nThe curriculum also covers practical use of AI tools, prompting, fact-checking, AI-generated content and the safe, ethical and responsible use of artificial intelligence.
\nThat makes the programme one of the more concrete attempts in the Western Balkans to move AI literacy from occasional workshops and extracurricular activities into formal secondary education.
\nIntroducing a new subject also means preparing the people who will teach it.
\nVEZILKA organised the first online training sessions for secondary-school teachers on September 2 and 3, bringing together educators from schools across North Macedonia.
\nThe organisation says it will continue supporting teachers as the programme is introduced and plans to develop additional educational resources around AI.
\nThat part of the rollout may prove just as important as the curriculum itself.
\nGenerative AI tools have entered classrooms much faster than formal education systems have been able to respond. Students already use systems such as ChatGPT, Gemini and other AI assistants for research, writing, translation and problem-solving, often without consistent guidance from schools.
\nA formal subject gives schools an opportunity to deal with those questions directly rather than treating AI only as a technology students encounter outside the classroom.
\nThe inclusion of fact-checking and responsible use is particularly relevant.
\nKnowing how to generate an answer from an AI system is increasingly a basic digital skill. Knowing whether that answer is accurate is a different matter.
\nGenerative AI systems can produce incorrect information, fabricated sources and convincing content that lacks evidence. Students also face an online environment where AI-generated images, video and audio are becoming harder to distinguish from authentic material.
\nTeaching AI literacy therefore involves more than showing students how to operate a tool.
\nIt requires understanding where AI systems can fail, how outputs should be verified, what information should not be shared with external platforms and where human judgement remains necessary.
\nThe new Macedonian curriculum appears designed around that broader interpretation of AI literacy rather than simply teaching students to use popular applications.
\nThe secondary-school programme is also part of a broader education and training programme being developed through VEZILKA, North Macedonia’s AI Factory Antenna initiative.
\nVEZILKA is preparing AI courses for several groups, including educators, researchers, engineers, public-sector employees, sector professionals and citizens without a technical background.
\nIts wider training programme begins in September 2026 and includes introductory AI-literacy courses as well as more specialised programmes.
\nAI Balkans has previously reported on VEZILKA primarily through North Macedonia’s connection to Europe’s AI computing infrastructure and efforts to strengthen Macedonian-language AI.
\nThe introduction of a school subject represents a different part of the same strategy: developing people who can actually use and understand the technology.
\nInfrastructure can give researchers and companies access to computing resources. Education determines how widely those capabilities can eventually be understood and applied.
\nNorth Macedonia is not alone in trying to work out how AI belongs in education.
\nAcross the Balkans, schools and universities are increasingly being forced to make decisions about AI use, academic integrity, teacher training and digital literacy.
\nMost of those responses remain fragmented.
\nSome institutions are issuing guidelines. Others are training educators or experimenting with AI-supported teaching. Formal AI education at school level remains less common.
\nNorth Macedonia’s new elective will therefore be worth watching beyond the country itself.
\nThe first practical test begins during the 2026/27 school year: how many schools offer the subject, how many students choose it and whether teachers receive enough support to turn the curriculum into meaningful classroom learning.
\nFor now, the country has taken a clear step beyond simply discussing the need for AI literacy.
\nIt has put it on the timetable.
","featured_image_id":105,"intro_image_id":105,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":43,"ordering":0,"published_at":"2026-09-06 12:47:43","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-06 12:47:43","updated_at":"2026-09-06 12:47:43","key_takeaways":null,"faq":null,"sources":null,"category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/837710df9b378c847553a04607a72c07.webp","featured_image_alt":"Vezilka photo","featured_image_width":1200,"featured_image_height":799,"intro_image_path":"/api/uploads/2026/06/837710df9b378c847553a04607a72c07.webp","intro_image_width":1200,"intro_image_height":799,"seo_title":"North Macedonia Introduces AI Subject in Secondary Schools","meta_description":"North Macedonia is introducing a 36-hour AI elective for secondary-school students from the 2026/27 school year.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":123,"title":"When AI Betrays Law and Justice: Skopje Conference Brings Legal and AI Experts Together","slug":"when-ai-betrays-law-and-justice-skopje-conference-brings-legal-and-ai-experts-together","short_description":"","content":"Skopje will host an international scientific conference on 16 September bringing together legal scholars, researchers, digital policy experts and technology practitioners to examine how artificial intelligence is affecting law, human rights, constitutionalism and democratic accountability.
\n
The conference, “When AI Betrays Law and Justice: Human Rights, Constitutionalism and Democratic Accountability in the Algorithmic Age,” will take place at Hotel Panoramika and is organised by the Faculty of Law “Iustinianus Primus” at Ss. Cyril and Methodius University in Skopje.
\nIt is being held within the research project “Alignment of the Macedonian Legal System with the European Union Artificial Intelligence Legislation,” funded by North Macedonia’s Ministry of Education and Science.
\nThe event brings together an international group of speakers alongside academics and researchers from North Macedonia, with the programme moving from broader questions of democracy and justice to new empirical findings on how AI is understood and used in the country.
\n
Among the international speakers is Paul Nemitz, former Principal Adviser for Strategies for Digital Transition at the European Commission and Visiting Professor of Law at the College of Europe.
\nNemitz will address the relationship between Big Tech, artificial intelligence, democracy and the rule of law.
\nJudit Varga, Founding Patron of the Central European Academy at the University of Miskolc, will examine algorithmic decision-making and its implications for constitutional government.
\nThe international programme also includes Laura Kask, CEO of Estonian digitalisation consultancy Proud Engineers, whose contribution will focus on trust in digital societies and the challenges of designing institutions that remain accountable as digital systems become more influential.
\nMax Hubner, Founder and CEO of The Change Lawyer in the Netherlands, will focus on the relationship between artificial intelligence, people and justice.
\nTheir contributions will be joined by perspectives from academics at the Faculty of Law “Iustinianus Primus”.
\n
Professor Aleksandar Spasov will examine how AI affects fundamental concepts in law, while Professor Boban Misoski will address procedural justice in criminal trials and the possible role of artificial intelligence in judicial processes.
\nProfessor Lazar Jovevski will focus on AI, labour markets and the wider consequences of technological change for people and work.
\nTogether, these discussions place AI within a legal and institutional context rather than treating it only as a technological development.
\nThe central issue is increasingly one of responsibility. As AI-supported systems enter public administration, employment, business and legal processes, institutions need to determine where automated decision-making is appropriate, how those decisions can be challenged and who remains accountable when something goes wrong.
\n
A central part of the conference will be dedicated specifically to AI in North Macedonia, with the presentation of new empirical findings examining public perceptions, corporate practices and institutional readiness.
\nThe research team is coordinated by Professor Tanja Karakamisheva-Jovanovska from the Faculty of Law “Iustinianus Primus”.
\nThe team also includes Professor Hristina Runcheva Tasev, Professor Dragan Gocevski, Assistant Professor Konstantin Bitrakov and Assistant Katerina Zhateva, all from the Faculty of Law.
\nThe research section will also include expert contributions from Professor Vladimir Bozhinovski and Suad Seferi, President and Founder of AI NOW Society.
\nThe study is particularly relevant because much of the discussion around AI in North Macedonia and the wider Western Balkans has so far focused on strategies, regulation, digital transformation and adoption.
\nEmpirical research can provide a different picture.
\nIt can show how people actually perceive AI, how companies are approaching its use and whether institutions have the structures, knowledge and capacity required to manage it responsibly.
\nThat becomes increasingly important as North Macedonia works toward closer alignment with European Union legislation while AI systems are already entering workplaces, universities, companies and public institutions.
\nThe regional dimension of the discussion will also be addressed through “Beyond Adoption: AI Readiness in the Western Balkans,” presented by Suad Seferi.
\nThe contribution will look beyond the number of organisations using AI and toward a wider question: whether institutions across the Western Balkans are genuinely prepared to manage the technology.
\nAI adoption and AI readiness are not necessarily the same thing.
\nAn organisation may already use generative AI tools while lacking internal policies governing their use. A company may introduce automation without clearly assigning responsibility for its outputs. Public institutions may experiment with AI before developing procedures for transparency, human oversight or risk assessment.
\nReadiness therefore involves more than access to technology. It includes institutional capacity, governance, skills, legal understanding, infrastructure and the ability to manage risk.
\nFor Western Balkan countries, these issues are closely connected with European integration and the gradual alignment of domestic frameworks with EU digital and AI regulation.
\nThe conference is being led by academics from the Faculty of Law “Iustinianus Primus”, with Tanja Karakamisheva-Jovanovska serving as project coordinator. Opening remarks will also include Sasho Georgievski, Dean of the Faculty of Law.
\nThe wider group involved in the event reflects the different areas now being affected by AI: constitutional law, criminal justice, labour, public administration, digitalisation, institutional governance and technology policy.
\nThat multidisciplinary approach matters because many of the hardest questions surrounding AI cannot be answered by engineers alone.
\nWhen an automated system affects employment, legal rights, access to services or institutional decisions, technical performance is only one part of the problem. Law, accountability and human responsibility remain equally important.
\nFor North Macedonia, the conference also creates an opportunity to connect European discussions around AI regulation with evidence about what is actually happening inside the country.
\nThe findings presented in Skopje could help move the local debate away from general discussions about AI adoption and toward more specific questions about institutional preparedness, governance and responsibility.
\n“When AI Betrays Law and Justice” will take place on 16 September 2026 at Hotel Panoramika in Skopje.
\n
Participation is free, with places limited. Registration is open until 12 September 2026.
","featured_image_id":269,"intro_image_id":269,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":5,"view_count":162,"ordering":0,"published_at":"2026-09-04 12:56:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-04 12:56:03","updated_at":"2026-09-04 12:56:53","key_takeaways":null,"faq":null,"sources":null,"category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/f4feaa8c83cf7349d9d43e70474b2e10.webp","featured_image_alt":"Poster for “When AI Betrays Law and Justice” with Lady Justice, an AI face, and Skopje’s riverfront.","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/09/f4feaa8c83cf7349d9d43e70474b2e10.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Skopje Conference Examines AI, Law and Accountability","meta_description":"International experts and Macedonian researchers meet in Skopje to discuss AI, law, human rights and new research on North Macedonia’s AI readiness.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":108,"title":"AINOW v0.94 Expands From Education Into Business With 3,270-Prompt AI Library","slug":"ainow-v0-94-expands-from-education-into-business-with-3-270-prompt-ai-library","short_description":"","content":"North Macedonia-based AINOW Society has released version 0.94 of its AI literacy platform, adding a new professional AI library alongside its existing education product and expanding its resources for universities.
\nThe release reorganises edu.ainow.mk around two applications: AINOW Education for schools and higher education, and the new AINOW Business platform for professionals and organisations.
\nBoth remain free to use, require no user account and have been designed to operate without a permanent internet connection.
\nThe move broadens a project that was originally focused primarily on helping teachers understand and use artificial intelligence. With v0.94, AINOW is extending the same model to professional work while keeping the software open source and available in Macedonian, Albanian and English.
\nAINOW Education now operates under /edu/ and retains the learning material and tools developed for teachers and school administrators.
The platform includes a 19-chapter AI literacy guide covering subjects from basic AI concepts and prompting to retrieval, performance, responsible use and safety.
\nIts education library now contains 1,106 prompts built around work carried out in schools and universities, including lesson preparation, communication with parents, individual education plans, administrative documentation and institutional evaluation.
\nVersion 0.94 makes Higher Education a dedicated category alongside Teachers and Administration.
\nThe new section provides material for lecturers, researchers, doctoral supervision, student services, quality assurance and university leadership, with 60 prompts per supported language across the six areas.
\nThe education platform also includes a 60-term AI glossary and a directory of 57 AI tools.
\nThe expansion reflects a wider problem across the region: much of the practical material available for using generative AI in education is still developed primarily in English and for larger education systems. AINOW's decision to maintain the platform in Macedonian and Albanian gives local institutions material that can be used without translating an English-language resource first.
\n
The biggest addition to v0.94 is AINOW Business, a separate application built around professional AI use.
\nIt launches with 3,270 prompts across 33 fields, including software development, law, medicine, finance, agriculture, journalism, public health, logistics and management.
\nThe platform also contains 2,190 predefined AI agent configurations.
\nThese provide structured roles, objectives and instructions that can be copied into compatible AI systems rather than requiring users to construct an agent configuration themselves.
\nAINOW has added several ways of navigating that material.
\nA Situations section currently provides eight guided scenarios for tasks such as negotiations and incident response. Users describe the circumstances they are dealing with and the platform helps construct a relevant prompt.
\n\nPlaybooks provide more structured, step-by-step workflows, while an Agent Builder generates customised instructions based on information supplied by the user.
\nThe Business application itself is free. Consulting, training and organisational AI adoption services are offered separately through AINOW Society's technology partner, CTRLAB.
\nOne of the more distinctive aspects of the project is its approach to distribution.
\nBoth applications are designed to run locally without a conventional server installation. A copy of the platform can be stored on a computer or USB drive and opened locally through its index.html file.
That makes the architecture relevant beyond software design.
\nSchools and organisations across the Balkans do not operate with identical infrastructure. Older computers, restricted installation permissions and inconsistent connectivity remain practical barriers in some institutions, particularly outside major urban centres.
\nThe Business application originally used a more conventional React and bundler-based architecture. For the v0.94 release, its modules and dependencies were adapted so the final application could operate from local files.
\nReact remains part of the application, but dependencies needed by the software are stored locally instead of being retrieved from third-party services when the application runs.
\nThe platform's fonts are now hosted locally as well.
\nThe same architecture supports AINOW's privacy model.
\nNeither Education nor Business requires registration. Language preferences, favourites and scratchpad information remain in the user's browser rather than being stored in a central AINOW account.
\nAccording to the organisation, the applications do not use analytics, tracking pixels or external telemetry to monitor how people use the platform.
\nThat design is particularly relevant for the education product, where teachers may be working with information connected to students, but AINOW has applied the same model to its professional platform.
\nIt is a different approach from the increasingly common model of AI software built around mandatory cloud accounts and persistent user profiles.
\nThe release was not limited to new features.
\nAINOW says it also conducted an audit of its platform statistics and multilingual interface after finding inconsistencies that had accumulated during development.
\nThe Business interface had previously advertised 3,370 prompts across 34 areas. An audit of the underlying content found the correct figures to be 3,270 prompts across 33 areas, which are now used throughout the platform.
\nThe team also found obsolete translation entries left behind after features had been removed, as well as around 40 interface placeholders that still appeared in English for Macedonian and Albanian users.
\nFollowing the cleanup, all three languages now use the same 901 active interface strings.
\nAnother discrepancy concerned the education guide, which was listed as having 20 chapters in one part of the application and 19 elsewhere. The verified number is 19.
\nPublishing those corrections alongside the larger feature release is a relatively small detail, but a useful one for an open-source project whose credibility depends partly on the accuracy of the resources it distributes.
\nAINOW Education and AINOW Business are released under the GNU GPLv3 licence, allowing the software to be inspected, modified and redistributed under the licence terms.
\nFor schools in particular, downloadable releases mean the hosted website is not the only way to access the project. A local copy can be distributed between computers without requiring each device to maintain internet access.
\n\n\n
With v0.94, AINOW is testing whether the model it developed for AI literacy in education can also work for professional adoption.
\nThe project now covers classrooms, universities and workplaces from the same multilingual, offline-capable base.
\nThe more important test comes next: whether organisations and contributors outside AINOW's existing education community begin using and extending it.
\nThe full AINOW Society release announcement is available on the organisation’s official blog
","featured_image_id":242,"intro_image_id":242,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":8,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":1,"show_table_of_contents":1,"read_time_minutes":6,"view_count":45,"ordering":0,"published_at":"2026-08-16 16:22:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-16 16:22:19","updated_at":"2026-08-16 16:30:56","key_takeaways":null,"faq":null,"sources":null,"category_name":"North Macedonia","category_slug":"north-macedonia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/9a8e84f250b054e8c7e59735b92aed7d.webp","featured_image_alt":"AINOW Society webpage offers free Education and Business platforms, including 3,270 business prompts.","featured_image_width":1487,"featured_image_height":696,"intro_image_path":"/api/uploads/2026/08/9a8e84f250b054e8c7e59735b92aed7d.webp","intro_image_width":1487,"intro_image_height":696,"seo_title":"AINOW v0.94 Adds Business AI Platform With 3,270 Prompts","meta_description":"AINOW v0.94 adds a Business platform with 3,270 prompts and 2,190 AI agents while expanding higher education and offline access.","robots":"index","canonical_url":"","og_title":"","og_description":""}]},{"category":{"id":9,"name":"Albania","slug":"albania"},"posts":[{"id":166,"title":"AllWeb Invites AI Agents to Speak at Tirana Conference","slug":"allweb-invites-ai-agents-to-speak-at-tirana-conference","short_description":null,"content":"The regional digital conference is opening its stage to AI-powered agents, inviting developers and startups to propose presentations for its November event.
\nAllWeb Digital Conf is inviting artificial intelligence agents to take the stage as conference speakers for the first time, opening a call for technology companies, startups and creative businesses ahead of its 6 November 2026 event in Tirana, Albania.
\nThe initiative, announced on 12 October, will allow selected AI agents to deliver presentations to a live business audience. Rather than showcase products, the organisers want participating agents to share practical knowledge and examples that professionals can apply in their work.
\nIt is an unusual experiment for a conference traditionally built around human speakers. The question AllWeb wants to explore is whether an AI agent can deliver a session that is genuinely useful, engaging and relevant to an audience of business and technology professionals.
\nThe call is open to teams developing AI agents capable of presenting ideas and interacting with conference participants.
\nApplicants will be expected to explain their proposed presentation topic, provide practical case studies and describe how their agent would engage with the audience.
\nSuggested areas include digital marketing, customer experience, productivity, innovation and business growth.
\nThe organisers say proposals will be assessed based on their relevance and potential value to attendees, rather than the technical novelty of the AI system alone.
\n“For years, we’ve brought experts to the AllWeb stage to share what they know about business, marketing and technology,” said Gerti Boshnjaku, co-organiser of AllWeb Digital Conf and CEO of Business Magazine Albania.
\nBoshnjaku said the organisers now want to give AI agents the same challenge, focusing on useful knowledge and practical examples instead of technology demonstrations.
\nThe distinction matters. An AI agent capable of producing a presentation is not necessarily capable of responding reliably to questions, adapting to an audience or explaining complex subjects accurately.
\nAllowing agents onto the conference stage creates an opportunity to examine those capabilities in a setting where audiences expect more than a prepared script.
\nAllWeb Digital Conf has been part of the regional technology and business events landscape since its establishment in North Macedonia in 2012.
\nSince 2016, editions of the conference have also been organised in Albania by Business Magazine Albania, bringing together professionals working in digital marketing, communications, technology and entrepreneurship.
\nDarko Buldioski, Programme Director of AllWeb Digital Conf, described the decision to invite AI agents as speakers as a departure from the traditional conference format.
\n“Ten years from now, this might feel completely normal,” Buldioski said, reflecting on the possibility of AI agents becoming a regular part of professional events.
\nThe 2026 conference will take place under the theme “All for AI & AI for All”, reflecting the growing role of artificial intelligence in business and professional workflows.
\nAccording to the organisers, more than 350 business leaders, marketers, entrepreneurs and technology professionals are expected at the event.
\nThe conference is scheduled for 6 November 2026 at the Namazgjah Complex in Tirana.
\nFor developers and startups, the call offers an opportunity to demonstrate their AI agents beyond conventional product presentations, with an emphasis on delivering knowledge to a professional audience.
\nThe organisers have not yet announced the application deadline or direct submission link. Further details about the selection process and participation requirements are expected through the official conference website.
\nInterested companies and development teams can follow announcements at https://allweb.digital/.
\nThe practical test will come in November, when selected AI agents are expected to demonstrate whether they can meet the demands of a live conference audience.
","featured_image_id":378,"intro_image_id":378,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":9,"author_id":10,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":3,"view_count":21,"ordering":0,"published_at":"2026-10-14 08:54:06","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-14 08:54:06","updated_at":"2026-10-14 08:54:06","key_takeaways":"[\"AllWeb Digital Conf has opened a call for AI agents to deliver presentations at its 6 November 2026 conference in Tirana, Albania.\",\"The organisers are looking for agents capable of sharing useful business insights, case studies and interactive presentations rather than conventional product demonstrations.\",\"Technology companies, startups and creative businesses can propose AI agents. The official application deadline and submission link have not yet been announced.\"]","faq":"[{\"q\":\"What is the AllWeb Call for AI Agents?\",\"a\":\"It is an initiative inviting companies and developers to propose AI-powered agents as speakers at AllWeb Digital Conf 2026.\"},{\"q\":\"When and where will AllWeb 2026 take place?\",\"a\":\"The conference will take place on 6 November 2026 at the Namazgjah Complex in Tirana, Albania.\"},{\"q\":\"Who can apply to present an AI agent?\",\"a\":\"The call targets technology companies, startups and creative businesses developing AI agents capable of delivering presentations and interacting with an audience.\"},{\"q\":\"What kind of presentations are organisers looking for?\",\"a\":\"Presentations should provide practical insights and case studies related to marketing, customer experience, productivity, innovation or business growth.\"},{\"q\":\"How can interested teams apply?\",\"a\":\"The application link and deadline have not yet been announced. Interested teams should check allweb.digital for official updates.\"}]","sources":"[{\"title\":\"AllWeb Digital Conf — official conference website\",\"url\":\"https://allweb.digital/\"},{\"title\":\"Albania Tech — AllWeb Opens Its Stage to AI Agents for the First Time.\",\"url\":\"https://albaniatech.org/allweb-opens-its-stage-to-ai-agents-for-the-first-time/\"},{\"title\":\"Business Magazine Albania — AllWeb AI Agents announcemen\",\"url\":\"https://businessmag.al/allweb-hap-per-here-te-pare-skenen-per-ai-agents/\"}]","category_name":"Albania","category_slug":"albania","author_name":"Guest Contributor","author_slug":"guest-contributor","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/b458321579a93f02cc96e6777b3a2e99.webp","featured_image_alt":"AllWeb","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/b458321579a93f02cc96e6777b3a2e99.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"AllWeb Invites AI Agents to Speak at Tirana Conference","meta_description":"AllWeb invites AI agents to speak at its November 2026 conference in Tirana, opening a new opportunity for developers and startups.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":132,"title":"Albania Plans Mandatory AI Training for Public Administration","slug":"albania-plans-mandatory-ai-training-for-public-administration","short_description":"","content":"Albania is preparing a national artificial intelligence programme that will make AI training mandatory for public administration, as the government moves to expand the use of the technology across state institutions.
\nPrime Minister Edi Rama announced the plans during Albania’s Second Innovation Forum on September 8, where government officials, businesses, universities and representatives of the ICT sector discussed AI skills, financing, data and adoption across the economy and public sector.
\nThe government says the national AI programme will be open to individuals, while participation in the training programme will be compulsory for public administration employees.
\nAlbania has also begun integrating AI modules into parts of the administration.
\nAccording to Rama, the first stage is focused on legal directorates, with the government planning to expand the systems across other levels of public administration.
\nThe intention, he said, is to provide civil servants with AI-based assistance as part of their everyday work.
\nThe announcement adds another layer to Albania’s increasingly ambitious use of artificial intelligence in government.
\nRather than limiting AI adoption to isolated pilots, the government is beginning to connect training, administrative tools, research funding and economic policy under a broader national programme.
\nThe details of that programme are still being prepared.
\nEconomy and Innovation Minister Delina Ibrahimaj said the government wants institutions, businesses, universities and the wider innovation ecosystem to contribute to its development.
\nThe ministry is positioning AI as a way to increase productivity, support business development and strengthen the competitiveness of Albania’s economy.
\nNo final document, implementation timetable or detailed governance framework for the programme was announced at the forum.
\nThat means some of the most important questions remain open, including which public-sector roles will be covered by mandatory training, what competencies officials will be expected to develop and how the government will evaluate the use of AI tools inside public institutions.
\nThe decision to begin with legal directorates is particularly notable.
\nAI systems used for drafting, analysing or processing legal and administrative material can improve speed and access to information, but they also create questions around accuracy, accountability, confidential information and human oversight.
\nThe government has not yet provided technical details about the modules being deployed or which AI models and providers are involved.
\nRama said the integration would eventually extend beyond legal departments and across the broader administration.
\nA separate local media report from the forum said Rama also referred to work on an AI-supported public procurement system and growing use of Albania’s e-government services.
\nThe direction is clear: Albania increasingly sees AI not simply as a private-sector technology, but as part of the infrastructure of government.
\nThe government also plans to increase grants through Albania’s Innovation Fund for applied research and partnerships between universities and industry.
\nRama identified cooperation between academia and industry as an area where Albania continues to lag.
\nThe government says around 9,000 students are currently studying technology-related fields, while the country’s technology exports have reached €222 million, according to figures presented at the forum.
\nThose numbers are being used to support a broader argument that Albania can build a larger technology sector alongside its digital transformation of government.
\nBut translating student numbers and public investment into domestic AI capability will require more than training programmes.
\nAccess to computing infrastructure, quality datasets, research capacity, skilled engineers and stronger links between universities and industry will all influence how much of Albania’s AI strategy can eventually be developed locally.
\nAlbania has already gained international attention for its aggressive approach to digital government and the use of AI in public administration.
\nThe new programme suggests that the country is now trying to move from individual projects toward a more coordinated national approach.
\nThat makes the governance side increasingly important.
\nMandatory AI training could help public-sector employees understand the systems they are expected to use, but training alone does not answer questions about transparency, data protection, procurement, responsibility for automated outputs or the circumstances in which AI should not be used.
\nThose issues are likely to become more important as Albania continues its EU accession process and aligns its regulatory framework with European standards.
\nThe next significant step will therefore be the publication of the National Programme for Artificial Intelligence itself.
\nThat document should show whether Albania’s growing ambitions are matched by clear responsibilities, safeguards, investment priorities and measurable targets for how AI will be deployed across government and the wider economy.
","featured_image_id":290,"intro_image_id":290,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":9,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":38,"ordering":0,"published_at":"2026-09-12 10:45:34","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-12 10:45:34","updated_at":"2026-09-12 10:45:34","key_takeaways":null,"faq":null,"sources":null,"category_name":"Albania","category_slug":"albania","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/118a33cea103309332f2fd74ce23e876.webp","featured_image_alt":"Two panelists speak at a “FORUM / INOVACION” AI event before a teal screen.","featured_image_width":1600,"featured_image_height":1067,"intro_image_path":"/api/uploads/2026/09/118a33cea103309332f2fd74ce23e876.webp","intro_image_width":1600,"intro_image_height":1067,"seo_title":"Albania Plans Mandatory AI Training for Public Staff","meta_description":"Albania is preparing a national AI programme, with mandatory AI training for public administration and wider deployment across government.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":120,"title":"Tirana Schools Begin Installing AI-Powered Security Cameras","slug":"tirana-schools-begin-installing-ai-powered-security-cameras","short_description":"","content":"Tirana has started installing AI-powered camera systems in schools as part of Albania’s wider Smart City programme, with the technology expected to be in use when students return for the new school year in September.
\nAlbania’s Interior Ministry confirmed on August 27 that the system has already been installed at Qemal Stafa High School, one of the first schools included in the programme.
\n![\"Tirana-School](\"/api/uploads/2026/08/52793a579c5ac4c8d7748619d3477f89.webp\")
Interior Minister Besfort Lamallari visited the school to inspect the new system, saying artificial intelligence and advanced monitoring technology would become part of the security measures used to protect students and staff.
The new school year in Albania begins on September 14.
\nThe school programme was first detailed earlier this year.
\nAround 380 smart cameras are planned across 20 schools in Tirana, with individual schools expected to receive between 11 and 31 cameras depending on their size and number of students.
\nThe cameras will be installed both inside and outside school buildings and will allow school environments to be monitored in real time.
\nThe programme also includes monitoring rooms, additional detection systems and training for staff responsible for operating the technology.
\nAuthorities say the cameras will work alongside existing security measures rather than replace them.
\nAt Qemal Stafa High School, for example, two State Police officers are already assigned to the school and are expected to remain in place during the coming academic year.
\nThe government has described the system as using artificial intelligence to improve monitoring and help identify potential security risks.
\nEarlier descriptions of the project suggested the system could assist with detecting suspicious individuals and activities around school premises.
\nHowever, detailed public information about the AI functions remains limited.
\nIt is not yet clear exactly what behaviour the software is designed to identify, whether biometric or facial recognition technology is involved, how alerts are generated, or how much human review takes place before action is taken.
\nThose details matter more when surveillance technology is introduced into schools, where most of the people being monitored are children.
\nThe government has presented the programme primarily as a security measure, arguing that better monitoring can create a safer environment for students, teachers and parents.
\nBut the introduction of AI-assisted surveillance in schools also creates questions around privacy and data protection.
\nLegal experts speaking about the project earlier this year called for clear rules governing the technology and consultation with parents.
\nImportant questions include who can access footage, how long recordings are stored, whether AI-generated alerts are kept separately, and what happens when the system incorrectly identifies behaviour as suspicious.
\nAlbania already has a legal framework for personal data protection, but AI-enabled surveillance introduces additional technical issues that conventional camera systems do not necessarily create.
\nThe government has not yet published detailed information addressing all of those questions for the school deployment.
\nThe cameras in schools are only one part of Albania’s broader Smart City initiative.
\nThe government has been developing a national Command and Control Centre intended to bring together surveillance, data analysis and public-security systems.
\nOfficials say the wider programme is designed to help police and other authorities identify risks and respond to incidents more quickly.
\nSchools are now becoming one of the first environments where ordinary citizens will encounter the technology directly.
\nThat makes the Tirana rollout worth watching beyond the question of whether cameras can improve security.
\nThe programme will also show how Albania intends to balance the growing use of AI in public spaces with privacy, transparency and the rights of people being monitored.
\nWith the first systems already installed and schools reopening on September 14, those questions are likely to become more important once the technology moves from testing into everyday use.
","featured_image_id":264,"intro_image_id":264,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":9,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":25,"ordering":0,"published_at":"2026-08-30 11:42:24","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-30 11:42:24","updated_at":"2026-08-30 11:42:24","key_takeaways":null,"faq":null,"sources":null,"category_name":"Albania","category_slug":"albania","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/eb0b3b3553e5dc2c741bdf0270f9f759.webp","featured_image_alt":"An AI-powered security camera hangs from the wall in a school hallway lined with blue lockers.","featured_image_width":1200,"featured_image_height":627,"intro_image_path":"/api/uploads/2026/08/eb0b3b3553e5dc2c741bdf0270f9f759.webp","intro_image_width":1200,"intro_image_height":627,"seo_title":"Tirana Schools Begin Installing AI-Powered Security Cameras","meta_description":"Tirana is introducing AI-powered cameras across schools under Albania’s Smart City programme, raising new questions about security and privacy.","robots":"index","canonical_url":"tirana-schools-ai-powered-cameras-smart-city","og_title":"","og_description":""},{"id":30,"title":"Albania Plans AI-Powered Healthcare by 2026","slug":"albania-plans-ai-powered-healthcare-by-2026","short_description":"","content":"Albania is preparing to introduce artificial intelligence into its national healthcare system in 2026, marking one of the country’s most important steps toward digital transformation in public health. The announcement was made by Prime Minister Edi Rama, who said that the 2026 healthcare budget will prioritize both artificial intelligence and infrastructure investments.
\nThe plan comes as Albania continues to modernize hospitals, emergency centers and medical services across the country. According to the government’s direction, artificial intelligence will not be treated as a separate technology experiment, but as part of a wider reform of the healthcare system. The goal is to improve diagnostics, connect health institutions digitally and raise the quality of care closer to European standards.
\nOne of the main areas where AI is expected to be used is medical imaging. Artificial intelligence can support doctors by analyzing scans, detecting patterns and identifying possible health risks faster than traditional manual processes alone. This does not mean that AI will replace doctors, but it can become a powerful support tool for specialists, especially in cases where fast and accurate interpretation is important.
\nAnother planned area is predictive analysis. With better digital systems, Albania could use health data to identify trends, anticipate pressure on hospitals, improve planning and support preventive medicine. Predictive tools can help health institutions understand where demand may increase, which services need more resources and how patients can be treated earlier.
\nThe government has also linked the AI healthcare plan with interoperability. This means that different parts of the healthcare system should be able to communicate with each other digitally. Hospitals, clinics, laboratories and central health institutions often work with separate systems, which can slow down care and create administrative problems. A national digital ecosystem could make patient information easier to manage and improve coordination between institutions.
\nThe project is expected to be developed through cooperation with a European AI healthcare partner and supported with EU funding. This is important because healthcare AI requires high standards for safety, privacy and reliability. Albania will need to ensure that any AI tools introduced into hospitals are tested, supervised and compatible with European health and data protection principles.
\nThe announcement also fits into Albania’s broader goal of improving healthcare infrastructure by 2030. Rama has mentioned hospital modernization projects, including the new Korçë Hospital, the Durrës Emergency Center and municipal hospitals in Mati and Tropojë. These physical investments are now being connected with digital modernization, showing that the government wants to combine better buildings, better equipment and smarter technology.
\nFor citizens, the promise of AI in healthcare is faster service, better diagnostics and more efficient public health management. For doctors and medical staff, AI could reduce pressure by helping with repetitive tasks, data analysis and decision support. For the state, it could improve planning and reduce inefficiencies in the healthcare system.
\nHowever, the reform will also bring serious challenges. Health data is highly sensitive, and any AI system used in medicine must be transparent, secure and carefully monitored. Albania will need strong rules on who can access patient data, how algorithms are tested and what happens if an AI-supported decision is wrong.
\nThe success of the project will depend not only on technology, but also on training. Doctors, nurses, administrators and IT professionals will need to understand how these systems work and how to use them responsibly. Without proper training, even advanced technology can fail to deliver real improvements.
\nAlbania’s plan to introduce artificial intelligence into healthcare shows that AI is moving from political discussion into practical public services. If implemented carefully, it could help modernize the country’s health system and bring it closer to European digital standards. The real test will be whether the new tools improve everyday care for patients, not only whether they sound impressive on paper.
","featured_image_id":66,"intro_image_id":66,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":9,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":1,"show_table_of_contents":1,"read_time_minutes":4,"view_count":15,"ordering":0,"published_at":"2026-06-25 11:45:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 11:45:54","updated_at":"2026-06-25 18:34:46","key_takeaways":null,"faq":null,"sources":null,"category_name":"Albania","category_slug":"albania","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/6ae9f253a90dde2799b3dcbeb6b55836.webp","featured_image_alt":"A scientist in a white coat examines medical data and a brain scan on a large digital display.","featured_image_width":1600,"featured_image_height":1066,"intro_image_path":"/api/uploads/2026/06/6ae9f253a90dde2799b3dcbeb6b55836.webp","intro_image_width":1600,"intro_image_height":1066,"seo_title":"Albania Plans AI-Powered Healthcare by 2026 | AI Balkans","meta_description":"Albania is preparing to introduce artificial intelligence into its national healthcare system in 2026, marking one of the country’s most important…","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":29,"title":"Albania Signs Europe’s AI and Human Rights Convention","slug":"albania-signs-europe-s-ai-and-human-rights-convention","short_description":"","content":"Albania has signed the Council of Europe’s Framework Convention on Artificial Intelligence, Human Rights, Democracy and the Rule of Law, joining a growing group of countries that are committing to regulate AI through a human-rights-based approach. The signing took place on 15 June 2026 and was announced by the Council of Europe shortly after.
\nThis is an important step for Albania because the convention is the first international legally binding treaty focused specifically on artificial intelligence. Its purpose is to make sure that AI systems are developed and used in ways that respect human rights, democratic principles and the rule of law.
\nFor Albania, the signing comes at a time when artificial intelligence is becoming a more visible part of public policy. The country has already attracted international attention for experimenting with AI in government services and public procurement. Now, by joining the Council of Europe framework, Albania is also signaling that AI development must be connected to legal responsibility and institutional safeguards.
\nThe convention does not focus only on technical standards. It focuses on the impact AI can have on people and society. This includes risks connected to discrimination, privacy, transparency, access to justice, democratic processes and the protection of citizens from unfair automated decisions.
\nBy signing the convention, Albania is aligning itself with a wider European approach to artificial intelligence. This is especially relevant because Albania is also pursuing closer integration with the European Union. As AI regulation becomes part of the European legal and political agenda, countries that want to move closer to European standards will need to show that they can manage both innovation and risk.
\nThe convention also matters for public trust. Artificial intelligence can bring faster services and better decision-making, but citizens are often concerned about how AI systems use personal data and whether automated tools can be biased or unfair. A human-rights framework can help create clearer expectations about how AI should be used by governments and companies.
\nFor public institutions, the signing means that Albania will need to think seriously about oversight. AI systems used in public administration, policing, courts, healthcare, education or social services cannot be treated like ordinary software. They may affect people’s rights and opportunities, so they require transparency, accountability and human control.
\nThe agreement also strengthens the idea that artificial intelligence should not be used to weaken democracy. AI tools can influence public debate, generate misleading content, manipulate information or affect electoral processes. The Council of Europe framework recognizes these risks and places democratic integrity at the center of AI governance.
\nFor businesses and technology developers in Albania, the convention could create a clearer direction for future regulation. Companies that build AI systems will increasingly need to consider not only performance and profit, but also fairness, data protection and explainability. This may create new obligations, but it can also help build trust in Albanian AI products and services.
\nThe signing does not automatically solve Albania’s AI governance challenges. Implementation will be the key issue. Albania will need laws, institutions, technical capacity and independent oversight mechanisms to make the principles of the convention meaningful in practice.
\nStill, the move is significant. It shows that Albania wants to participate in international AI governance rather than remain outside the process. As artificial intelligence becomes more powerful, countries will be judged not only by how quickly they adopt it, but also by how responsibly they manage it.
\nAlbania’s signature on the Council of Europe AI Convention places the country within a broader European effort to ensure that technology serves people, instead of putting citizens’ rights at risk. The next stage will be translating that commitment into national rules, public-sector practices and real protections for citizens.
","featured_image_id":65,"intro_image_id":65,"featured_image_credit":"COE","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":9,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":16,"ordering":0,"published_at":"2026-06-25 11:43:26","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 11:43:26","updated_at":"2026-06-25 11:43:26","key_takeaways":null,"faq":null,"sources":null,"category_name":"Albania","category_slug":"albania","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/91813223e0c9898789ab84f158056b84.webp","featured_image_alt":"Two officials sign documents beside Albanian and European Union flags.","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/06/91813223e0c9898789ab84f158056b84.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Albania Signs Europe’s AI and Human Rights Convention","meta_description":"Albania has signed the Council of Europe’s Framework Convention on Artificial Intelligence, Human Rights, Democracy and the Rule of Law, joining a…","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":28,"title":"Albania Prepares First AI Law With AKSHI at the Center","slug":"albania-prepares-first-ai-law-with-akshi-at-the-center","short_description":"","content":"Albania is preparing its first dedicated law on artificial intelligence, a major step toward creating a national legal framework for the development, use and supervision of AI systems. The draft law has been issued for public consultation and aims to align Albania’s approach with the European Union’s AI Act.
\nThe draft gives a central role to the National Agency for Information Society, known as AKSHI, which is expected to serve as the coordinating authority for artificial intelligence in the country. This means AKSHI would have responsibilities related to implementation, monitoring and supervision of AI systems and general-purpose AI models.
\nThe law is designed to regulate how AI systems are developed, placed on the market and used in Albania. It also applies to some foreign providers if their AI systems produce results used inside Albania. This broad scope reflects the reality of modern technology, where AI services can be built in one country but used by citizens, companies or institutions in another.
\nOne of the most important parts of the draft is the classification of AI systems based on risk. Low-risk systems would face lighter requirements, while high-risk systems would need to follow stricter rules. These rules include risk management, technical documentation, transparency, logging, post-market monitoring and human oversight.
\nHigh-risk AI systems are especially important because they can affect people’s rights, safety or access to important services. Such systems may be used in areas like employment, education, public administration, law enforcement, healthcare or access to financial services. The draft law recognizes that when AI has real consequences for people, stronger safeguards are necessary.
\nThe draft also lists practices that would be prohibited. These include AI systems that use manipulative or deceptive techniques to influence people’s behavior, systems that exploit people’s weaknesses because of age, disability or economic status, and systems that create social scoring of individuals.
\nAnother sensitive area is biometric identification. The draft places restrictions on real-time biometric identification systems in public spaces. For law enforcement purposes, such systems would be allowed only in specific cases, such as searching for victims of kidnapping or trafficking, preventing serious threats to life and security, or identifying suspects of serious crimes. Prior court authorization would be required.
\nThe draft also addresses facial recognition databases. It prohibits systems that create such databases through the indiscriminate collection of images from the internet or security cameras. This is an important protection because facial recognition technology can create serious privacy and surveillance risks if used without limits.
\nAnother area covered by the draft is synthetic content and deepfakes. As AI-generated images, videos, voices and texts become more realistic, governments are under pressure to create transparency rules. Albania’s draft law begins to define these concepts and creates a basis for regulating AI-generated content.
\nAt the same time, the draft is not only restrictive. It also includes measures to support innovation. One example is the creation of regulatory AI sandboxes. These are controlled environments where companies, researchers and startups can develop, train and test AI systems under supervision before bringing them fully to market.
\nThis could be especially useful for startups and small businesses. AI regulation can be complex and expensive, and smaller companies often struggle with compliance. The draft includes simplified forms of technical documentation for small and medium-sized enterprises, which could help innovation continue while still maintaining basic safeguards.
\nThe law is also important for Albania’s European integration path. By aligning with the EU AI Act, Albania is preparing its legal system for future European standards. This could help Albanian institutions, companies and technology providers cooperate more easily with European partners.
\nHowever, implementation will be difficult. Albania will need trained regulators, technical experts, clear procedures and strong cooperation between institutions. AI supervision requires both legal knowledge and technical capacity, because authorities must understand how systems work, how risks are measured and how violations can be detected.
\nThe draft law shows that Albania is moving from experimentation toward formal AI governance. After gaining global attention for its AI-based public-sector initiatives, the country is now trying to build rules that can manage the risks of artificial intelligence while still encouraging innovation.
\nIf passed and implemented effectively, the law could become the foundation of Albania’s AI policy for the coming years. It would give citizens stronger protections, give companies clearer rules and bring Albania closer to the European model of trustworthy, human-centered artificial intelligence.
","featured_image_id":64,"intro_image_id":64,"featured_image_credit":"Gettyimages","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":9,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":24,"ordering":0,"published_at":"2026-06-25 11:37:59","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 11:37:59","updated_at":"2026-06-25 11:37:59","key_takeaways":null,"faq":null,"sources":null,"category_name":"Albania","category_slug":"albania","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/94d75ca42ef057fa0c1eb1fb0458cbc5.webp","featured_image_alt":"Edi Rama talking, with albanian flag in the background","featured_image_width":1024,"featured_image_height":682,"intro_image_path":"/api/uploads/2026/06/94d75ca42ef057fa0c1eb1fb0458cbc5.webp","intro_image_width":1024,"intro_image_height":682,"seo_title":"Albania Prepares First AI Law With AKSHI at the Center","meta_description":"Albania is preparing its first dedicated law on artificial intelligence, a major step toward creating a national legal framework for the development,…","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":21,"title":"Albania at a Crossroads: AI Promises, Political Turbulence, and Growing Public Pressure","slug":"albania-at-a-crossroads-ai-promises-political-turbulence-and-growing-public-pressure","short_description":"","content":"TIRANA — Albania is once again in the regional spotlight, and this time the country is being pulled in two very different directions: bold digital promises from the government and mounting political pressure in the streets. Prime Minister Edi Rama’s administration continues to present artificial intelligence as a tool for modernization and anti-corruption reform, but the public mood has become more tense as protests, corruption allegations, and institutional distrust deepen.
\nThe most unusual symbol of Albania’s tech ambitions is Diella, the AI-powered figure promoted as part of the government’s effort to make public procurement more transparent. Officials have framed the move as a leap toward cleaner administration and a more efficient state, but critics argue that the real problem is not the lack of software — it is the lack of trust in the system itself. In the Balkans, where politics and state institutions are often judged by results rather than slogans, that distinction matters.
\nAt the same time, Albania has been shaken by renewed unrest. In Tirana, protesters have taken to the streets demanding accountability over corruption scandals and challenging Rama’s government more openly than before. The latest tensions show that the political atmosphere remains fragile, with opposition pressure growing and public frustration spilling into repeated demonstrations.
\nFor Albania, the timing is sensitive. The government wants to project stability, reform, and EU-readiness, but the country is also dealing with a more skeptical public and a louder opposition. That makes the AI narrative both attractive and risky: it offers a modern image for the outside world, but at home it is being tested against the old Balkan realities of power, corruption, and street politics.
","featured_image_id":56,"intro_image_id":56,"featured_image_credit":"Edi Rama","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":9,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":2,"view_count":25,"ordering":0,"published_at":"2026-06-25 10:36:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 10:36:32","updated_at":"2026-06-25 10:51:07","key_takeaways":null,"faq":null,"sources":null,"category_name":"Albania","category_slug":"albania","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/4e29da25e688cdcbeb7bb067b7be1092.webp","featured_image_alt":"Edi Rama talking in front of EU Desk, with Albanian Flag in the background","featured_image_width":1024,"featured_image_height":576,"intro_image_path":"/api/uploads/2026/06/4e29da25e688cdcbeb7bb067b7be1092.webp","intro_image_width":1024,"intro_image_height":576,"seo_title":"Albania at a Crossroads: AI Promises, Political Turbulence, and Growing Public Pressure","meta_description":"TIRANA — Albania is once again in the regional spotlight, and this time the country is being pulled in two very different directions: bold digital…","robots":"index","canonical_url":"","og_title":"","og_description":""}]},{"category":{"id":12,"name":"Bosnia & Herzegovina","slug":"bosnia-herzegovina"},"posts":[{"id":126,"title":"Federation of Bosnia and Herzegovina Backs AI Use in University Teaching","slug":"federation-of-bosnia-and-herzegovina-backs-ai-use-in-university-teaching","short_description":"","content":"Universities in the Federation of Bosnia and Herzegovina are receiving government support to introduce artificial intelligence tools into teaching and improve the digital skills of academic staff.
\nThe Federal Ministry of Education and Science published the final results of its latest higher-education support programme on September 4, including a dedicated funding category for the use of AI tools in teaching and the development of lecturers’ digital skills.
\nThe programme is part of a wider funding initiative aimed at the development of higher education in the Federation of Bosnia and Herzegovina.
\nOther areas supported through the call include additional pedagogical training for university assistants, closer links between study programmes and labour-market needs, and international summer-school programmes.
\nThe AI component, however, reflects a wider change already under way across universities in the region: institutions are moving beyond debates about whether students should use generative AI and beginning to consider how the technology should be integrated into teaching itself.
\nWhen the funding call was announced in June, the ministry explicitly listed support for universities applying AI tools in teaching alongside improving lecturers’ digital capabilities. Applications closed in July, with the final results published this month.
\nThat matters because much of the early response to generative AI in higher education has focused on students.
\nUniversities have spent the past few years dealing with questions around AI-assisted assignments, plagiarism, examinations and academic integrity. Less attention has been given to the skills lecturers themselves need if AI becomes part of everyday teaching.
\nThose skills go beyond learning how to use a chatbot.
\nAcademic staff need to understand where AI-generated information can be unreliable, what types of data should not be entered into external systems, how AI affects assessment design and when students should be expected to disclose its use.
\nThey also need to decide where AI genuinely contributes to learning and where it simply makes a task faster without improving understanding.
\nBy including both tools and teacher skills in the same programme, the ministry is effectively treating the two issues as connected.
\nThe initiative comes as Bosnia and Herzegovina slowly develops a broader policy conversation around artificial intelligence.
\nUntil recently, much of the country's AI activity came from universities, technology companies, researchers and individual projects rather than coordinated public policy.
\nThat is beginning to change.
\nEducation is one of the areas where governments can influence AI adoption relatively quickly because public institutions already have established funding programmes, training structures and relationships with universities.
\nThe question is what universities do with that support.
\nIntroducing AI tools without clear academic rules could create new problems around assessment, privacy and unequal access. Training lecturers without giving them space to redesign courses would have limited effect as well.
\nThe strongest university approaches emerging elsewhere in Europe increasingly combine access to technology with clear rules about responsibility, transparency and academic judgement.
\nFor universities in the Federation, the funding creates an opportunity to develop those practices before AI use becomes even more deeply embedded in student work.
\nThat could mean redesigning assignments, training staff to evaluate AI-generated content, creating guidance on appropriate use and experimenting with AI-assisted teaching in areas where it can provide measurable benefits.
\nIt could also provide lessons for other higher-education institutions across Bosnia and Herzegovina, where education policy is divided across different levels of government.
\nThe September results do not settle how individual universities will ultimately use AI.
\nThey do, however, mark an important shift.
\nAI in higher education is beginning to move from something students adopt independently to something public institutions are prepared to fund, organise and teach around.
\nThe next question is what the selected universities build with that support, and whether the projects produce practices that can be used more widely across the country.
","featured_image_id":277,"intro_image_id":277,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":12,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":25,"ordering":0,"published_at":"2026-09-06 12:37:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-06 12:37:20","updated_at":"2026-09-06 12:37:23","key_takeaways":null,"faq":null,"sources":null,"category_name":"Bosnia & Herzegovina","category_slug":"bosnia-herzegovina","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/c1c22777d3f58fd373096a97bda99a23.webp","featured_image_alt":"A humanoid robot with glowing blue lights browses books in a warmly lit library.","featured_image_width":1343,"featured_image_height":753,"intro_image_path":"/api/uploads/2026/09/c1c22777d3f58fd373096a97bda99a23.webp","intro_image_width":1343,"intro_image_height":753,"seo_title":"Bosnia Federation Backs AI Use in University Teaching","meta_description":"Universities in the Federation of Bosnia and Herzegovina are receiving support to use AI tools in teaching and improve staff digital skills.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":110,"title":"Bosnia’s Parliament Adopts Roadmap for Responsible AI Use","slug":"bosnia-s-parliament-adopts-roadmap-for-responsible-ai-use","short_description":"","content":"Bosnia and Herzegovina’s Parliamentary Assembly has adopted a roadmap for using artificial intelligence in its daily operations, setting out rules, staff training and a first group of practical AI projects inside the country’s national legislature.
\nThe roadmap makes the Parliamentary Assembly the first legislative institution in Bosnia and Herzegovina to establish a strategic framework for integrating AI into its work, according to the United Nations Development Programme, which has supported the initiative.
\nRather than introducing a single AI system, the plan lays out how the institution intends to manage the technology as it becomes part of routine parliamentary work.
\nThe first planned pilots include AI-assisted transcription of parliamentary sessions, media analysis, language and editorial support, and AI assistants intended to help staff in the Parliamentary Assembly’s Secretariat. The projects are expected to operate with safeguards covering data protection, human oversight and responsible use.
\nThe roadmap is built around three priorities: developing AI skills among staff, creating internal policies and governance rules, and gradually introducing AI tools into selected parliamentary processes.
\nThat institutional approach follows signs that AI was already entering day-to-day work informally.
\nAn AI readiness assessment carried out with UNDP found that staff were already using AI tools, but also identified a lack of common standards, internal policies and institutional capacity for managing their use. The assessment used the Inter-Parliamentary Union’s AI Maturity Framework for Parliaments and included surveys, interviews and practical workshops.
\nMore than 40 employees of the Parliamentary Assembly’s Secretariat later took part in a two-day training programme covering practical use of generative AI as well as the legal, ethical and security issues that come with introducing the technology into public institutions.
\nThe training was not the institution’s first step.
\nIn May, members of parliament, delegates and parliamentary staff took part in a separate programme titled “Artificial Intelligence in Practice at the Parliamentary Assembly of Bosnia and Herzegovina,” organised by the Secretariat with UNDP support. The programme covered AI risks and opportunities, regulation and practical applications in parliamentary work.
\nThe move comes as public institutions across the Western Balkans face increasing pressure to decide how generative AI should be used internally before wider regulation is fully established.
\nFor Bosnia and Herzegovina, the issue also intersects with the country’s European integration process. During the May training programme, UNDP representatives pointed to the Parliamentary Assembly’s future role in adopting legislation governing AI as Bosnia and Herzegovina aligns more closely with European rules.
\nThat gives the roadmap a significance beyond administrative efficiency.
\nAI-assisted transcription or editorial work may appear relatively limited compared with higher-risk uses of artificial intelligence in areas such as policing, healthcare or social services. But introducing formal governance at the parliamentary level establishes procedures for how public employees use AI, how outputs are checked and where human responsibility remains.
\nThe Parliamentary Assembly is not giving AI systems authority to make parliamentary decisions. The current focus is on supporting staff and improving administrative processes, with human oversight remaining part of the framework.
\nThat distinction is important in a region where many public institutions are experimenting with commercial AI tools faster than internal rules are being written.
\nThe next stage will be implementation.
\nUNDP says work will now move toward developing the first pilot projects, applying the roadmap’s priorities and further strengthening the institution’s capacity to manage AI use.
\nFor Bosnia and Herzegovina, the real test will come when those pilots leave the planning stage and become part of everyday parliamentary work, particularly in areas where sensitive institutional data, accountability and the accuracy of AI-generated material become practical rather than theoretical concerns.
","featured_image_id":249,"intro_image_id":249,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":12,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":38,"ordering":0,"published_at":"2026-08-19 13:12:30","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-19 13:12:30","updated_at":"2026-08-19 13:12:30","key_takeaways":null,"faq":null,"sources":null,"category_name":"Bosnia & Herzegovina","category_slug":"bosnia-herzegovina","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/c0902cc9115bcdb03b38ed53a59c77f8.webp","featured_image_alt":"Lawmakers attend a UNDP session on artificial intelligence at Bosnia and Herzegovina’s Parliament.","featured_image_width":1600,"featured_image_height":800,"intro_image_path":"/api/uploads/2026/08/c0902cc9115bcdb03b38ed53a59c77f8.webp","intro_image_width":1600,"intro_image_height":800,"seo_title":"Bosnia’s Parliament Adopts Roadmap for Responsible AI Use","meta_description":"Bosnia and Herzegovina’s parliament has adopted an AI roadmap covering staff training, governance and pilots for transcription and administrative work.","robots":"index","canonical_url":"bosnia-parliament-adopts-ai-roadmap","og_title":"","og_description":""},{"id":80,"title":"Sarajevo AI Summit Opens Research Stage for 2026","slug":"sarajevo-ai-summit-opens-research-stage-for-2026","short_description":"","content":"Sarajevo’s largest artificial intelligence gathering is expanding its focus beyond technology companies and conference speakers, inviting students and researchers to present their own work at this year’s Kiss the Future AI Summit.
\nThe summit, scheduled for October 15 and 16, has opened submissions for a new Research Stage covering artificial intelligence, intelligent systems, energy and sustainability. Applications remain open until July 31.
\nStudents, early-career researchers and research teams can submit completed studies, work-in-progress ideas or research posters. Selected participants will present their work during the summit, while one emerging researcher will receive mentorship through Sarajevo’s Blum Institute.
\nThe addition matters because many technology events across the region still focus heavily on global trends, corporate products and well-known speakers. Giving local researchers space to present their work could help connect universities, young engineers and companies that often operate separately.
\nThe summit organisers say they expect more than 2,000 participants, over 70 speakers and representatives from more than 30 companies across several thematic stages. The first announced speakers include professionals from GitHub, Databricks, Red Hat, Supabase, ElevenLabs and other international technology organisations.
\nBut the more important question for Bosnia and Herzegovina is what remains after the stages are dismantled.
\nThe country has a growing technology sector, strong engineering talent and an increasingly visible AI community. What it still lacks is a mature national ecosystem capable of consistently turning research into products, public-sector tools and companies that can grow without immediately moving their work abroad.
\nThe Research Stage appears designed to address part of that gap.
\nSubmissions can take the form of oral presentations, research posters or “tiny papers”—short papers of up to 500 words for early-stage ideas and preliminary findings. Research topics may include generative AI, large language models, ethical AI governance, robotics, autonomous systems, smart grids, renewable-energy forecasting and climate technology.
\nThat range also reflects a wider shift in the AI conversation. The focus is gradually moving away from chatbots alone and towards how artificial intelligence might be used in energy, industry, transport and other sectors where Bosnia already has technical experience.
\nThis is particularly relevant for a country with a long engineering tradition but limited investment in modern research infrastructure.
\nBlum Institute, which organises the summit together with technology company Bloomteq, describes itself as Bosnia and Herzegovina’s first institute dedicated to artificial intelligence research and development. It works with universities including the University of Sarajevo’s Faculty of Electrical Engineering, the Faculty of Traffic and Communications and the International University of Sarajevo.
\nThose partnerships could help establish a stronger bridge between academic work and commercial application. However, the scale of that impact will depend on whether collaboration continues throughout the year, rather than remaining tied to a two-day event.
\nResearchers selected through the open call will be notified in August, with final presentation materials due in September. Their work will be presented in Sarajevo on October 16.
\nFor young researchers in Bosnia and the wider Balkans, the opportunity is not only to appear on a conference programme. It is also a chance to place locally developed ideas in front of companies, institutions and researchers who might otherwise never see them.
\nWhether the summit can turn those connections into lasting projects will be the more important measure of success.
","featured_image_id":199,"intro_image_id":199,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":12,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":22,"ordering":0,"published_at":"2026-07-15 08:08:35","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-15 08:08:35","updated_at":"2026-07-15 08:08:35","key_takeaways":null,"faq":null,"sources":null,"category_name":"Bosnia & Herzegovina","category_slug":"bosnia-herzegovina","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/c8576ead9ce87ce825a4c1ff649ac14e.webp","featured_image_alt":"Kiss the Future AI Summit Sarajevo","featured_image_width":1279,"featured_image_height":720,"intro_image_path":"/api/uploads/2026/07/c8576ead9ce87ce825a4c1ff649ac14e.webp","intro_image_width":1279,"intro_image_height":720,"seo_title":"Sarajevo AI Summit Opens Research Stage for 2026","meta_description":"Sarajevo’s Kiss the Future AI Summit invites students and researchers to present AI, energy and intelligent systems work by July 31.","robots":"index","canonical_url":"","og_title":"","og_description":""}]},{"category":{"id":11,"name":"Serbia","slug":"serbia"},"posts":[{"id":143,"title":"Serbian Teachers Test AI Assistant Built for Classrooms","slug":"serbian-teachers-test-ai-assistant-built-for-classrooms","short_description":"","content":"Serbian education nonprofit Petlja Foundation is testing an artificial intelligence assistant designed specifically for teachers, as schools experiment with ways to use AI for lesson preparation, classroom materials and student projects.
\nThe Petlja AI Assistant is built on Microsoft Azure and Azure OpenAI Service and is currently available to around 250 teachers in Serbia. Between 50 and 60 use it actively, according to Petlja director Milan Simić.
\nRather than giving students a general-purpose chatbot, the project focuses primarily on supporting teachers.
\nOne of those teachers is Sanja Rajić, who teaches mathematics and programming at the Palanka Gymnasium in Bačka Palanka. She began testing an early version of the assistant in 2023.
\nRajić uses the tool to prepare lesson structures, generate exercises for different levels of knowledge and create variations of tests. Formal lesson preparation that could previously take hours can be produced as a starting point by the assistant and then reviewed and adapted by the teacher.
\n
That distinction is central to how the system is being used. Teachers remain responsible for checking the generated material and deciding what is appropriate for their students.
\nThe tool has also found its way into practical student projects.
\nStudents at the Palanka Gymnasium have used the assistant while working on robotics, programming and computer vision projects. In one case, students building a mechanical finger used it to help troubleshoot a 3D-printed component when they could not find the required screw locally.
\nTwo students from the school also worked on BrAInUp, an educational well-being application created for Serbia's 2026 Student Companies Competition, with Rajić and the AI assistant supporting parts of the technical development.
\nPetlja was founded in 2017, the same year Serbia introduced programming as a mandatory subject from fifth grade.
\nThe foundation initially focused on creating programming and computer science materials that students could use directly through a web browser. Its platform has since expanded into teacher training, digital education and AI-supported learning.
\nPetlja reports more than 100,000 active users over the past three months, with more than 18,000 course participants during the current school year and more than 50 courses available in Serbian.
\nIts work with Serbia's education system includes teacher training, materials for the Digital World subject, resources for specialised IT classes and development of AI tools for education.
\nThese activities are carried out through projects involving Serbia's Ministry of Education, the United Nations Development Programme and other partners.
\nMicrosoft has supported Petlja through technology, expertise and financial assistance, while Azure provides the infrastructure used by the current AI Assistant.
\nRunning an AI assistant for a large number of teachers creates another problem: cost.
\nEvery interaction with a commercial AI model carries an operating expense. Petlja is therefore working through the UNICEF TLM4EL project on an open-source model adapted for Serbian education.
\nThe plan is to eventually run the model on donated hardware rather than relying entirely on commercial cloud AI services.
\nPetlja says this could lower operating costs and allow more of the system's data to remain under local control. There is not yet a confirmed timeline or funding arrangement for a large-scale rollout.
\nThat part of the project could prove more significant than the assistant itself.
\nFor education systems across the Western Balkans, introducing generative AI is only one part of the problem. Schools also have to decide who controls the infrastructure, where educational data is processed, how tools work in local languages and who is responsible for checking their output.
\nPetlja's experiment provides one model: start with teachers, keep human review in the process and build the technology around the education system rather than introducing a general AI tool and expecting schools to adapt to it.
\nThe next test will be whether that model can move beyond several dozen active users and operate sustainably across a much larger part of Serbia's school system.
","featured_image_id":314,"intro_image_id":314,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":11,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":149,"ordering":0,"published_at":"2026-09-26 09:41:32","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-26 09:41:32","updated_at":"2026-09-26 09:41:32","key_takeaways":"[\"Around 250 Serbian teachers currently have access to Petlja's AI Assistant, with 50 to 60 using it actively.\",\"Teachers are using the system for lesson preparation, exercises, tests and student technology projects, while reviewing and adapting its output.\",\"Petlja is also working on an open-source model for Serbian education that could eventually run locally and reduce dependence on commercial AI services.\"]","faq":"[{\"q\":\"What is the Petlja AI Assistant?\",\"a\":\"It is an AI tool developed by Serbia's Petlja Foundation to support teachers with lesson planning, educational materials and other classroom tasks.\"},{\"q\":\"How many teachers are using it?\",\"a\":\"Around 250 teachers currently have access. Petlja says approximately 50 to 60 are active users.\"},{\"q\":\"Does the AI replace teachers?\",\"a\":\"No. The current model is designed as a support tool. Teachers are expected to review, modify and approve generated materials before using them in class.\"},{\"q\":\"Is Petlja developing its own AI model?\",\"a\":\"Petlja is working through the UNICEF TLM4EL project on adapting an open-source model for Serbian education, with plans for local deployment on donated hardware.\"}]","sources":"[{\"title\":\"Microsoft published the detailed Petlja case study on 24 September 2026\",\"url\":\"https://news.microsoft.com/source/emea/features/petlja-microsoft-ai-serbian-classrooms/\"}]","category_name":"Serbia","category_slug":"serbia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/da3c6e043da6db9d73c50cda0c5179c2.webp","featured_image_alt":"A teacher speaks to students beside a laptop and document camera in a bright classroom technology lab.","featured_image_width":1600,"featured_image_height":1069,"intro_image_path":"/api/uploads/2026/09/da3c6e043da6db9d73c50cda0c5179c2.webp","intro_image_width":1600,"intro_image_height":1069,"seo_title":"Serbian Teachers Test AI Assistant Built for Classrooms","meta_description":"Serbia's Petlja is testing an AI assistant for teachers and developing a locally hosted model designed for the country's education system.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":118,"title":"Serbia Starts Humanoid Robot Production in Šabac","slug":"serbia-starts-humanoid-robot-production-in-abac","short_description":"","content":"Serbia has opened a new production line for humanoid robots in Šabac, bringing one of the country's most ambitious robotics projects into operation.
\nThe facility was launched on August 29 by Minth Group in partnership with Chinese robotics company AGIBOT. Serbia's Development Agency described it as the first serial production operation of humanoid robots in Europe.
\nThe first phase of the project represents an investment of €20 million and is located at Minth's existing production site in Šabac.
\nSerbian President Aleksandar Vučić attended the opening together with representatives of Minth Group, AGIBOT, the Serbian government and local authorities. The first robot produced at the facility was presented during the launch.
\nThe current operation is primarily an assembly line for humanoid robots.
\n![\"\"](\"/api/uploads/2026/08/75e886209a0005ec39ffc2b76be3b6cd.webp\")
Serbian officials say the initial target is to assemble more than 5,000 robots annually at the Šabac facility, while the companies behind the project plan to expand production significantly over time.
\nMinth has also indicated that its Serbian operations could eventually include other advanced technologies, including unmanned aerial vehicles and battery systems.
\nThe larger plan centres on a proposed Robotics Industrial Park in Inđija.
\nThe project is expected to involve around €200 million in total investment, with future production capacity of up to 20,000 humanoid robots and robotic dogs per year.
\nThose numbers remain planned production targets rather than current output.
\nAGIBOT has become one of China's prominent companies in the rapidly developing humanoid robotics market.
\nThe company develops robots designed for industrial, commercial and service environments, combining artificial intelligence with computer vision, sensors and autonomous movement.
\nThe Serbian project gives AGIBOT a manufacturing presence in Europe while allowing Minth to expand beyond its traditional automotive manufacturing business.
\nFor Serbia, the partnership is significant because it brings physical AI manufacturing into an industrial base that has traditionally been associated with automotive components and other conventional manufacturing.
\nMost AI development across the Balkans has so far been concentrated in software, IT services, startups, education and research.
\nHumanoid robotics requires something different.
\nBuilding these systems involves software and AI models, but also batteries, sensors, motors, electronics, precision components and complex manufacturing infrastructure.
\nThat makes the Šabac project particularly interesting for the region.
\n![\"serbia-shabac1.webp\"](\"/api/uploads/2026/08/49c3175930a1d13a8cd11842fed1f90b.webp\")
If production expands beyond assembly and the planned robotics park in Inđija is completed, Serbia could become part of a wider European supply chain for humanoid robotics at a time when the industry is still taking shape.
There are still questions about how quickly the operation can scale and how much of the manufacturing process will eventually take place locally.
\nFor now, however, the project has moved beyond announcements.
\nThe production line is open, the first robot has been assembled in Šabac, and Minth and AGIBOT are preparing for a much larger robotics operation in Serbia.
","featured_image_id":260,"intro_image_id":260,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":11,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":25,"ordering":0,"published_at":"2026-08-30 11:33:50","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-30 11:33:50","updated_at":"2026-08-30 11:33:50","key_takeaways":null,"faq":null,"sources":null,"category_name":"Serbia","category_slug":"serbia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/bf8f26222c2e62b68acff733a18e8b34.webp","featured_image_alt":"Two men and a humanoid robot stand between oversized Serbian folk mascots before a “Serbia” backdrop.","featured_image_width":1024,"featured_image_height":683,"intro_image_path":"/api/uploads/2026/08/bf8f26222c2e62b68acff733a18e8b34.webp","intro_image_width":1024,"intro_image_height":683,"seo_title":"Serbia Starts Humanoid Robot Production in Šabac","meta_description":"Serbia has opened a humanoid robot production line in Šabac through a partnership between Minth Group and China's AGIBOT.","robots":"index","canonical_url":"serbia-humanoid-robot-production-sabac-agibot","og_title":"","og_description":""},{"id":84,"title":"Serbia Joins China-Led Global AI Organization","slug":"serbia-joins-china-led-global-ai-organization","short_description":"","content":"Serbia has joined 28 other countries in establishing a new intergovernmental organization for artificial intelligence, placing Belgrade among the founding members of a China-backed initiative that wants a larger role in global AI governance.
\nThe agreement establishing the World Artificial Intelligence Cooperation Organization, known as WAICO, was signed in Shanghai on July 16. The organization will be headquartered in the Chinese city and will operate as an independent international body focused on cooperation around the development and governance of AI.
\nChina’s Foreign Minister Wang Yi signed the agreement on behalf of Beijing. Other founding members include Russia, Belarus, Brazil, Cuba, Venezuela and Pakistan, alongside countries from Africa and Asia.
\n![\"Chinese](\"/api/uploads/2026/07/df814665fc8b718d3bbfc81194b37a8e.webp\")
\nSerbia is the only Western Balkan country reported among the 29 founding members.
\nThe Serbian government had approved the country’s participation in June, when it adopted the basis for negotiating and signing the founding agreement. At the time, the government said membership would support scientific and technological cooperation, knowledge exchange and joint research and innovation projects.
\nThe formal signing now places Serbia inside an organization that China is presenting as a new platform for countries seeking a greater voice in how artificial intelligence is developed and regulated.
\nAccording to the official Chinese government announcement, WAICO will promote cooperation on AI innovation, safety and governance while working to make the technology more accessible to developing countries.
\nThe organization also intends to address the uneven distribution of computing infrastructure, data, technical knowledge and other resources that have left much of the Global South dependent on a small number of technology companies and countries.
\nMembership will remain open to other states.
\nThe initiative was announced around the opening of the World Artificial Intelligence Conference in Shanghai, where Chinese President Xi Jinping called for a global system of AI governance based on wider participation and greater access to the technology.
\nChina’s message is that AI rules should not be determined only by a small group of wealthy countries or private companies. At the same time, the creation of a new organization headquartered in Shanghai gives Beijing an opportunity to shape international AI policy around its own strategic interests.
\nFor Serbia, joining WAICO adds another layer to an already complex foreign and technology policy.
\nThe country is a candidate for European Union membership and is expected to bring its legislation closer to the EU AI Act, which establishes risk-based rules for artificial intelligence. Serbia has also developed close technological and economic relationships with China, including cooperation in digital infrastructure, surveillance technology, innovation and scientific research.
\nThose relationships are not necessarily incompatible. WAICO has not been presented as a regulatory body with binding rules comparable to the EU AI Act, and its detailed powers, funding arrangements and decision-making structures are not yet clear.
\nStill, Serbia’s role as a founding member raises questions about how it will navigate different approaches to AI governance.
\nThe European Union has built its policy around regulation, fundamental rights and legal accountability. China is placing greater emphasis on state-led development, international cooperation and access for developing economies. The United States continues to rely more heavily on industry-led innovation, national security restrictions and selective federal oversight.
\nSerbia now has a formal place inside one of those emerging international initiatives while continuing to pursue EU membership.
\nThat position could give Belgrade access to new research partnerships, training programmes and AI infrastructure. It could also allow Serbia to represent the interests of smaller and developing countries in discussions that are often dominated by the world’s largest economies.
\nThe value of membership will depend on what WAICO becomes in practice.
\nThe organization has been formally established, but little has been disclosed about how it will operate, what commitments members have accepted or whether it will produce standards, finance projects or coordinate national policies.
\nIt is also unclear how much influence smaller members will have compared with China and the organization’s other larger founding states.
\nSerbia has spent several years presenting itself as a regional leader in artificial intelligence. It adopted its first national AI strategy in 2019, established an AI research institute and has invested in national computing infrastructure and public-sector applications.
\nJoining WAICO gives that strategy an international dimension, but it also places Serbia within a wider contest over who will shape the rules and institutions governing AI.
\nThe next indication of the organization’s direction will come when its membership structure, leadership and first programmes are announced. Until then, Serbia’s decision is politically significant, even if the practical benefits remain undefined.
","featured_image_id":208,"intro_image_id":208,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":11,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":40,"ordering":0,"published_at":"2026-07-18 11:25:17","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-18 11:25:17","updated_at":"2026-07-18 11:25:17","key_takeaways":null,"faq":null,"sources":null,"category_name":"Serbia","category_slug":"serbia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/8e098fedb8d9bceba9a7dc75015509de.webp","featured_image_alt":"Shanghai Serbia Attendance","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/07/8e098fedb8d9bceba9a7dc75015509de.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Serbia Joins China-Led Global AI Organization","meta_description":"Serbia has joined 28 countries in establishing WAICO, a China-backed intergovernmental organization for AI cooperation and governance.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":27,"title":"Serbia Strengthens AI Infrastructure With NVIDIA Tech","slug":"serbia-strengthens-ai-infrastructure-with-nvidia-tech","short_description":"","content":"Serbia is working to build the foundations for a new era of artificial intelligence, with digital infrastructure becoming one of the country’s most important priorities. As AI begins to transform companies, institutions and entire industries, the key question is no longer only who has the best software, but who controls the data, computing power and infrastructure needed to develop advanced systems.
\nRecent reporting from Serbia highlights NVIDIA’s role in this wider AI ecosystem, showing how global hardware and computing technologies are becoming central to national AI strategies. Artificial intelligence requires powerful processors, secure data centers and specialized platforms capable of handling large-scale calculations. Without that infrastructure, even the best ideas and research teams can remain limited.
\nFor Serbia, this is especially important because the country wants to move from basic digitalization toward more advanced technological development. Building AI infrastructure means creating the conditions for researchers, startups, companies and public institutions to develop their own solutions instead of relying only on foreign platforms.
\nAI infrastructure includes several elements. The first is computing power, which allows models to be trained, tested and deployed. The second is data infrastructure, because AI systems depend on large, well-organized and secure datasets. The third is technical expertise, since advanced infrastructure is useful only if there are engineers, researchers and institutions capable of using it properly.
\nSerbia has already invested in data centers, digital government systems and supercomputing capacity. These investments are now becoming more relevant because artificial intelligence depends on exactly these foundations. A country that has strong data and computing infrastructure can support more complex projects in healthcare, education, industry, agriculture, public administration and science.
\nNVIDIA technology is important because the company’s chips and platforms are widely used in AI development around the world. Advanced AI models often rely on graphic processing units and specialized hardware that can process huge amounts of information quickly. By connecting with this type of technology, Serbia can strengthen its ability to participate in modern AI development.
\nThe focus on infrastructure also shows that Serbia is thinking beyond simple AI applications such as chatbots or automation tools. The bigger goal is to create an ecosystem where domestic institutions and companies can build, test and run more serious AI systems. That could include medical image analysis, smart energy systems, industrial optimization, Serbian-language AI models and public-sector data tools.
\nFor the private sector, stronger infrastructure could reduce barriers to innovation. Many startups and smaller companies cannot afford expensive computing resources on their own. If Serbia can provide access through national or institutional platforms, more local teams could experiment with AI and bring products to market.
\nFor universities and researchers, better infrastructure could mean stronger scientific output and deeper cooperation with international partners. AI research is increasingly dependent on access to modern hardware. Without it, talented researchers may be forced to work with foreign institutions or leave the country to develop advanced projects.
\nPublic institutions could also benefit, but only if AI is introduced responsibly. Better infrastructure can help the state improve services, analyze data and plan policies more effectively. However, public use of AI must be transparent, secure and subject to human oversight, especially when citizens’ data and rights are involved.
\nSerbia’s AI ambitions will depend on whether infrastructure investments are matched with education, regulation and practical implementation. Hardware alone is not enough. The country will need trained professionals, clear rules, strong cybersecurity and cooperation between the state, academia and business.
\nThe latest focus on AI infrastructure and NVIDIA-related technology suggests that Serbia is preparing for a more serious phase of artificial intelligence development. Instead of treating AI only as a future opportunity, the country is beginning to build the technical base needed to use it today.
\nIf these efforts continue, Serbia could strengthen its position as a regional AI hub. The real test will be whether infrastructure becomes accessible and useful for local innovators, researchers and institutions — and whether it leads to products, services and public benefits that citizens and businesses can actually feel.
","featured_image_id":63,"intro_image_id":63,"featured_image_credit":"nVidia","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":11,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":8,"ordering":0,"published_at":"2026-06-25 11:30:42","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 11:30:42","updated_at":"2026-06-25 11:30:42","key_takeaways":null,"faq":null,"sources":null,"category_name":"Serbia","category_slug":"serbia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/7818c02adce7f0b79961667d13857df1.webp","featured_image_alt":"Nvidia Logo connected in some mindmap","featured_image_width":1000,"featured_image_height":600,"intro_image_path":"/api/uploads/2026/06/7818c02adce7f0b79961667d13857df1.webp","intro_image_width":1000,"intro_image_height":600,"seo_title":"Serbia Strengthens AI Infrastructure With NVIDIA Tech","meta_description":"Serbia is working to build the foundations for a new era of artificial intelligence, with digital infrastructure becoming one of the country’s…","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":26,"title":"Serbia Taps Kai-Fu Lee on AI Transformation","slug":"serbia-taps-kai-fu-lee-on-ai-transformation","short_description":"","content":"Serbia has placed artificial intelligence high on its national development agenda, with President Aleksandar Vučić meeting Chinese AI expert Kai-Fu Lee in Belgrade to discuss how the technology is reshaping economies, education, healthcare, industry and public services.
\nThe meeting signals Serbia’s growing interest in artificial intelligence not only as a technology trend, but as a strategic tool for modernization. AI is increasingly seen as a force that can change how governments provide services, how companies compete and how education systems prepare young people for future jobs.
\nKai-Fu Lee is one of the most recognized global figures in artificial intelligence, known for his work in technology, investment and AI development. His visit to Serbia comes at a time when the country is trying to strengthen its digital infrastructure, attract international knowledge and position itself more seriously in the global AI race.
\nAccording to the public message after the meeting, the conversation focused on the way artificial intelligence is transforming key sectors of society. In the economy, AI can help companies automate processes, analyze data faster and improve productivity. In industry, it can support smarter production, predictive maintenance and better supply-chain management. In healthcare, it can assist doctors with diagnostics, patient data analysis and research. In education, it can change how students learn and how teachers personalize lessons.
\nFor Serbia, the most important question is how to turn these possibilities into practical benefits. The country has a strong IT sector and a growing number of technology professionals, but AI development also requires infrastructure, investment, data access, regulation and specialized education. Meetings with international experts can help Serbia understand what steps are needed to avoid falling behind larger markets.
\nThe discussion also points to the importance of public-sector modernization. Artificial intelligence could help governments process information faster, reduce administrative delays and improve services for citizens. However, such use must be carefully managed because public institutions handle sensitive data and decisions that affect people’s rights.
\nThat is why Serbia’s AI strategy cannot depend only on technology. It must also include ethical rules, transparency, cybersecurity and protection of citizens. If AI is used in public services, citizens need to know when automated systems are involved, how decisions are made and who is responsible if mistakes happen.
\nThe meeting with Kai-Fu Lee also has a wider geopolitical and economic meaning. Countries around the world are competing to attract AI investment, talent and partnerships. Serbia is trying to show that it wants to be part of that global conversation and not only a consumer of imported technology.
\nFor Serbian businesses, the message is clear: artificial intelligence is becoming a serious economic factor. Companies that learn how to use AI responsibly may become more competitive, while those that ignore it may struggle to keep up. This applies not only to software firms, but also to manufacturing, logistics, finance, media, education and healthcare.
\nThe next challenge for Serbia will be turning high-level discussions into concrete programs. That means investing in AI education, supporting startups, improving access to computing resources and helping companies adopt AI tools safely. It also means preparing public institutions to use technology in a way that is efficient but accountable.
\nThe meeting with Kai-Fu Lee shows that Serbia wants to be active in shaping its AI future. Whether that ambition becomes reality will depend on how quickly the country can combine expertise, infrastructure, regulation and real implementation across the economy and public sector.
","featured_image_id":62,"intro_image_id":62,"featured_image_credit":"Kaust","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":11,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":14,"ordering":0,"published_at":"2026-06-25 11:27:46","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 11:27:46","updated_at":"2026-06-25 11:27:46","key_takeaways":null,"faq":null,"sources":null,"category_name":"Serbia","category_slug":"serbia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/f4007877928e21adcbef058118035341.webp","featured_image_alt":"Kai fu lee talking in some event","featured_image_width":1600,"featured_image_height":1067,"intro_image_path":"/api/uploads/2026/06/f4007877928e21adcbef058118035341.webp","intro_image_width":1600,"intro_image_height":1067,"seo_title":"Serbia Taps Kai-Fu Lee on AI Transformation | AI Balkans","meta_description":"Serbia has placed artificial intelligence high on its national development agenda, with President Aleksandar Vučić meeting Chinese AI expert Kai-Fu Lee…","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":25,"title":"Serbia Launches SAIFA AI Infrastructure Project","slug":"serbia-launches-saifa-ai-infrastructure-project","short_description":"","content":"Serbia has officially launched the SAIFA project, short for Serbian Artificial Intelligence Factory Antenna, a new initiative designed to connect the country more closely with Europe’s artificial intelligence and high-performance computing ecosystem. The project is coordinated by the School of Electrical Engineering at the University of Belgrade and supported through the EuroHPC Joint Undertaking.
\nSAIFA is part of a wider European effort to expand access to advanced AI infrastructure, computing resources, data tools and expert support. Through this project, Serbia aims to strengthen its national AI capacity and provide researchers, startups, public institutions and companies with better access to the resources needed to develop and test artificial intelligence solutions.
\nThe project is important because modern AI development requires more than just software knowledge. Advanced AI systems often need large datasets, strong computing power, technical expertise, testing environments and cooperation between universities, industry and government. SAIFA is intended to become a bridge between Serbia’s local innovation ecosystem and the broader European AI network.
\nBy joining this network, Serbia is positioning itself not only as a user of artificial intelligence technologies, but also as a country that wants to participate in their development. The initiative could help domestic researchers and businesses work on more complex AI projects, including applications in healthcare, energy, agriculture, sustainability, culture and language technologies.
\nThe role of the University of Belgrade’s School of Electrical Engineering is especially significant. As coordinator, the faculty is expected to bring together academic expertise, technical infrastructure and cooperation with other institutions. This gives the project a strong research foundation and connects it with Serbia’s existing scientific and engineering community.
\nSAIFA is also linked to Europe’s wider strategy of building AI Factories and AI Factory Antennas. These initiatives are meant to make advanced computing resources more accessible across the continent, especially for countries and organizations that may not have enough infrastructure on their own. For Serbia, this could mean easier access to European-level tools and knowledge without having to build everything independently from the beginning.
\nThe project could be particularly useful for startups and small companies. Many young AI businesses have ideas and technical talent but lack access to powerful computing resources or specialized support. If SAIFA succeeds, it could lower some of those barriers and make it easier for local teams to test, improve and deploy AI products.
\nPublic institutions may also benefit from the project. AI can support better public services, data analysis, planning and decision-making, but only if it is developed and used responsibly. A national AI hub could help institutions understand how to apply AI in a safer and more structured way.
\nThe launch of SAIFA comes at a time when Serbia is also preparing its first law on artificial intelligence. Together, these developments show that the country is working on two sides of the same challenge: building the infrastructure needed for AI innovation, while also preparing rules to manage the risks.
\nFor Serbia’s technology sector, the project represents a chance to move higher in the AI value chain. Instead of being seen mainly as a regional outsourcing or software services market, Serbia could strengthen its position as a place where more advanced AI research, testing and development can happen.
\nThe success of SAIFA will depend on how widely its services are used and whether it can create real cooperation between academia, industry, startups and the public sector. If the project becomes accessible and practical, it could become one of the central pillars of Serbia’s AI development in the coming years.
\nIn the broader regional context, SAIFA also gives Serbia a stronger role in Europe’s AI infrastructure map. As artificial intelligence becomes increasingly important for economic competitiveness, scientific research and public policy, projects like this could determine which countries are able to keep pace with technological change and which remain only consumers of foreign AI solutions.
","featured_image_id":61,"intro_image_id":61,"featured_image_credit":"SAIFA team","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":11,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":5,"ordering":0,"published_at":"2026-06-25 11:19:29","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 11:19:29","updated_at":"2026-06-25 11:19:29","key_takeaways":null,"faq":null,"sources":null,"category_name":"Serbia","category_slug":"serbia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/8543055ad58d0f259550a04c726fa72d.webp","featured_image_alt":"Team of the Saifa Launch in Serbia","featured_image_width":1024,"featured_image_height":577,"intro_image_path":"/api/uploads/2026/06/8543055ad58d0f259550a04c726fa72d.webp","intro_image_width":1024,"intro_image_height":577,"seo_title":"Serbia Launches SAIFA AI Infrastructure Project | AI Balkans","meta_description":"Serbia has officially launched the SAIFA project, short for Serbian Artificial Intelligence Factory Antenna, a new initiative designed to connect the…","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":24,"title":"Serbia’s First AI Law Expected by 2026","slug":"serbia-s-first-ai-law-expected-by-2026","short_description":"","content":"BELGRADE — Serbia is preparing to adopt its first comprehensive law on artificial intelligence, marking one of the country’s most important steps so far in regulating the fast-growing AI sector. The proposed law is expected to be introduced by the end of 2026 and will create a formal legal framework for how artificial intelligence systems are developed, used and supervised in both the public and private sectors.
\nThe new regulation is expected to be largely aligned with the European Union’s AI Act, currently considered one of the most advanced artificial intelligence laws in the world. For Serbia, this means the country is trying to move closer to European digital standards while also creating rules that fit the needs of its own technology market.
\nThe future law will focus especially on AI systems that can directly affect citizens’ rights, safety or access to important services. These include systems used in healthcare, education, employment, security, recruitment, finance and public administration. In practice, this could include algorithms that support medical decisions, software used in hiring processes, tools that assess loan applications, or AI systems used by institutions to process large amounts of personal or sensitive data.
\nOne of the key ideas behind the planned law is the classification of AI systems according to risk. Low-risk systems would face fewer obligations, while high-risk systems would have to meet stricter standards related to transparency, safety, documentation and human oversight. Certain uses of AI could also be restricted or prohibited if they are considered too dangerous or harmful.
\nA central principle of the future framework is that artificial intelligence should not replace human responsibility in sensitive decisions. Instead, AI should be used as a support tool, while the final decision in important cases should remain under human control. This is especially relevant in sectors where mistakes or bias could have serious consequences for individuals.
\nThe law is also expected to give Serbian companies clearer rules for developing and selling AI-based products. Until now, Serbia has relied mostly on strategy documents, general digital policy and existing sectoral regulations. A dedicated AI law would create more legal certainty for startups, IT companies, researchers and investors.
\nSupporters of the new framework argue that regulation does not have to slow innovation. On the contrary, clear rules could help companies build more trustworthy products and prepare for cooperation with European partners. Since many Serbian technology companies already work with EU clients, alignment with European standards could become an advantage rather than a burden.
\nAt the same time, the law will need to avoid creating unnecessary pressure on smaller companies and startups. If compliance becomes too expensive or complicated, young AI firms could struggle to compete. That is why one of the biggest challenges for lawmakers will be to protect citizens without blocking innovation.
\nThe preparation of the law is connected to Serbia’s broader artificial intelligence strategy, which aims to strengthen the domestic AI ecosystem by 2030. The country has already shown interest in building AI capacity through education, research, infrastructure and cooperation with European initiatives.
\nIf adopted as expected, the law would place Serbia among the countries in the region taking a more structured approach to artificial intelligence governance. It would also signal that AI is no longer treated only as a technological opportunity, but as an area that requires clear rules, accountability and public trust.
\nFor citizens, the most important question will be whether the law can ensure that AI systems are safe, fair and transparent. For businesses, the key issue will be whether regulation can provide stability without limiting growth. For the state, the challenge will be implementation: writing a law is only the first step, while enforcement, expertise and institutional capacity will determine how effective the framework becomes in practice.
","featured_image_id":60,"intro_image_id":60,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":11,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":14,"ordering":0,"published_at":"2026-06-25 11:17:03","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 11:17:03","updated_at":"2026-06-25 11:17:03","key_takeaways":null,"faq":null,"sources":null,"category_name":"Serbia","category_slug":"serbia","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/98d1ca489b25da009fe1d82b40c9c094.webp","featured_image_alt":"AI 2026 preview of tools with woman handing some of the tools","featured_image_width":1200,"featured_image_height":630,"intro_image_path":"/api/uploads/2026/06/98d1ca489b25da009fe1d82b40c9c094.webp","intro_image_width":1200,"intro_image_height":630,"seo_title":"Serbia’s First AI Law Expected by 2026 | AI Balkans","meta_description":"BELGRADE — Serbia is preparing to adopt its first comprehensive law on artificial intelligence, marking one of the country’s most important…","robots":"index","canonical_url":"","og_title":"","og_description":""}]},{"category":{"id":13,"name":"Montenegro","slug":"montenegro"},"posts":[{"id":160,"title":"Western Balkans Digital Summit 2026 to Focus on AI","slug":"western-balkans-digital-summit-2026-to-focus-on-ai","short_description":null,"content":"The Western Balkans Digital Summit will bring government officials, technology experts and business representatives to Herceg Novi, Montenegro, from October 12 to 14, with artificial intelligence, cybersecurity and regional digital integration among the central topics.
\nThe eighth edition will take place at One&Only Portonovi under Montenegro's 2026 presidency of the Berlin Process. More than 150 participants are expected, including representatives of the six Western Balkan economies, the European Commission, academic institutions, businesses and civil society organizations.
\n
The event is organized by Montenegro's Ministry of Public Administration in partnership with the Regional Cooperation Council (RCC), which has supported the summit since its introduction in 2018.
\nAccording to the Montenegrin government's announcement, this year's programme is intended to address practical challenges in developing a more connected, secure and competitive regional digital economy.
\nThe summit's programme is organized around five main areas: digital economy, digital skills, digital infrastructure, artificial intelligence and connectivity, and trust and security.
\nAI discussions will focus on the implementation of advanced technologies and their role in supporting regional connectivity. The wider programme will also examine how businesses, governments and educational institutions can develop the skills and infrastructure needed for digital transformation.
\nCybersecurity will be another central topic, with attention to strengthening the region's resilience against digital threats and improving cooperation between institutions.
\nThe Regional Cooperation Council has also identified digital public services and online child protection among the subjects expected to feature in the discussions.
\nFor the Western Balkans, these issues extend beyond technology adoption. Differences in institutional capacity, digital infrastructure and access to skilled professionals affect how quickly countries can introduce new digital services and benefit from AI applications.
\nIn an interview published by the Ministry of Public Administration on October 8, Montenegro's Minister of Public Administration, Marash Dukaj, outlined the country's expectations for the summit.
\nHe said regional cooperation should produce practical results in areas such as cross-border data exchange, digital identity, public services and cybersecurity.
\nOne priority is closer alignment with the European Union's Digital Single Market, including work toward more affordable mobile roaming between the Western Balkans and EU countries.
\nDukaj also pointed to the need for stronger cooperation on cybersecurity, arguing that increasingly interconnected public systems require coordinated responses to digital threats.
\nMontenegro has announced plans to expand its digital public services and develop an AI assistant for its e-government platform.
\nThe minister stressed that AI in public administration must remain subject to human oversight, particularly when systems are involved in legal or administrative decisions.
\nThese are national priorities presented by Montenegro ahead of the summit, rather than agreements already reached by participating governments.
\nThe Western Balkans Digital Summit has served as a meeting point for regional governments and technology stakeholders since 2018. The previous edition took place in Skopje in October 2025.
\nThis year's gathering comes as the region continues working to align its digital policies and infrastructure with European frameworks.
\nFor businesses, improved digital connectivity and interoperable public services could reduce administrative barriers when operating across borders.
\nFor governments, common approaches to cybersecurity, digital identity and AI governance could support more reliable public services and regional cooperation.
\nHowever, the extent to which these ambitions translate into practical results will depend on agreements, financing and implementation after the summit.
\nThe most relevant outcomes will be any concrete commitments involving AI governance, cybersecurity cooperation, digital public services and regional connectivity.
\nParticipants are also expected to discuss how to strengthen digital skills and prepare the workforce for changing technological requirements.
\nWith AI increasingly entering public administration and business operations, the challenge for Western Balkan economies is to move from policy discussions toward measurable improvements in services, institutional capacity and digital infrastructure.
\nThe October 12–14 summit provides another opportunity to establish shared priorities. Its longer-term significance will depend on what participating institutions agree to implement after leaving Herceg Novi.
","featured_image_id":364,"intro_image_id":364,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":13,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":87,"ordering":0,"published_at":"2026-10-09 08:34:38","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-09 08:34:38","updated_at":"2026-10-09 08:34:38","key_takeaways":"[\"The eighth Western Balkans Digital Summit takes place in Herceg Novi, Montenegro, from October 12 to 14, with over 150 participants expected.\",\"Artificial intelligence, cybersecurity, infrastructure, digital skills and economic integration form the programme's five central areas.\",\"Montenegro wants the summit to advance practical regional cooperation, including digital public services and closer integration with the EU digital market.\"]","faq":"[{\"q\":\"When and where is the Western Balkans Digital Summit 2026?\",\"a\":\"The summit is scheduled for October 12–14, 2026, at One&Only Portonovi in Herceg Novi, Montenegro.\"},{\"q\":\"What are the main topics?\",\"a\":\"AI, cybersecurity, digital economy, digital skills, infrastructure, connectivity and regional digital integration.\"},{\"q\":\"Who organizes the summit?\",\"a\":\"Montenegro's Ministry of Public Administration, under government patronage and in partnership with the Regional Cooperation Council.\"},{\"q\":\"Can participants register?\",\"a\":\"The government announced that registration is open through its official event registration link.\"}]","sources":"[{\"title\":\"Government of Montenegro, October 7 announcement\",\"url\":\"https://www.gov.me/clanak/otvorene-prijave-za-osmi-digitalni-samit-zapadnog-balkana-u-herceg-novom\"},{\"title\":\"Regional Cooperation Council, summit information\",\"url\":\"https://www.rcc.int/events/1670/rcc-co-organizes-8th-western-balkans-digital-summit\"},{\"title\":\"Montenegro Ministry of Public Administration, October 8 interview\",\"url\":\"https://www.gov.me/clanak/dukaj-crna-gora-zeli-da-kroz-samit-ubrza-povezivanje-regiona-sa-digitalnim-trzistem-eu\"},{\"title\":\"Western Balkans Digital Summit website\",\"url\":\"https://digitalsummitwb6.com/\"}]","category_name":"Montenegro","category_slug":"montenegro","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/cebed43670cecd766b9558cfff22f5b9.webp","featured_image_alt":"Montenegro","featured_image_width":1200,"featured_image_height":799,"intro_image_path":"/api/uploads/2026/10/cebed43670cecd766b9558cfff22f5b9.webp","intro_image_width":1200,"intro_image_height":799,"seo_title":"Western Balkans Digital Summit 2026 to Focus on AI","meta_description":"The Western Balkans Digital Summit opens in Montenegro on October 12, bringing AI, cybersecurity, digital skills and EU integration into focus.","robots":"index","canonical_url":"western-balkans-digital-summit-2026-ai","og_title":"","og_description":""},{"id":89,"title":"Montenegro Hosts Eastern Europe’s Machine Learning Summer School","slug":"montenegro-hosts-eastern-europe-s-machine-learning-summer-school","short_description":"","content":"Montenegro will host the 2026 Eastern European Machine Learning Summer School in Cetinje from July 27 to August 1, bringing together students, researchers and technology professionals from across the region.
\nThe six-day programme will include lectures, practical sessions and discussions covering artificial intelligence and machine learning. It will be held at Hotel Grand in Cetinje and conducted in English.
\nEEML is primarily an educational and research programme rather than a conventional technology conference. It is designed to connect emerging researchers and students with experienced AI professionals and international experts.
\nThis year’s edition is supported by organisations and institutions from across the region, including the Montenegrin AI Association, the Institute for Artificial Intelligence Research and Development of Serbia and AI Albania.
\nFor Montenegro, hosting EEML is an opportunity to strengthen its position within the growing Western Balkan AI community. It also gives regional students and researchers access to advanced training and professional networks that are often difficult to find locally.
\nThe event comes as Western Balkan countries increasingly discuss artificial intelligence as part of their economic and education strategies. However, the region still faces limited research funding, a shortage of specialised programmes and the continued departure of skilled professionals.
\nEEML cannot resolve those challenges alone, but it can help build stronger connections between regional universities, researchers and AI organisations.
\nFor one week, Cetinje will become an important meeting point for Eastern Europe’s machine-learning community. The greater test will be whether the partnerships created there continue after the programme ends.
","featured_image_id":216,"intro_image_id":216,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":13,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":2,"view_count":22,"ordering":0,"published_at":"2026-07-26 09:12:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-26 09:12:15","updated_at":"2026-07-26 09:13:17","key_takeaways":null,"faq":null,"sources":null,"category_name":"Montenegro","category_slug":"montenegro","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/bf8d46002f32d15ec63f173ea97ee719.webp","featured_image_alt":"ECML logo and event dates overlay a sunset aerial view of Cetinje, Montenegro.","featured_image_width":1024,"featured_image_height":434,"intro_image_path":"/api/uploads/2026/07/bf8d46002f32d15ec63f173ea97ee719.webp","intro_image_width":1024,"intro_image_height":434,"seo_title":"Montenegro Hosts Machine Learning Summer School","meta_description":"Montenegro hosts EEML 2026 in Cetinje, bringing together regional students, researchers and experts in AI and machine learning.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":54,"title":"Game Changer 2026 Brings AI Talks to Montenegro","slug":"game-changer-2026-brings-ai-talks-to-montenegro","short_description":"","content":"Montenegro’s coastal city of Tivat became a regional meeting point for technology, innovation, investment, and artificial intelligence during Game Changer Montenegro 2026.
\nThe festival, held from 18 to 20 June 2026, brought together more than 300 participants from the fields of technology, business, tourism, investment, startups, and innovation. Across its conference programme, the event opened discussions on several themes shaping the future of the region, including artificial intelligence, fintech, cybersecurity, sustainable technology, startup development, and digital transformation.
\n![\"Panel](\"/api/uploads/2026/06/916cfb57183df26474c560d6dcaed51f.webp\")
\nFor Montenegro, the event was more than a technology conference. It showed how smaller Western Balkan countries can use focused regional gatherings to position themselves inside bigger conversations about Europe’s digital economy.
\nArtificial intelligence was one of the central topics in the wider programme. The discussions reflected a growing regional shift: AI is no longer being treated only as a global trend or a topic for large technology markets. It is increasingly becoming part of conversations around investment, public services, tourism, infrastructure, cybersecurity, and business competitiveness in the Western Balkans.
\nThis matters for Montenegro because the country is not trying to compete with major AI economies on scale. Instead, its opportunity may be in becoming a focused regional connector. With its position on the Adriatic coast, growing tourism sector, and increasing interest in digital services, Montenegro can use events like Game Changer to attract investors, founders, technology companies, and policy voices from across the region.
\nThe festival also connected AI with other strategic sectors. Fintech, cybersecurity, sustainable technology, and startup development are all areas where AI is already becoming a practical tool. Banks and fintech companies are using AI to improve fraud detection, customer service, and risk analysis. Cybersecurity teams are using AI to detect threats faster, while attackers are also using the same technologies to create new risks. Tourism businesses can use AI for personalization, pricing, translation, customer support, and operational planning.
\n![\"Game](\"/api/uploads/2026/06/9f56213d6696d4576e0354449b14eb37.webp\")
\nFor a country like Montenegro, these connections are important. The future of AI in the region will not depend only on whether countries build large language models or national AI labs. It will also depend on whether businesses, public institutions, and local ecosystems understand how to use AI in real sectors.
\nGame Changer Montenegro 2026 also gave visibility to the role of community in digital transformation. Technology ecosystems are not built only through laws, grants, or infrastructure. They also need spaces where founders, investors, companies, public institutions, and young professionals can meet and exchange ideas. In that sense, the festival helped create a stronger regional bridge between business, technology, and public conversation.
\nThe Western Balkans are currently entering a new stage of AI development. North Macedonia is moving through national AI infrastructure and language technology discussions. Serbia and Albania are preparing AI regulation. Kosovo is facing the challenge of AI-generated disinformation. Bosnia and Herzegovina is discussing responsible AI use in media. Montenegro’s role may be different: building a place where technology, investment, tourism, and digital infrastructure meet.
\nThe key question now is whether events like Game Changer can lead to long-term ecosystem results. Conferences can create visibility, but the real impact depends on what happens after them. Montenegro will need stronger links between startups, universities, investors, government institutions, and regional partners if it wants to turn AI conversations into concrete projects.
\nStill, Game Changer Montenegro 2026 showed that the country wants to be part of the region’s technology future. For the Western Balkans, that is an important signal. AI development in the region will not come from one country alone. It will come from multiple local ecosystems building different strengths and connecting them across borders.
\nMontenegro may not be the largest AI market in the Balkans, but it can become a meaningful regional platform for AI, innovation, and digital economy discussions.
\nAnd in the next phase of the AI race, that kind of positioning may matter more than size.
","featured_image_id":113,"intro_image_id":113,"featured_image_credit":"","featured_image_overlay":"Gamechanger 2026","featured_image_overlay_position":"bottom-left","category_id":13,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":51,"ordering":0,"published_at":"2026-06-29 12:11:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-29 12:11:24","updated_at":"2026-06-29 16:02:55","key_takeaways":null,"faq":null,"sources":null,"category_name":"Montenegro","category_slug":"montenegro","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/80767e20752bfdc3b4fe9aeeca1847e5.webp","featured_image_alt":"Gamechanger Montenegro Gathering","featured_image_width":1200,"featured_image_height":630,"intro_image_path":"/api/uploads/2026/06/80767e20752bfdc3b4fe9aeeca1847e5.webp","intro_image_width":1200,"intro_image_height":630,"seo_title":"Game Changer 2026 Brings Regional AI Talks to Montenegro","meta_description":"Montenegro hosted Game Changer 2026, bringing regional voices together for AI, technology, innovation, and digital transformation talks.","robots":"index","canonical_url":"","og_title":"","og_description":""}]},{"category":{"id":10,"name":"Kosovo","slug":"kosovo"},"posts":[{"id":146,"title":"Kosovo Looks to AI to Promote Tourism","slug":"kosovo-looks-to-ai-to-promote-tourism","short_description":"","content":"Kosovo is looking to artificial intelligence and digital tools as part of a new effort to rethink how the country promotes itself as a tourism destination.
\nThe government used this year’s World Tourism Day conference in Pristina to launch the concept “RE: TOURISM – AI in the promotion of tourism in Kosovo”, bringing artificial intelligence directly into discussions about the country’s tourism strategy.
\n![\"Albin](\"/api/uploads/2026/09/5fe9c3407ad51d20d4d0eae6a105d043.webp\")
The conference was organised by Kosovo’s Ministry of Culture and Tourism on September 25 and opened by Prime Minister Albin Kurti.
\nAccording to the government, the idea is to connect Kosovo’s tourism potential with digitalisation and AI as the way people search for information, plan trips and choose destinations increasingly changes.
\nThe initiative is also linked to the broader international theme promoted by UN Tourism around the use of digital technologies and artificial intelligence in redesigning tourism.
\nKosovo has been trying to position tourism as a larger part of its economic development.
\nKurti said visitor numbers have grown by an average of around 20% annually over the past four years and pointed to the country's Tourism Strategy 2030 as the framework for expanding the sector. The figure was presented by the government at the conference.
\nThe new RE: TOURISM concept adds technology to that strategy.
\nThe government has not yet announced a specific AI platform, application or procurement programme connected to the initiative. Instead, the conference focused on how digitalisation and AI could change the way Kosovo presents destinations and reaches international visitors.
\nThat distinction matters. For now, this is a policy and promotion direction rather than the launch of a finished AI product.
\nPotential applications could eventually include how tourism information is organised and delivered, multilingual visitor services, digital destination discovery and more personalised promotion. The government announcement itself, however, does not specify which AI systems will be deployed.
\nTourism is becoming an increasingly relevant AI use case for the Western Balkans.
\n![\"Kosovo](\"/api/uploads/2026/09/1dc5c049803f0f3d0d0543301b60c9e1.webp\")
Unlike sectors that require large computing infrastructure or advanced AI research, tourism offers countries in the region a more immediate opportunity to use existing technologies in areas such as translation, visitor information, recommendations, marketing and service delivery.
That makes the sector particularly relevant for smaller economies where tourism already contributes directly to local businesses and employment.
\nKosovo's government has also linked tourism development with employment opportunities, particularly for young people and women, while calling for cooperation between public institutions, municipalities, businesses, local communities and the diaspora.
\nThe country's diaspora already plays an important role in bringing visitors and investment into Kosovo, another factor the government highlighted during the conference.
\nThe next question is what RE: TOURISM becomes in practice.
\nKosovo now has a stated direction linking AI and digitalisation with its Tourism Strategy 2030, but the practical impact will depend on whether that direction produces actual tools, datasets, partnerships or public services.
\nFor AI Balkans, that will be the part worth watching.
\nThe use of AI in regional tourism will matter less as a marketing theme than as a test of whether governments and businesses can turn widely available AI technologies into services that visitors genuinely use.
","featured_image_id":324,"intro_image_id":324,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":10,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":31,"ordering":0,"published_at":"2026-09-27 16:02:04","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-27 16:02:04","updated_at":"2026-09-27 16:02:04","key_takeaways":"[\"Kosovo launched the RE: TOURISM concept on September 25, focused on AI and digitalisation in tourism promotion.\",\"The initiative is connected to Kosovo’s broader Tourism Strategy 2030.\",\"No specific AI product or platform has yet been announced, so the development is currently a strategic direction rather than a deployed system.\"]","faq":null,"sources":null,"category_name":"Kosovo","category_slug":"kosovo","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/0834bb48a71ca049e15f34944bca7fb5.webp","featured_image_alt":"Kosovo's prime minister addresses an audience before an AI-themed backdrop and the Union i Turizmit Kosovës logo.","featured_image_width":1024,"featured_image_height":683,"intro_image_path":"/api/uploads/2026/09/0834bb48a71ca049e15f34944bca7fb5.webp","intro_image_width":1024,"intro_image_height":683,"seo_title":"Kosovo Looks to AI to Promote Tourism","meta_description":"Kosovo is bringing AI and digitalisation into its tourism strategy through the new RE: TOURISM initiative launched in Pristina.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":133,"title":"Kosovo Details AI Plans for eKosova 2.0","slug":"kosovo-details-ai-plans-for-ekosova-2-0","short_description":"","content":"Kosovo has provided new details on how artificial intelligence will become part of the next generation of its eKosova public services platform, with the government planning more integrated services, stronger data infrastructure and a shift toward cloud-based government systems.
\n![\"AI](\"/api/uploads/2026/09/75222c1314ae7403a384af264d9029e9.webp\")
Speaking at the AI & Tech International Summit 2026 in Prishtina on September 3, acting Prime Minister Albin Kurti said the government is moving toward eKosova 2.0, describing the next version of the platform as more user-centred and increasingly enabled by artificial intelligence.
The announcement builds on earlier plans for the platform but provides a clearer picture of the infrastructure Kosovo wants to put behind its next phase of digital government.
\nAccording to Kurti, public services are being redesigned around \"life events\", while the government is strengthening interoperability between systems and improving the infrastructure used to manage and exchange data.
\nKosovo is also modernising government networks, moving toward cloud infrastructure and increasing investment in cybersecurity.
\nThe existing eKosova platform already brings hundreds of public services into a common digital environment.
\nAI Balkans previously reported on plans to expand the system through eKosova 2.0, including the expected introduction of AI features and hundreds of additional administrative services.
\nThe latest government statement goes further by placing AI within a wider redesign of how those services are delivered.
\n![\"Kurti.jpg\"](\"/api/uploads/2026/09/e96dd44648317101d7c87323a854dde8.webp\")
Kurti said AI could make public services easier to access, more personalised and more proactive. He also pointed to potential uses inside institutions, including reducing administrative workload, improving how information is used and supporting decision-making.
No detailed list of AI systems or specific models that will be deployed through eKosova 2.0 was announced.
\nThe government has also not provided a timetable for when individual AI-enabled features will become available to citizens.
\nThat distinction matters. There is a substantial difference between announcing that a digital platform will support AI and deploying AI systems that directly interact with citizens or assist public officials in administrative decisions.
\nOne of the more significant details in the government's latest plans is the move toward organising digital services around life events.
\nInstead of requiring citizens to understand which individual institution handles each procedure, this approach can group services around situations such as starting a business, having a child, moving home or accessing other public services.
\nThat model depends heavily on government systems being able to communicate with each other.
\nKosovo's emphasis on interoperability and data infrastructure therefore matters as much as the AI element itself.
\nWithout reliable connections between public databases and institutions, AI assistants or personalised digital services would have limited information to work with and could add another interface without solving the underlying administrative fragmentation.
\nThe move toward cloud infrastructure adds another dimension to Kosovo's digital transformation.
\nGovernment systems increasingly require computing capacity that can scale as more services move online and larger amounts of data are processed.
\nBut moving public infrastructure toward the cloud also increases the importance of cybersecurity, access controls, data governance and decisions about where government information is stored.
\nKurti identified cybersecurity as one of the areas where Kosovo has already been investing and said further work is part of the foundations for the next stage of digital government.
\nThe same safeguards will become more important as AI is introduced.
\nPublic-sector AI systems may process information that citizens cannot choose to withhold from government. That creates a different level of responsibility from the use of consumer AI tools.
\nQuestions around privacy, transparency, human oversight and responsibility for incorrect outputs will therefore need to develop alongside the technology.
\nKurti acknowledged that broader challenge during his speech, saying AI adoption should be accompanied by accountability, privacy, cybersecurity, transparency and human oversight.
\n![\"eKosovo.jpg\"](\"/api/uploads/2026/09/61e54622d247cbbaf77dcbd76f989a43.webp\")
The government's ambitions extend beyond public administration.
\nKurti said AI could help Kosovo's technology companies move further up the global value chain, shifting from providing technology services toward developing more advanced products and platforms.
\nHe also linked that ambition to education, arguing that young people need to be prepared not only to use new technologies but to build them and compete in an economy increasingly shaped by AI.
\nFor Kosovo, where technology services and a young digital workforce already play an important economic role, that could be an important part of the country's wider AI strategy.
\nBut much of the government's vision remains at the direction-setting stage.
\nThe next test will be implementation: which AI-enabled services appear first, how citizens are informed when AI is involved, what data those systems can access and what safeguards are established when automated tools become part of official government processes.
\nThose details will determine whether eKosova 2.0 becomes simply a larger digital portal or a more fundamental redesign of how citizens interact with the state.
","featured_image_id":291,"intro_image_id":291,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":10,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":46,"ordering":0,"published_at":"2026-09-12 10:53:15","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-12 10:53:15","updated_at":"2026-09-12 10:53:15","key_takeaways":null,"faq":null,"sources":null,"category_name":"Kosovo","category_slug":"kosovo","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/c854498a1b26b92adb0b65b16f18b478.webp","featured_image_alt":"Three speakers sit before a presentation showing eKosova, launched in 2021 with 330 electronic services.","featured_image_width":1600,"featured_image_height":1066,"intro_image_path":"/api/uploads/2026/09/c854498a1b26b92adb0b65b16f18b478.webp","intro_image_width":1600,"intro_image_height":1066,"seo_title":"Kosovo Details AI Plans for eKosova 2.0","meta_description":"Kosovo reveals new eKosova 2.0 plans covering AI-enabled services, cloud infrastructure, interoperability, data and cybersecurity.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":107,"title":"Kosovo to Host AI & Tech Summit 2026 in Prishtina This September","slug":"kosovo-to-host-ai-tech-summit-2026-in-prishtina-this-september","short_description":"","content":"Kosovo will host AI & Tech Summit 2026 in Prishtina on September 3 and 4, bringing together public institutions, universities, technology companies, startups and artificial intelligence experts for two days focused on AI, digital transformation and innovation.
\nThe event will take place at AAB College, which has been building institutional cooperation around the summit with several Kosovo government ministries in recent weeks. AAB says the programme is intended to connect academia, government and the technology industry around the growing use of artificial intelligence and digital technologies.
\nAn accompanying AI & Tech Expo is also planned, where companies, startups, universities and organisations are expected to present AI-based solutions and other emerging technologies.
\nThe summit has gained a public-sector dimension through cooperation discussions between AAB and Kosovo's government.
\nAAB Rector Blerim Olluri met Lulëzon Jagxhiu, Kosovo's Minister of Digitalization and Public Administration, on July 6 to discuss the organisation of the summit.
\nTheir discussions linked the event with digital transformation in public administration, electronic government services, artificial intelligence and technological innovation.
\nJagxhiu's position is also confirmed by Kosovo's government cabinet, which lists him as Minister of Digitalization and Public Administration.
\nThe involvement of the ministry gives the conference a practical policy angle. Governments across the region are increasingly being pushed to decide where AI belongs in public administration, which services could benefit from automation and what institutional safeguards will be required as those systems become more widely used.
\nAAB has also discussed cooperation on the summit with Hajrulla Çeku, Kosovo's Minister of Education and Science.
\nA meeting between Çeku and Olluri on July 3 covered the possible co-organisation of the event, as well as the integration of artificial intelligence into higher education, development of digital skills and preparation of students for changes in the labour market.
\nKosovo government sources also identify Çeku as the country's Minister of Education and Science.
\nAccording to AAB, the summit is expected to bring together universities, international AI experts, technology companies, public institutions, international organisations and startups. Discussions are expected to cover artificial intelligence, digital transformation, education, scientific research and innovation.
\nAAB has also expanded cooperation around the event to Kosovo's Ministry of Industry, Entrepreneurship, Trade and Innovation, with the institution announcing further discussions on the summit on July 17.
\nThe event arrives as artificial intelligence moves beyond the technology sector and into government, universities and traditional industries across the Balkans.
\nFor Kosovo, that raises questions that extend beyond access to the latest generative AI tools.
\nUniversities face pressure to adapt courses and digital skills programmes. Public institutions must determine where AI can realistically improve services. Local companies and startups, meanwhile, are competing in a technology market increasingly shaped by AI platforms developed outside the region.
\nBringing those groups into the same event could give the summit relevance beyond a conventional technology conference.
\nAAB describes its ambition as positioning Kosovo as a regional centre for dialogue around artificial intelligence, digital transformation and innovation. Whether the summit can contribute to that goal will depend on the partnerships, projects and initiatives that emerge after the conference rather than on the event itself.
\nThe AI & Tech Summit 2026 will take place at AAB College in Prishtina on September 3–4, alongside the planned AI & Tech Expo.
","featured_image_id":241,"intro_image_id":241,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":10,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":34,"ordering":0,"published_at":"2026-08-15 09:22:47","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-15 09:22:47","updated_at":"2026-08-15 09:22:47","key_takeaways":null,"faq":null,"sources":null,"category_name":"Kosovo","category_slug":"kosovo","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/b3586f69da732b4dc35604b1a453fd0d.webp","featured_image_alt":"Red poster for the AI and TECH International Summit 2026, 3–4 September at AAB College in Prishtina.","featured_image_width":1000,"featured_image_height":667,"intro_image_path":"/api/uploads/2026/08/b3586f69da732b4dc35604b1a453fd0d.webp","intro_image_width":1000,"intro_image_height":667,"seo_title":"Kosovo to Host AI & Tech Summit 2026 in Prishtina","meta_description":"Kosovo will host AI & Tech Summit 2026 in Prishtina on September 3–4, bringing together government, universities, startups and AI experts.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":82,"title":"Kosovo to Make Its Debut at the International AI Olympiad","slug":"kosovo-to-make-its-debut-at-the-international-ai-olympiad","short_description":"","content":"Kosovo will participate in the International Olympiad in Artificial Intelligence for the first time, sending a team of four students to compete in Astana, Kazakhstan, from August 2 to 8.
\nThe team consists of Vesa Greiçevci, Parid Haxholli, Enesa Kiçmari and Entela Mirena, according to a report published by Bota Sot. Professor Fiton Hoxha will lead the delegation.
\nTheir participation places Kosovo among 103 countries and territories registered for this year’s competition, which has attracted more than 120 teams. The figure represents a considerable expansion for an event that held its first edition in Bulgaria only two years ago.
\nThe International Olympiad in Artificial Intelligence confirmed that participants qualify through national selection processes and compete as official teams representing their countries or territories.
\nFor Kosovo, the debut carries significance beyond the competition itself.
\nKazakhstan does not recognise Kosovo’s independence, while Kosovo is not a member of UNESCO, which has granted its patronage to this year’s Olympiad. Despite that political background, the Kosovo team is expected to participate alongside delegations from around the world.
\nThe competition is designed for secondary-school students and combines theoretical knowledge with practical work in artificial intelligence. Participants face individual tasks involving machine learning, data analysis and problem-solving, alongside a team challenge.
\nThis year’s team competition begins with an online preparation phase before moving to simulation rounds and live robot finals in Astana. The strongest teams will be asked to apply AI methods in a practical robotics setting rather than simply answer questions about the technology.
\nThe format reflects a broader change in how countries are approaching AI education. Artificial intelligence is increasingly becoming a competitive academic discipline, similar to mathematics, physics and informatics, with national selection programmes and international competitions now emerging around it.
\nFor smaller countries, these events offer more than medals. They give students access to international networks, advanced learning opportunities and contact with peers working on similar technical problems.
\nKosovo has already built experience in international mathematics and informatics competitions. Members of its AI team have previously appeared in other academic contests. Haxholli represented Kosovo at the International Mathematical Olympiad in 2025, while Mirena has competed for Kosovo at European mathematics and informatics olympiads.
\nThat background suggests the delegation is drawing from an existing community of students with strong foundations in mathematics, programming and logical reasoning.
\nHowever, participating in an international competition is only one part of developing AI talent. The longer-term question is whether schools, universities and public institutions across Kosovo can create continuous opportunities for students interested in the field.
\nMuch of the current AI education debate in the Balkans still revolves around whether students should be allowed to use tools such as ChatGPT. Competitions such as the IOAI move the discussion further, towards understanding how AI systems work, how models are evaluated and how students can build with the technology rather than simply consume it.
\nThe official organisers describe accessibility as one of the priorities for the 2026 edition. Forty participating countries and territories will receive financial assistance, more than twice the number supported in the previous competition.
\n![\"IOAI](\"/api/uploads/2026/07/90e2877587f34e14e1010e6d17828cce.webp\")
\nGoogle, Huawei, Jane Street, IMC, DeepInfra and other technology companies are listed among the international sponsors.
\nKosovo’s first appearance will therefore be both a competition and an early test of how the country can position its young talent within the growing international AI education community.
\n![\"IOAI](\"/api/uploads/2026/07/92a10b66342d4759ffbdc84d29ad23e7.webp\")
\nThe results will be decided in Astana in August. What matters afterward is whether this first delegation becomes the beginning of a permanent national pathway for students interested in artificial intelligence.
","featured_image_id":203,"intro_image_id":203,"featured_image_credit":"IOAI 2025 from their archive","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":10,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":19,"ordering":0,"published_at":"2026-07-18 11:09:17","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-18 11:09:17","updated_at":"2026-07-18 11:09:17","key_takeaways":null,"faq":null,"sources":null,"category_name":"Kosovo","category_slug":"kosovo","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/c38b4d73ddb1a9d85df1870a265b5753.webp","featured_image_alt":"Attendees gather at the 2024 International AI Olympiad (IOAI) closing ceremony in a brightly lit hall.","featured_image_width":1600,"featured_image_height":1067,"intro_image_path":"/api/uploads/2026/07/c38b4d73ddb1a9d85df1870a265b5753.webp","intro_image_width":1600,"intro_image_height":1067,"seo_title":"Kosovo to Debut at the International AI Olympiad","meta_description":"Kosovo will send four students to its first International AI Olympiad, joining 103 countries and territories in Astana this August.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":35,"title":"AI Sparks New Debate on Kosovo’s Education Future","slug":"ai-sparks-new-debate-on-kosovo-s-education-future","short_description":"","content":"Artificial intelligence is becoming one of the most important topics in Kosovo’s education debate, as students, teachers and researchers increasingly face the question of how AI should be used in learning.
\nThe discussion is no longer theoretical. AI tools are already being used by students to write, summarize, translate, research and prepare assignments. At the same time, teachers and education institutions are still trying to decide whether these tools should be treated mainly as a risk or as a new opportunity for improving learning.
\nThe rise of AI in education is changing the way knowledge is produced and evaluated. In the past, homework and essays were often used to measure whether students understood a topic. Today, a student can generate a draft, explanation or summary within seconds. This forces schools and universities to rethink how they assess real understanding.
\nFor Kosovo, this shift creates both opportunities and challenges. On one hand, artificial intelligence can help students access knowledge faster, improve writing, understand difficult concepts and explore subjects beyond the limits of traditional textbooks. On the other hand, if students use AI only to avoid independent work, it could weaken critical thinking and reduce the quality of learning.
\nThe most important issue is not whether AI should be banned, but how it should be used. Education systems that try to completely block AI may quickly fall behind, because the technology is already part of everyday life. Instead, Kosovo’s schools and universities need clear rules that explain when AI use is acceptable, how it should be cited and what kind of work must remain personal and original.
\nTeachers also need support. Many educators are being asked to respond to AI without enough training or institutional guidance. Some may see it only as a cheating tool, while others may want to use it but do not yet know how. Training teachers to work with AI could help transform the technology from a problem into a teaching resource.
\nAI can also support teachers directly. It can help prepare lesson plans, create exercises, summarize materials, adapt explanations for different student levels and generate examples for classroom discussion. Used carefully, this could reduce administrative pressure and give teachers more time for direct work with students.
\nAnother important issue for Kosovo is language. Many AI tools work better in English than in Albanian or Serbian. This means that students who use AI in Kosovo may receive different quality depending on the language they use. Better local-language digital resources will be necessary if AI is to become truly useful in the education system.
\nThe arrival of AI also raises questions about equality. Students with better internet access, better devices and paid AI tools may gain an advantage over others. If institutions do not create a fair framework, artificial intelligence could widen existing education gaps.
\nKosovo now has an opportunity to modernize education by treating AI as part of digital literacy. Students should not only learn how to use AI, but also how to question it, verify its answers and understand its limits. AI can make mistakes, invent information and reflect bias, so human judgment remains essential.
\nThe future of education in Kosovo will depend on whether institutions respond actively or allow AI use to develop randomly. If handled well, artificial intelligence can help students become more curious, independent and creative. If ignored, it may become an uncontrolled shortcut that weakens learning.
\nThe debate around AI and education shows that Kosovo is entering a new phase of digital transformation. The classroom is becoming one of the first places where society must decide how humans and intelligent machines will work together.
","featured_image_id":72,"intro_image_id":72,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":10,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":33,"ordering":0,"published_at":"2026-06-25 13:47:49","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 13:47:49","updated_at":"2026-06-25 13:47:49","key_takeaways":null,"faq":null,"sources":null,"category_name":"Kosovo","category_slug":"kosovo","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/910ec8fdb21338c911d03a79fd24e337.webp","featured_image_alt":"View from Prishtina Library","featured_image_width":1200,"featured_image_height":800,"intro_image_path":"/api/uploads/2026/06/910ec8fdb21338c911d03a79fd24e337.webp","intro_image_width":1200,"intro_image_height":800,"seo_title":"AI Sparks New Debate on Kosovo’s Education Future","meta_description":"Artificial intelligence is becoming one of the most important topics in Kosovo's education debate, as students, teachers and researchers.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":34,"title":"Kosovo Startups Gain Momentum in AI Rankings","slug":"kosovo-startups-gain-momentum-in-ai-rankings","short_description":"","content":"Kosovo’s technology ecosystem is continuing to gain visibility, with June 2026 startup rankings showing that the country has a growing base of digital companies working in areas connected to artificial intelligence, software development, data and automation.
\nAlthough Kosovo is still a small market, its startup scene has become increasingly important in the wider Western Balkans technology landscape. The country has a young population, strong interest in digital skills and a growing number of companies offering technology services to international clients.
\nThe latest startup ecosystem rankings list several Kosovo-based companies among the country’s most visible startups, including firms active in software, data and AI-related services. This shows that artificial intelligence in Kosovo is not limited to public debates or academic discussions. It is also becoming part of the private-sector innovation environment.
\nOne of the most important elements of Kosovo’s AI potential is its talent base. Many young professionals in the country work in software engineering, design, digital marketing, data analysis and remote technology services. These skills create a foundation for AI adoption because companies that already understand digital tools can move more quickly toward automation and machine-learning-based solutions.
\nAI can help Kosovo startups compete internationally. Instead of relying only on low-cost outsourcing, companies can use AI to build smarter products, improve customer support, automate workflows and analyze business data. This can move the ecosystem toward higher-value services.
\nFor small startups, AI tools can also reduce operational barriers. A young company can use artificial intelligence to generate prototypes, test ideas, create content, analyze user behavior and manage repetitive tasks. This gives smaller teams the ability to move faster and compete with larger companies.
\nHowever, the growth of AI in Kosovo’s startup sector will depend on access to investment, infrastructure and specialized knowledge. Building serious AI products requires more than using existing tools. It often requires high-quality data, technical expertise, computing resources and a clear understanding of customer problems.
\nKosovo’s startup ecosystem also needs stronger links between universities, companies and investors. AI development works best when research, talent and business needs are connected. Universities can support the ecosystem by preparing students with practical skills in machine learning, data science and responsible AI development.
\nAnother opportunity is regional cooperation. Kosovo companies can build AI products for Albanian-speaking markets, the Western Balkans and the diaspora. Local knowledge, language understanding and cultural context can become advantages if companies use them to develop specialized tools.
\nAI adoption could also help traditional sectors in Kosovo. Startups can build solutions for education, healthcare, agriculture, finance, logistics and public administration. These are areas where digital tools can solve real problems and create measurable value.
\nAt the same time, Kosovo must avoid treating AI as only a buzzword. Not every digital product is an AI product, and not every company using automation is building artificial intelligence. For the ecosystem to mature, startups need to focus on real use cases, reliable technology and practical impact.
\nThe June 2026 visibility of Kosovo startups shows that the country has the ingredients for a stronger AI ecosystem: talent, entrepreneurial energy and growing digital demand. The next step is turning that potential into companies that build original products, attract investment and solve real problems.
\nIf Kosovo continues to invest in digital skills and startup support, artificial intelligence could become one of the sectors that helps the country move beyond traditional economic limits and compete more actively in the regional technology market.
","featured_image_id":71,"intro_image_id":71,"featured_image_credit":"Orainfo","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":10,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":28,"ordering":0,"published_at":"2026-06-25 13:36:59","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 13:36:59","updated_at":"2026-06-25 13:36:59","key_takeaways":null,"faq":null,"sources":null,"category_name":"Kosovo","category_slug":"kosovo","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/6e6f8d21dded51665be1db56c40c4ed1.webp","featured_image_alt":"Five young adults collaborate around laptops beside a Kosovo Startup Foundation banner.","featured_image_width":1536,"featured_image_height":1024,"intro_image_path":"/api/uploads/2026/06/6e6f8d21dded51665be1db56c40c4ed1.webp","intro_image_width":1536,"intro_image_height":1024,"seo_title":"Kosovo Startups Gain Momentum in AI Rankings | AI Balkans","meta_description":"Kosovo's technology ecosystem is continuing to gain visibility, with June 2026 startup rankings showing that the country has a growing base.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":33,"title":"Kosovo Brings AI Into Digital Services With e-Kosova 2.0","slug":"kosovo-brings-ai-into-digital-services-with-e-kosova-2-0","short_description":"","content":"Kosovo’s next phase of digital government is expected to include artificial intelligence as part of the upgraded e-Kosova 2.0 platform, a major public-service modernization project that aims to introduce hundreds of new online services for citizens and businesses.
\nThe project is designed to expand Kosovo’s national digital services portal and make public administration faster, simpler and more connected. One of the key parts of the announcement is the planned integration of AI-driven features, which could help citizens navigate services more easily and help institutions process information more efficiently.
\nThe upgraded platform is expected to include around 400 new administrative services. This would represent a significant expansion of Kosovo’s digital government system and could reduce the need for citizens to visit offices, wait in lines or submit the same documents multiple times.
\nArtificial intelligence could support the platform in several practical ways. AI assistants could help users find the correct service, explain procedures, guide citizens through applications and answer common administrative questions. For institutions, AI could help organize requests, classify documents and support faster internal workflows.
\nThe move reflects a wider global trend: governments are no longer only putting forms online, but trying to create smarter digital services. The goal is not just digitalization, but a better user experience. Citizens expect public platforms to work more like modern digital products, with clear instructions, faster responses and fewer unnecessary steps.
\nFor Kosovo, e-Kosova 2.0 could become an important step in building trust in digital public services. If citizens can complete more procedures online, the state can reduce administrative pressure and improve access for people outside major urban centers.
\nThe project also has economic importance. Businesses often lose time dealing with paperwork, permits, certificates and official procedures. A stronger digital platform could make it easier to open businesses, update records, access documents and communicate with institutions.
\nHowever, adding artificial intelligence to public services requires caution. Government platforms handle sensitive personal and business data, so cybersecurity and privacy must be central to the system. AI tools must be reliable, transparent and carefully supervised, especially when they are used in official processes.
\nAnother challenge is accessibility. Digital services should be useful for all citizens, including older people, people with disabilities and those with limited digital skills. If AI features are too complicated or unclear, they could create frustration instead of solving problems.
\nTraining public officials will also be essential. Technology alone cannot improve administration if institutions do not change the way they work. Civil servants need to understand how the new platform functions, how AI tools support their tasks and how to protect citizen data.
\nThe success of e-Kosova 2.0 will depend on implementation. A platform with hundreds of services can only work well if those services are stable, updated and connected to real institutional processes. If online services still require physical follow-up, the benefits will be limited.
\nStill, the inclusion of artificial intelligence shows that Kosovo wants to move toward a more advanced digital government model. The country is not only expanding the number of online services, but also exploring how smarter systems can make administration more efficient.
\nIf implemented properly, e-Kosova 2.0 could become one of Kosovo’s most important digital transformation projects. It could also provide a practical example of how artificial intelligence can be used outside politics, not as a controversy, but as a tool to improve everyday services for citizens and businesses.
","featured_image_id":70,"intro_image_id":70,"featured_image_credit":"Linkedin","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":10,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":32,"ordering":0,"published_at":"2026-06-25 13:33:46","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 13:33:46","updated_at":"2026-06-25 13:33:46","key_takeaways":null,"faq":null,"sources":null,"category_name":"Kosovo","category_slug":"kosovo","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/5cf381409bf4e3fe7a58fe0ff263c5e6.webp","featured_image_alt":"A yellow neon sign reading “Kosova 2.0” glows on an indoor wall beside a framed poster.","featured_image_width":1600,"featured_image_height":1200,"intro_image_path":"/api/uploads/2026/06/5cf381409bf4e3fe7a58fe0ff263c5e6.webp","intro_image_width":1600,"intro_image_height":1200,"seo_title":"Kosovo Brings AI Into Digital Services With e-Kosova 2.0","meta_description":"Kosovo's next phase of digital government is expected to include artificial intelligence as part of the upgraded e-Kosova 2.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":23,"title":"Kosovo’s AI Landscape Faces a New Test as Synthetic Content Spreads Online","slug":"kosovo-s-ai-landscape-faces-a-new-test-as-synthetic-content-spreads-online","short_description":"","content":"PRISHTINA — Kosovo’s artificial intelligence scene is moving in two opposite directions: on one side, a growing interest in AI, digital services, and training for young people; on the other, a rising wave of AI-generated disinformation that is beginning to shape the country’s online space. For a country trying to modernize quickly, the challenge is no longer whether AI will arrive, but how to manage its impact before it gets ahead of the public conversation.
\nRecent reports show that AI-generated images and synthetic media have been circulating online in Kosovo, including content involving public figures. That has raised concerns about how easily manipulated visuals can spread during periods of public attention, especially when people are already consuming news through social platforms and short-form video. The problem is not only technical; it is also political, social, and media-related, because false material can travel faster than verification.
\nAt the same time, Kosovo is not standing still on the innovation side. Local startup activity is showing stronger interest in artificial intelligence and digital services, while youth-focused initiatives such as Techathon – Teens in AI have brought more than 70 young participants into the conversation. That suggests that the country’s AI story is not just about risk, but also about a new generation building skills, experimenting with tools, and exploring practical uses for automation and digital products.
\nAnother important development is the launch of the EU AI Act Readiness Assessment for Kosovo, which signals that the conversation is beginning to move from excitement to governance. That kind of assessment matters because it pushes institutions to think about regulation, oversight, and responsible use before AI systems become deeply embedded in public and private life. In a small market like Kosovo, where digital trust is still fragile, that early preparation could prove just as important as innovation itself.
\nFor now, Kosovo’s AI story is a warning and an opportunity at the same time. The warning is that synthetic content is already creating new risks for public trust. The opportunity is that Kosovo still has a chance to shape its AI future through education, startup growth, and better governance before the technology becomes harder to control.
","featured_image_id":58,"intro_image_id":58,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":10,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":2,"view_count":7,"ordering":0,"published_at":"2026-06-25 10:56:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 10:56:26","updated_at":"2026-06-25 10:56:43","key_takeaways":null,"faq":null,"sources":null,"category_name":"Kosovo","category_slug":"kosovo","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/8b38ee656d0ed5a6b9b303422fa0aec2.webp","featured_image_alt":"AI Fiber Data","featured_image_width":1600,"featured_image_height":946,"intro_image_path":"/api/uploads/2026/06/8b38ee656d0ed5a6b9b303422fa0aec2.webp","intro_image_width":1600,"intro_image_height":946,"seo_title":"Kosovo’s AI Landscape Faces a New Test as Synthetic Content Spreads Online","meta_description":"PRISTINA — Kosovo’s artificial intelligence scene is moving in two opposite directions: on one side, a growing interest in AI, digital…","robots":"index","canonical_url":"","og_title":"","og_description":""}]},{"category":{"id":6,"name":"Global AI","slug":"global-ai"},"posts":[{"id":175,"title":"Anthropic Launches Claude Duet, a Proactive AI Assistant","slug":"anthropic-claude-duet-proactive-ai-agent","short_description":null,"content":"Anthropic introduced Claude Duet on October 21, 2026, a new AI assistant designed to monitor connected applications, identify tasks that need attention and complete work without requiring users to remain in an active conversation.
\nUnlike conventional AI chatbots that wait for instructions, Claude Duet can initiate contact, prepare daily briefings and continue working while users are away from their computers.
\nThe product is entering a limited research preview for Claude Pro and Max subscribers in the United States, with access managed through a waitlist. It is available as a separate application for iOS, Android and the web.
\nClaude Duet is powered by Anthropic's Claude Opus 5.5 model and is designed primarily for professionals and small business owners.
\n
Each morning, the assistant prepares a briefing based on information from connected applications, including email, calendars and workplace communication tools. It identifies upcoming commitments, unfinished tasks and information that may require a decision.
\nThroughout the day, Duet checks connected accounts every few hours and contacts users when it identifies something important.
\nFor example, an assistant connected to a business email account and calendar could notice an upcoming client meeting, collect relevant information and prepare a briefing before the meeting takes place.
\nIt could also identify unpaid invoices, draft payment reminders or prepare a follow-up message after a customer interaction.
\nUsers can assign more substantial tasks, including creating financial forecasts, building dashboards, developing prototypes and generating web applications.
\nThese capabilities move Claude beyond answering questions and generating content. The assistant is intended to take responsibility for completing parts of a workflow.
\nDuet operates through a dedicated virtual computer in the cloud, equipped with its own browser.
\nThis allows the assistant to continue working even when the user's laptop or smartphone is switched off.
\nRather than requiring users to manage multiple conversations, Duet maintains a continuing interaction that incorporates information from their existing Claude account, including memory, skills and connected applications.
\nAnthropic says the assistant can work with more than 6,000 connectors and plugins available through the Claude Marketplace.
\n\nThese include services such as Gmail, Slack, QuickBooks, Stripe, HubSpot, Canva and Shopify.
\nA business owner could theoretically ask Duet to review outstanding invoices, prepare reminders, update customer records and compile a report using information from several connected services.
\nHowever, the availability of integrations does not guarantee that every workflow can be completed reliably or without additional configuration.
\nGiving an AI assistant access to business applications introduces security and accountability questions.
\nAn agent that can read correspondence, access financial records and prepare transactions has considerably more operational responsibility than a chatbot generating text.
\nAnthropic says Claude Duet requires explicit approval before performing sensitive actions, including sending messages, making payments, signing documents and deleting files.
\nCertain operations, such as trading securities or transferring money between accounts, remain prohibited even with user approval.
\nThe company also says Duet cannot directly access passwords or authentication tokens and uses additional safeguards intended to prevent sensitive information from being disclosed.
\nThese protections include a dedicated security classifier, defenses against malicious instructions embedded in external content and penetration testing of the system's infrastructure.
\nNevertheless, such safeguards do not eliminate the risks associated with autonomous agents.
\nAn assistant could misunderstand instructions, incorrectly prioritize information or produce an inaccurate document. Malicious content encountered through connected applications could also attempt to influence its behavior.
\nAnthropic acknowledges that the preview version can misinterpret conversations and make mistakes when deciding which issues require immediate attention.
\nFor businesses, this makes human review particularly important when AI-generated work involves customers, financial transactions or confidential information.
\nClaude Duet is not yet available to subscribers in North Macedonia or other Balkan markets.
\nThe initial research preview is restricted to US-based Pro and Max subscribers. Team and Enterprise plans are not included.
\nAnthropic has not announced a timetable for broader international availability.
\nDuring the preview, Duet shares the usage limits of the user's existing Claude subscription.
\nIts background checks consume those limits, meaning that additional activity by the assistant can reduce the capacity available for other Claude tasks.
\nDuet also does not continuously inspect connected applications in real time. Its periodic checks mean that important information may be detected some time after it arrives.
\nThe company describes the product as an early version, with some capabilities still under development.
\nAlthough Claude Duet is currently unavailable in the region, its introduction illustrates how AI tools are moving toward more active participation in business operations.
\nFor smaller companies, the potential appeal is straightforward. Tasks such as following up on invoices, preparing meeting materials, reviewing communications and compiling reports consume working hours even when they do not require specialist expertise.
\nDelegating parts of that work to an AI assistant could reduce administrative effort.
\nBut practical adoption requires more than connecting an account.
\nBusinesses would need to determine which information an AI agent may access, which tasks it can perform independently and which actions require human authorization.
\nThey would also need appropriate controls for personal data, confidential business records and access permissions.
\nThese considerations are particularly relevant for organizations that already depend on Microsoft 365, Google Workspace, customer management systems and cloud-based financial software.
\nThe immediate question for regional businesses is not whether they should adopt Claude Duet, which they currently cannot access, but how they would govern similar autonomous tools once they become available.
\nAnthropic has not provided a timeline for expanding Duet beyond the United States. Its performance during the research preview, particularly its reliability and ability to handle sensitive workflows, will be important factors to watch.
","featured_image_id":413,"intro_image_id":413,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":17,"ordering":0,"published_at":"2026-10-22 16:22:48","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-22 16:22:48","updated_at":"2026-10-22 16:22:48","key_takeaways":"[\"Anthropic launched Claude Duet on October 21, 2026, introducing a proactive AI assistant that monitors connected applications and works in the background.\",\"Duet can prepare briefings, identify unfinished tasks, create reports and complete multi-step workflows, while requiring approval for sensitive actions.\",\"The research preview is limited to US Pro and Max subscribers. Availability in the Balkans has not been announced.\"]","faq":"[{\"q\":\"What is Claude Duet?\",\"a\":\"Claude Duet is an AI assistant developed by Anthropic that can proactively identify tasks, work across connected applications and complete assignments in the background.\"},{\"q\":\"How is Claude Duet different from regular Claude?\",\"a\":\"Regular Claude primarily responds to user requests during conversations. Duet can check connected applications periodically, initiate contact when something needs attention and continue working on tasks without an active conversation.\"},{\"q\":\"Can Claude Duet send emails or make payments automatically?\",\"a\":\"Duet can prepare these actions, but it requires user approval before sending messages, making payments or performing other sensitive operations.\"},{\"q\":\"Is Claude Duet available in North Macedonia?\",\"a\":\"No. As of October 22, 2026, the research preview is available by waitlist only to Claude Pro and Max subscribers in the United States.\"}]","sources":"[{\"title\":\"Anthropic, Introducing Claude Duet, October 21, 2026\",\"url\":\"https://claude.com/resources/articles/claude-duet\"},{\"title\":\"Claude Duet, Official Product Page\",\"url\":\"https://claude.com/product/claude-duet\"},{\"title\":\"Claude Release Notes, October 2026\",\"url\":\"https://support.claude.com/en/articles/12138966-release-notes\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/10a217a592dfbbe51b4fda2e842053df.webp","featured_image_alt":"Claude Duet","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/10a217a592dfbbe51b4fda2e842053df.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Claude Duet Is Anthropic's New AI Agent","meta_description":"Anthropic launches Claude Duet, an AI agent that manages tasks in the background. Explore its features, security controls and availability.","robots":"index","canonical_url":"anthropic-claude-duet-proactive-ai-agent","og_title":"","og_description":""},{"id":171,"title":"DeepSeek Upgrades V4.1 Flash and Opens Search API","slug":"deepseek-upgrades-v4-1-flash-and-opens-search-api","short_description":null,"content":"DeepSeek has released an upgraded version of its V4.1 Flash artificial intelligence model and opened its proprietary web search infrastructure to developers. The October 17 announcement introduces DeepSeek-V4.1-Flash-1017, with improvements to coding and autonomous task execution, alongside two new APIs for retrieving information from the web.
\nThe updated model is already available through DeepSeek's website, mobile application and API. Existing developers can access it using the deepseek-flash model identifier without changing their integration or paying higher model rates.
The simultaneous release of web search capabilities gives developers another option for building AI assistants and applications that need current information rather than relying exclusively on knowledge stored during model training.
\nDeepSeek says V4.1 Flash 1017 completes tasks faster, produces shorter outputs and performs better across several technical evaluations than the previous version released in September.
\nThe improvements focus particularly on AI agents, systems that can use tools, execute instructions and complete sequences of tasks with limited human intervention.
\nAccording to the company's published benchmarks, the updated model achieved a score of 42.3 on Terminal-Bench 4.0, compared with 31.2 for the previous release. Its ProgramBench score increased from 20.3 to 29.1, while its DeepSWE v1.1 result rose from 74.2 to 77.0.
\nThese are company-reported results, rather than independent assessments of the new release. Performance in real development environments may differ depending on the tools, evaluation framework and complexity of the task.
\nDeepSeek has also optimized the model for agent frameworks including DeepSeek Harness, Tencent CodeBuddy and WorkBuddy. Developers using DeepSeek Harness are advised to upgrade to version 0.2.1-rc.1 or later.
\nOne area DeepSeek is emphasizing is the development of complete software projects from relatively short instructions.
\nThe company says it has used reinforcement learning with self-feedback to improve the model's ability to handle longer development workflows, including game creation and 3D modelling.
\nIn its release materials, DeepSeek demonstrated several projects developed with the model, including a browser-based racing game, a recreation of Portal 2 using the Godot game engine, a 3D model of Beijing's Forbidden City and an interactive historical scene called Valley of Ages.
\nThe examples illustrate the type of extended work the model is being trained to handle. They do not, by themselves, establish how consistently it performs across unfamiliar projects or production software environments.
\nAlongside the model upgrade, DeepSeek has launched public access to its Search and Fetch APIs.
\nThe Search API allows applications to query the web and receive structured results. The Fetch API retrieves the content of an individual webpage and returns it in Markdown format, making it easier for AI systems to process and summarize information.
\nAccording to DeepSeek, its internal search infrastructure already handles hundreds of millions of requests daily and indexes hundreds of billions of webpages.
\nThe company says its search engine supports multiple languages and uses filtering mechanisms intended to reduce search-engine spam, manipulation aimed at AI-generated answers and low-quality AI-produced material. These claims have not yet been independently established through publicly available comparative testing.
\nBoth interfaces can be used independently of the language model through authenticated API requests. They are particularly relevant to AI agents, research assistants and retrieval-augmented generation applications that need access to information published after a model's training period.
\nDeepSeek has priced both Search and Fetch at $0.75 per 1,000 successful requests.
\nCharges apply to requests that return usable results or webpage content. Failed requests, timeouts and searches without results are not billed under the published rules.
\nThe company is temporarily waiving these search charges for eligible server-side tool calls made through DeepSeek Harness, Claude Code and Codex using the DeepSeek API. Billing for those calls is scheduled to begin on October 31 at 16:00 UTC.
\nThe model itself retains its existing API pricing structure, including lower off-peak rates. Developers already using deepseek-flash automatically receive the upgraded model.
That distinction matters for development costs. Model inference continues to be billed according to token usage, while the new search tools are charged separately per successful request.
\nDeepSeek has published the V4.1 Flash 1017 model weights on Hugging Face and ModelScope, allowing developers and researchers to examine the release and explore deployments beyond the company's hosted API, subject to the applicable model terms and infrastructure requirements.
\nThe model builds on the V4.1 Flash architecture introduced in September, which combines a mixture-of-experts design with native visual understanding.
\nPublishing the updated weights gives developers a way to evaluate the model separately from DeepSeek's hosted service. However, the Search and Fetch APIs remain distinct hosted services, and releasing model weights does not mean the company's proprietary search index has been made publicly available.
\nFor organizations building coding assistants or web-connected AI tools, the October release introduces two practical changes: an upgraded model available through an existing API identifier and a separately priced route to live web information.
\nThe next test will be how the upgraded model performs under independent evaluation and whether DeepSeek's search service can consistently deliver the relevance, freshness and reliability claimed in its documentation. For developers using the temporarily free search integrations, the immediate deadline is October 31, when usage-based billing begins.
","featured_image_id":392,"intro_image_id":392,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":27,"ordering":0,"published_at":"2026-10-17 16:31:05","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-17 16:31:05","updated_at":"2026-10-17 16:31:05","key_takeaways":"[\"DeepSeek releases V4.1 Flash 1017. The October 17 upgrade improves company-reported coding and agent benchmarks while maintaining existing model API pricing.\",\"Search and Fetch APIs are now publicly available. Developers can access DeepSeek's proprietary web search infrastructure to retrieve current information and webpage content.\",\"Search pricing is $0.75 per 1,000 successful requests. Eligible server-side search calls through DeepSeek Harness, Claude Code and Codex remain temporarily free until October 31, 2026.\",\"The upgraded model weights are available. DeepSeek has published V4.1 Flash 1017 on Hugging Face and ModelScope, although its proprietary search infrastructure remains a separate hosted service.\"]","faq":"[{\"q\":\"What is DeepSeek V4.1 Flash 1017?\",\"a\":\"DeepSeek V4.1 Flash 1017 is an upgraded AI model released on October 17, 2026. It introduces improvements to coding, task execution and agent workflows, according to DeepSeek's published evaluations.\"},{\"q\":\"Is DeepSeek V4.1 Flash 1017 available now?\",\"a\":\"Yes. The model is available through DeepSeek's website, mobile application and API. Developers can access it using the existing deepseek-flash model identifier.\"},{\"q\":\"What is the DeepSeek Search API?\",\"a\":\"It is a web search interface that allows developers to retrieve structured search results from DeepSeek's proprietary search engine. A separate Fetch API retrieves webpage content in Markdown format.\"},{\"q\":\"How much does DeepSeek Search cost?\",\"a\":\"DeepSeek charges $0.75 per 1,000 successful Search API requests and the same rate for Fetch API requests. Eligible integrations through certain agent tools are temporarily exempt from these charges.\"},{\"q\":\"Is DeepSeek V4.1 Flash 1017 open source?\",\"a\":\"DeepSeek has publicly released the model weights on Hugging Face and ModelScope. Developers should review the applicable licensing terms before using or redistributing them. Its proprietary web search infrastructure is not part of that model-weight release.\"}]","sources":"[{\"title\":\"DeepSeek V4.1 Flash 1017 official release\",\"url\":\"https://api-docs.deepseek.com/news/news261017\"},{\"title\":\"DeepSeek V4.1 Flash 1017 on Hugging Face\",\"url\":\"https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/086832e75052ff4af4d3e01a309dbaec.webp","featured_image_alt":"DeepSeek","featured_image_width":1600,"featured_image_height":893,"intro_image_path":"/api/uploads/2026/10/086832e75052ff4af4d3e01a309dbaec.webp","intro_image_width":1600,"intro_image_height":893,"seo_title":"DeepSeek Upgrades V4.1 Flash and Opens Search API","meta_description":"DeepSeek releases V4.1 Flash 1017 with improved coding performance and opens Search and Fetch APIs priced at $0.75 per 1,000 requests.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":169,"title":"Mistral Releases Timestral for AI-Powered Forecasting","slug":"mistral-releases-timestral-for-ai-powered-forecasting","short_description":null,"content":"French AI company Mistral has released Timestral, a new family of artificial intelligence models designed to predict future trends from historical data, targeting applications in energy, finance, manufacturing and supply chain management.
\nAnnounced on October 15, 2026, Timestral introduces four models ranging from 5 million to 209 million parameters. Unlike general-purpose AI systems built for conversation, reasoning or content generation, these models specialise in time-series forecasting, analysing patterns in numerical data to estimate what may happen next.
\nAccording to Mistral, the largest Timestral model achieved the highest results among pretrained, zero-shot forecasting models across three major public benchmarks: GIFT-Eval, TIME and fev-bench.
\nThe company has released the model weights alongside an inference implementation, allowing developers and organisations to run the models on their own infrastructure, including standard CPUs.
\nTime-series forecasting is widely used in industries where decisions depend on understanding how values change over time.
\nElectricity providers forecast demand to balance production and consumption. Manufacturers analyse sensor readings to anticipate equipment failures. Retailers estimate future sales to manage inventory, while financial institutions use historical data to assess cash flows, trading volumes and market volatility.
\nThese tasks have traditionally relied on statistical methods or machine learning systems trained for specific datasets.
\nTimestral follows a different approach. As a time-series foundation model, it is pretrained on a broad collection of patterns and can generate predictions for previously unseen datasets without requiring task-specific training.
\nThis is known as zero-shot forecasting.
\nMistral says Timestral can analyse multiple related variables simultaneously, rather than treating each series independently. It also supports additional information that may influence future outcomes, such as known schedules, calendar events or other relevant measurements.
\nFor example, an electricity-demand model could consider historical consumption alongside temperature readings and information about upcoming holidays.
\nInstead of producing only a single prediction, Timestral generates probabilistic forecasts. These express a range of possible outcomes, giving organisations more information about uncertainty when planning operations.
\nMistral has released Timestral in four configurations.
\n| Model | \nParameters | \nMain advantage | \n
|---|---|---|
| Timestral-5M | \n5 million | \nLightweight, fast inference | \n
| Timestral-28M | \n28 million | \nLower latency with greater capacity | \n
| Timestral-123M | \n123 million | \nBalance of accuracy and efficiency | \n
| Timestral-209M | \n209 million | \nHighest reported forecasting accuracy | \n
All four models support historical input sequences of up to 32,768 time steps and generate probabilistic forecasts across 21 quantiles.
\nDespite their different sizes, they share the same underlying architecture and training approach.
\nMistral reports that the smallest model can complete a representative forecasting task in approximately 10 milliseconds on an eight-thread server CPU. The largest completes the same task in approximately 140 milliseconds.
\nThese figures come from the company's own testing and will vary depending on hardware, dataset size and forecasting requirements.
\nThe ability to operate without dedicated graphics processors could make Timestral attractive for businesses that need forecasting capabilities without investing in expensive AI computing infrastructure.
\nMistral evaluated Timestral against existing forecasting systems using three established public benchmarks.
\n
GIFT-Eval measures forecasting performance across datasets from different domains. TIME evaluates predictions across a broad collection of forecasting tasks, while fev-bench includes additional real-world forecasting scenarios and supporting variables.
\nAccording to the company's published results, Timestral-209M ranked first among pretrained and zero-shot models across all three benchmarks.
\nMistral also reports that its 28-million-parameter model outperformed the substantially larger Chronos-2 model on GIFT-Eval and TIME, while Timestral-209M achieved stronger results than TimesFM-3 and Falcon-2.0 using fewer parameters.
\nThe results suggest that model size alone does not determine forecasting quality.
\nHowever, the comparisons apply to the evaluated models and benchmark configurations. They do not establish that Timestral will outperform every specialised forecasting system or conventional statistical method in production.
\nMistral has submitted results to public leaderboards and published model information to support reproducibility. A full technical report was still forthcoming at the time of the announcement.
\nA notable aspect of Timestral is its training process.
\nMistral says most of the training data was synthetically generated rather than collected directly from real-world systems.
\nThe company created artificial time-series data containing trends, seasonal patterns, noise and relationships between variables.
\nTimestral was initially pretrained on synthetic data before undergoing a second training stage combining synthetic and real-world datasets.
\nThis approach allows developers to create large and varied training collections without depending entirely on historical records from individual companies or industries.
\nIt can also help expose models to patterns that may be uncommon in available real-world datasets.
\nNevertheless, synthetic training data cannot guarantee reliable performance when actual market conditions, industrial processes or customer behaviour differ significantly from the patterns represented during training.
\nMistral has made all four Timestral model weights available through Hugging Face under a Modified MIT licence.
\nThe company has also published a GitHub repository containing its inference implementation, released separately under Apache 2.0.
\nDevelopers can download the models and run forecasts using Python and PyTorch, either on CPUs or supported GPUs.
\nFor organisations working with commercially sensitive information, Timestral supports deployment within private infrastructure, including on-premises systems and isolated environments.
\nMistral is additionally offering model adaptation through Mistral Forge, its enterprise training platform.
\nThis enables organisations to fine-tune Timestral using proprietary datasets, adapt its architecture to industry-specific variables and update models as forecasting conditions change.
\nSuch capabilities may be particularly important in sectors where operational data cannot easily be transferred to external cloud services.
\nThe release places Mistral in a growing field of companies developing foundation models for specialised data rather than focusing exclusively on language and multimodal AI.
\nForecasting models are less visible to consumers than conversational assistants, but their performance can have direct financial and operational consequences.
\nAn inaccurate prediction of electricity demand can affect energy costs. Poor inventory forecasts can increase storage expenses or create shortages. In industrial environments, failures to anticipate changes in sensor measurements can contribute to equipment downtime.
\nThese applications require more than strong benchmark results. Forecasting systems must be tested against local datasets, evaluated under changing conditions and monitored after deployment.
\nTimestral's combination of relatively small model sizes, probabilistic predictions and private deployment options gives developers another approach to these problems.
\nFor now, the central question is how well the models perform outside public benchmarks, particularly in industries where data quality, unusual events and rapidly changing conditions can undermine historical patterns.
\nMistral says a detailed technical report will follow its initial release, providing further information about the models' training and evaluation.
","featured_image_id":385,"intro_image_id":385,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":6,"view_count":37,"ordering":0,"published_at":"2026-10-16 08:21:24","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-16 08:21:24","updated_at":"2026-10-16 08:21:24","key_takeaways":"[\"Mistral released Timestral on October 15, 2026, introducing four AI models designed specifically for time-series forecasting, ranging from 5 million to 209 million parameters.\",\"The company reports that Timestral-209M leads three major public forecasting benchmarks among pretrained, zero-shot models, while requiring fewer parameters than several competing systems.\",\"Timestral supports multivariate and probabilistic forecasting, can operate on standard CPUs and is available for private deployment using downloadable model weights under a Modified MIT licence.\",\"The models target practical applications including energy-demand prediction, financial forecasting, manufacturing operations and supply chain planning.\"]","faq":"[{\"q\":\"What is Mistral Timestral?\",\"a\":\"Timestral is a family of AI foundation models designed for time-series forecasting. It analyses historical numerical data to predict future values and provide estimates of uncertainty.\"},{\"q\":\"What makes Timestral different from ChatGPT or other language models?\",\"a\":\"Timestral is designed for forecasting numerical sequences rather than generating conversational responses, documents or code.\"},{\"q\":\"Is Timestral open source?\",\"a\":\"Mistral has released the inference code under Apache 2.0 and the downloadable model weights under a Modified MIT licence. The model-weight licence includes additional conditions.\"},{\"q\":\"Can Timestral run without a GPU?\",\"a\":\"Yes. Mistral provides CPU-compatible inference for all four models, although actual performance depends on the hardware and forecasting workload.\"},{\"q\":\"How accurate is Timestral?\",\"a\":\"Mistral reports leading results across GIFT-Eval, TIME and fev-bench among pretrained, zero-shot models. Accuracy on specific business datasets requires separate testing.\"},{\"q\":\"Where can developers download Timestral?\",\"a\":\"The model weights are available through Mistral's official Hugging Face collection, and the inference implementation is published on GitHub.\"}]","sources":"[{\"title\":\"Mistral official Timestral announcement, October 15, 2026\",\"url\":\"https://github.com/mistralai/timestral\"},{\"title\":\"Timestral model weights on Hugging Face\",\"url\":\"https://huggingface.co/collections/mistralai/timestral\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/9759ee6814b7438348be43e3b1fe3f8d.webp","featured_image_alt":"Timestral","featured_image_width":1200,"featured_image_height":716,"intro_image_path":"/api/uploads/2026/10/9759ee6814b7438348be43e3b1fe3f8d.webp","intro_image_width":1200,"intro_image_height":716,"seo_title":"Mistral Releases Timestral for AI Forecasting","meta_description":"Mistral launches Timestral, four lightweight AI forecasting models with open weights, CPU support and leading reported benchmark performance.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":168,"title":"Figma Expands Into AI-Powered Software Development","slug":"figma-expands-into-ai-powered-software-development","short_description":null,"content":"Figma is bringing working code, AI agents and backend infrastructure into its design platform, introducing a series of features that could change how teams move from product concepts to functioning applications.
\nThe company announced the updates on October 15 at Config India 2026 in Bengaluru, unveiling an expanded development workflow built around code layers, Figma Specs, integration with the Payload backend framework and tighter connections between design systems and AI coding agents.
\nThe announcements extend Figma's ambitions beyond interface design. Its platform is increasingly positioned as a place where designers, developers and product teams can collaborate on working software without moving through entirely separate design, prototyping and development environments.
\nThe timing reflects a broader shift in software development, as AI coding assistants make it easier to generate applications from written instructions and visual references.
\nOne of the most significant announcements is the expansion of code layers, a feature that allows users to bring functioning code directly onto the Figma canvas.
\nInstead of working exclusively with static screens or simulated interactions, designers can now create and modify interactive prototypes backed by executable code.
\nFigma supports several ways to begin. Users can convert existing design frames into code-backed experiences, import a connected GitHub repository, capture a live website through its Chrome extension or describe an application to an AI agent.
\nCode layers were introduced earlier in 2026 and are now entering an open beta across Figma's plans.
\nThe distinction matters because traditional design prototypes often approximate how software should behave without implementing its actual functionality. Code-backed prototypes allow teams to test interactions and make changes closer to the development process.
\nHowever, this does not mean Figma can automatically turn any design into a production-ready application. The company says support for increasingly complex codebases remains an area of future development.
\nThe company also introduced Figma Specs, a Markdown-based environment for capturing product requirements, technical decisions and design context.
\nSpecs can exist as standalone documents or alongside visual designs on the canvas. Teams can connect specifications to existing Figma elements and use AI agents to help develop or revise them.
\nThe feature addresses a growing challenge in AI-assisted software development: providing coding agents with reliable information about what a product is supposed to do.
\nAn AI agent may generate a functional interface from a prompt, but the result can diverge from established design systems, technical constraints or product requirements if those instructions are incomplete.
\nWith Specs, Figma wants written requirements, visual references and technical context to remain connected throughout development.
\nThe system also supports importing existing Markdown documents, allowing teams to bring documentation into the design environment rather than recreating it.
\nSpecs is entering open beta on eligible paid plans, with access rolling out gradually.
\nAnother major development is the integration of Payload into Figma Make.
\nPayload is an open-source content management system and application framework whose team joined Figma in 2025.
\nWith the integration, Figma's AI-assisted application builder can generate a backend that includes a database, content management, file storage, authentication, access controls and APIs.
\nThese are essential components of many real-world applications.
\nA business application, for example, may require user accounts, stored records, administrative tools and secure access to information. Visual design tools alone cannot provide those capabilities.
\nFigma says Payload can now supply that infrastructure directly within its development workflow, with hosted services available through Figma and self-hosting remaining an option for developers.
\nThe integration is rolling out across Figma Make plans.
\nWhile this moves Figma closer to full-stack application development, production deployment still requires consideration of security, reliability, maintenance and operational costs. Generating a backend is not equivalent to independently verifying that an application is ready for public use.
\nFigma is also expanding the information available to AI agents working with design systems.
\nNew capabilities allow organizations to attach guidelines to published design libraries and connect code sources from GitHub repositories or npm packages.
\nThe objective is to help agents generate components that follow existing patterns rather than repeatedly inventing new ones.
\nFigma has also announced improvements to its Model Context Protocol (MCP) integrations, extending how external AI tools interact with the platform.
\nThis could be particularly relevant for organizations already using coding agents alongside established design systems. The more accurately an agent can interpret existing components, implementation rules and technical context, the less manual correction may be required.
\nSome capabilities remain in limited testing. Code-source integration is entering closed beta, while centralized design-system management is currently a preview.
\nThe announcements arrive as the distinction between design applications, AI coding tools and low-code development platforms becomes less clear.
\nAI-assisted development has made it possible to generate interfaces and application prototypes with relatively little conventional programming. Tools built around this approach are increasingly competing for users who previously depended on separate design and development processes.
\nFigma is taking a different route from standalone prompt-to-app services.
\nRather than treating design as an input that is handed off to a coding system, it is attempting to keep the product's visual design, technical specifications, working code and collaboration history within a connected environment.
\nThat could help teams maintain consistency as applications evolve. It could also make Figma a more direct competitor to AI application builders and parts of the traditional software development toolchain.
\nThere are important limits to that comparison. Professional software engineering involves testing, infrastructure management, security reviews, integration work and ongoing maintenance. Figma's announcements do not remove those responsibilities.
\nNor has the company demonstrated that its new workflow is more reliable or economical than established alternatives across complex production projects.
\nFigma's latest releases suggest that the company sees AI-assisted software development as a central part of its future, rather than an additional feature for designers.
\nThe most consequential change may be the connection between information that has historically lived in different places: product documentation, interface designs, source code and backend systems.
\nFor development teams, the practical question is whether bringing these elements together produces better software with fewer handoffs and less rework.
\nSeveral of the newly announced capabilities are still in beta, and the company is introducing usage-based AI credits for code-layer tasks after a promotional period ending November 4, 2026.
\nThe next measure of Figma's progress will be how these tools perform in real projects, particularly when teams move beyond prototypes into software that must be maintained, secured and updated over time.
","featured_image_id":383,"intro_image_id":383,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":6,"view_count":20,"ordering":0,"published_at":"2026-10-16 08:06:51","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-16 08:06:51","updated_at":"2026-10-16 08:06:51","key_takeaways":"[\"Figma has introduced expanded AI-assisted development capabilities, bringing working code, technical specifications and backend infrastructure closer to its design environment.\",\"Code layers and Figma Specs are entering open beta, while Payload integration adds database, authentication, content management and API functionality to Figma Make.\",\"The announcements position Figma more directly against AI application builders, although complex production software still requires engineering, security and maintenance work.\"]","faq":"[{\"q\":\"What did Figma announce in October 2026?\",\"a\":\"Figma announced code layers entering open beta, the introduction of Figma Specs, native Payload backend integration, additional design-system capabilities and expanded AI-agent connectivity at Config India 2026.\"},{\"q\":\"Can Figma now build complete applications?\",\"a\":\"Figma Make can generate applications with working backend functionality through Payload. However, production readiness depends on the application's complexity, security, testing and deployment requirements.\"},{\"q\":\"What are Figma code layers?\",\"a\":\"Code layers allow developers and designers to create, import and modify interactive prototypes backed by real code directly on the Figma canvas.\"},{\"q\":\"What is Figma Specs?\",\"a\":\"Figma Specs is a Markdown-based environment that connects product requirements, technical documentation and visual design context for teams and AI agents.\"},{\"q\":\"Are the new Figma features free?\",\"a\":\"Availability varies. Code layers are rolling out across plans, with a temporary promotional period for generating and editing code on paid plans through November 4, 2026. Specs is available on eligible paid and institutional plans. Usage-based AI credits apply to certain features.\"},{\"q\":\"Does Figma replace traditional development tools?\",\"a\":\"No. The new capabilities reduce some handoffs between design and development, but they do not eliminate the need for programming expertise, software testing, infrastructure management or security reviews.\"}]","sources":"[{\"title\":\"Figma's full announcement by CEO Dylan Field\",\"url\":\"https://www.figma.com/blog/build-by-design/\"},{\"title\":\"Introducing Figma Specs\",\"url\":\"https://www.figma.com/blog/websites-with-payload-in-figma/\"},{\"title\":\"Payload integration and backend capabilities\",\"url\":\"https://www.figma.com/blog/websites-with-payload-in-figma/\"},{\"title\":\"Official availability, limitations and pricing details\",\"url\":\"https://help.figma.com/hc/en-us/articles/43663368947863-What-s-new-from-Config-India-2026\"},{\"title\":\"Figma's official release notes\",\"url\":\"https://www.figma.com/release-notes/\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/a03f2c4b575d120113ca3b050ebfc4da.webp","featured_image_alt":"Figma AI","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/a03f2c4b575d120113ca3b050ebfc4da.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Figma Expands Into AI-Powered Software Development","meta_description":"Figma introduces code layers, Specs and Payload integration, bringing AI agents and backend development into its design platform.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":167,"title":"Google Releases Gemini 4 to Challenge AI Rivals","slug":"google-releases-gemini-4-to-challenge-ai-rivals","short_description":null,"content":"Google has expanded access to Gemini 4 Argon, its newest frontier artificial intelligence model, bringing advanced reasoning, coding and professional task automation capabilities to developers, businesses and paying consumers.
\nThe wider release on October 13, 2026, follows an initial announcement on September 30, when Google limited access to selected cybersecurity professionals. Gemini 4 Argon is now available through Google's developer platforms, enterprise services, Gemini app and AI-powered Search features.
\nThe launch puts Google in more direct competition with OpenAI and Anthropic for customers building software, automating business processes and deploying AI systems capable of completing complex tasks.
\nAccording to Google DeepMind, the new model is designed to maintain reasoning across extended workflows rather than simply produce answers to individual questions. Its targeted applications include software engineering, financial analysis, legal research, cybersecurity and scientific work.
\nGoogle has published benchmark comparisons showing Gemini 4 Argon outperforming several competing models in selected professional tasks. The results are substantial, although they do not establish a universal lead.
\n
On the Vals Index, which evaluates performance across knowledge-intensive professional work, Gemini 4 Argon achieved 68.9%, compared with 63.1% for GPT-6 Astra, 65.8% for Claude Fable 5.1 and 67.0% for Claude Opus 5.5.
\nThe model also performed strongly on DeepSWE v1.1, a software engineering benchmark, scoring 77.9%. GPT-6 Astra achieved 74.1%, while Claude Opus 5.5 scored 74.2%.
\nThe ranking changes on other tests.
\nOn FrontierSWE v2, GPT-6 Astra led with 65.5%, followed by Claude Opus 5.5 at 62.3%. Gemini 4 Argon scored 55.0%. Anthropic's Claude Opus 5.5 also retained an advantage on Terminal-bench 4.0, scoring 66.4% against Gemini's 57.4%.
\nScientific task evaluations produced similarly mixed results. GPT-6 Astra outperformed Gemini on Terminal-Bench Science 0.1, while Gemini led on LABBench 2 and RiemannBench.
\nThese figures come from Google's published model card and associated evaluation methodology. They provide useful comparisons under specific testing conditions, but should not be interpreted as independent confirmation that Gemini performs better across every application.
\nFor developers and enterprise customers, the practical question is which model performs reliably on their particular tasks, at an acceptable speed and cost.
\nGemini 4 Argon supports text, images, audio and video inputs within a combined context window of up to one million tokens.
\n
That capacity allows applications to process extensive documents, software repositories and other large collections of information. It does not, by itself, guarantee that the model will correctly interpret every detail in a long input.
\nGoogle is emphasizing Argon's ability to handle multi-step activities, including code migration, software maintenance and technical investigation.
\nThe model uses a sparse mixture-of-experts architecture and was trained using Google's Tensor Processing Units. Its design is intended to improve performance on difficult reasoning tasks without requiring every component of the model to operate on each request.
\nGoogle is also integrating the model into its broader ecosystem, including its Antigravity development platform and Gemini's task-oriented features.
\nFor software developers, this expands the possibility of assigning larger engineering activities to AI systems rather than requesting isolated code snippets. For organizations, it creates additional options for automating research, document analysis and business processes.
\nHowever, increased autonomy introduces new operational risks. Systems that can execute multiple actions require careful permissions, monitoring and verification, particularly when handling sensitive information or modifying production software.
\nPricing is another important part of the launch.
\nGemini 4 Argon is available at an introductory API rate of $2 per million input tokens and $10 per million output tokens until November 30, 2026.
\nGoogle says standard pricing will increase to $4 per million input tokens and $20 per million output tokens after that date. Cached input tokens receive a 95% discount during the introductory period.
\nThe temporary reduction gives developers an incentive to experiment with Argon and compare its performance against systems already used in their applications.
\nThe eventual cost will depend on the complexity of individual tasks, the amount of generated output, repeated model calls and other operational requirements.
\nA lower advertised token price does not necessarily translate into a cheaper completed workflow. An AI system that requires more processing time, repeated attempts or extensive verification may cost more in practice.
\nGoogle's control over its own AI infrastructure, developer services and consumer distribution channels nevertheless gives the company considerable influence over how the model reaches customers.
\nGoogle says Gemini 4 Argon is available globally to developers and enterprise customers through the Gemini API, Google AI Studio and relevant business services.
\nConsumer access is more restricted. The model is available to Google AI Pro and Ultra subscribers through the Gemini app and AI Mode in Search.
\nGoogle AI Ultra subscribers can also use Argon through the Antigravity development platform.
\nDevelopers using the API require paid access, while availability and usage limits depend on the service and subscription tier.
\nThe release is therefore broader than the initial restricted rollout, but it does not amount to unrestricted free access for all users.
\nGoogle's model card acknowledges that Gemini 4 Argon can still produce incorrect information and may experience slow responses or timeouts. Its stated knowledge cutoff is March 2026.
\nThe company also identifies more serious risks associated with the model's advanced capabilities.
\nAccording to its published safety assessment, Google could not rule out that Argon had reached certain critical capability thresholds relating to biological information and offensive cybersecurity. It also identified concerns involving situational awareness and the potential to circumvent oversight.
\nGoogle says it has implemented safeguards intended to reduce those risks before wider deployment.
\nThat distinction matters. The existence of safety controls does not eliminate the underlying capabilities or guarantee that every possible misuse has been prevented.
\nGoogle's decision to initially provide restricted access to vetted cybersecurity defenders illustrates the difficulties AI developers face as increasingly powerful systems become commercially available.
\nGemini 4 Argon's release strengthens Google's position in a market where model providers increasingly compete on their ability to complete useful work.
\nOpenAI, Anthropic and Google are developing systems that can write and maintain software, interact with tools, analyze large volumes of information and carry out activities across multiple applications.
\nThe competition is no longer determined solely by which model provides the best response to a question.
\nCustomers must also consider reliability, integration, security controls, execution speed and the total cost of completing a task.
\nGoogle enters this stage with substantial advantages in distribution. Its AI models can be integrated into products already used by consumers, developers and enterprises, from Search and Gemini to cloud infrastructure and software development tools.
\nIts latest benchmark results suggest that Gemini 4 Argon is competitive with leading models in several demanding areas. They also show that OpenAI and Anthropic retain advantages on other evaluations.
\nThe next test will come from sustained real-world use, independent evaluations and the response of competing model providers.
\nFor now, Google's wider release gives developers and businesses another serious option in a market where leadership increasingly depends on practical results rather than a single benchmark ranking.
","featured_image_id":379,"intro_image_id":379,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":6,"view_count":29,"ordering":0,"published_at":"2026-10-14 16:57:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-14 16:57:49","updated_at":"2026-10-14 17:03:59","key_takeaways":"[\"Gemini 4 Argon became broadly available on October 13, 2026, following its initial restricted announcement on September 30.\",\"Google's benchmarks place Gemini 4 ahead of GPT-6 Astra and several Claude models in professional knowledge work and selected coding tasks. Competing models remain ahead in other evaluations.\",\"Developers pay $2 per million input tokens and $10 per million output tokens through November 30, before standard pricing doubles.\",\"Google's own assessment identifies risks involving cybersecurity, biological information and model oversight, alongside safeguards intended to reduce them.\"]","faq":"[{\"q\":\"When was Google Gemini 4 released?\",\"a\":\"Google first announced Gemini 4 Argon on September 30, 2026, with restricted access. Broader availability began on October 13, 2026.\"},{\"q\":\"Is Gemini 4 better than GPT-6 and Claude?\",\"a\":\"Gemini 4 Argon leads several benchmarks published by Google, including the Vals Index and DeepSWE v1.1. However, GPT-6 Astra and Claude Opus 5.5 outperform it on other tests. No single model leads across every category.\"},{\"q\":\"Is Gemini 4 available for free?\",\"a\":\"The initial rollout requires eligible paid Google AI subscriptions or paid API access. Google has not announced unrestricted free access to Argon for all users.\"},{\"q\":\"How much does Gemini 4 cost?\",\"a\":\"Introductory API pricing is $2 per million input tokens and $10 per million output tokens through November 30, 2026. Standard pricing then increases to $4 and $20 respectively.\"},{\"q\":\"What is Gemini 4's context window?\",\"a\":\"Gemini 4 Argon supports a combined input-output context window of up to one million tokens, including text, images, audio and video inputs.\"},{\"q\":\"Can Gemini 4 make mistakes?\",\"a\":\"Yes. Google's model card acknowledges hallucinations, occasional slow responses and timeouts. Its stated knowledge cutoff is March 2026.\"}]","sources":"[{\"title\":\"Google's official launch announcement\",\"url\":\"https://blog.google/intl/en-in/products/introducing-gemini-4-argon/\"},{\"title\":\"Google DeepMind Gemini 4 Argon model card\",\"url\":\"https://deepmind.google/models/model-cards/gemini-4-argon/\"},{\"title\":\"Google's Gemini 4 safety and security report\",\"url\":\"https://blog.google/innovation-and-ai/technology/safety-security/gemini-4-model-safety-security/\"},{\"title\":\"Original Gemini 4 announcement\",\"url\":\"https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/b5345c91845d4644e95cae0ed5c0d2ef.webp","featured_image_alt":"Google Argon","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/b5345c91845d4644e95cae0ed5c0d2ef.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Google Releases Gemini 4 to Challenge AI Rivals","meta_description":"Google releases Gemini 4 Argon with advanced coding, reasoning and agent capabilities. See its benchmarks against GPT-6 and Claude, pricing and access.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":164,"title":"UK Uses AI to Find 1,352 Cybersecurity Flaws","slug":"uk-uses-ai-to-find-1-352-cybersecurity-flaws","short_description":null,"content":"The United Kingdom is expanding its use of artificial intelligence to protect government systems after two pilot programmes identified and addressed 1,352 cybersecurity vulnerabilities across 548 code repositories in 18 government departments.
\nThe findings were announced on October 12, 2026, as British authorities introduced further measures to strengthen national cyber defences and prepare for security threats involving increasingly autonomous AI systems.
\nThe initiative represents a practical application of advanced AI in government cybersecurity, demonstrating how AI models can help identify software weaknesses while raising new questions about the security of autonomous systems themselves.
\nThe UK's Government Cyber Unit has been testing frontier AI models to identify vulnerabilities in public-sector software and improve the detection of advanced cyber threats.
\nFollowing the results of two pilot programmes, the government announced plans to extend the use of AI-assisted vulnerability detection, particularly across systems considered critical to national security and public services.
\nAn earlier government case study provides insight into how the technology was deployed.
\nOver one month, technical teams worked across nine government organisations, using AI models to examine public code repositories. The exercise produced 407 findings, including serious vulnerabilities involving authentication bypass, potential data exposure and remote code execution.
\nAccording to the government, all critical vulnerabilities identified during that exercise were remediated, with no evidence that they had been exploited.
\nThe reported cost of AI model usage was approximately £13,000.
\nAmong the findings was a vulnerability involving an automated GitHub workflow used by a government digital service. A specially constructed comment on a pull request could have triggered unauthorised code execution, potentially exposing credentials and other sensitive information.
\nThe example demonstrates how AI-assisted analysis can uncover security weaknesses that arise from interactions between different software components.
\nHowever, the government's overall figure of 1,352 vulnerabilities does not come with a complete public breakdown of the severity of every finding or the validation methods used across both pilot programmes.
\n\nThe government's announcement also addresses a different category of cybersecurity risk: autonomous AI agents.
\nUnlike conventional AI assistants that primarily respond to requests, autonomous agents can perform sequences of actions, access connected systems and execute tasks with limited human intervention.
\nThese capabilities introduce new operational risks.
\nAn AI agent with access to sensitive systems could act outside its intended instructions, mishandle confidential information or perform actions with unintended consequences.
\nBritish authorities are working with the National Cyber Security Centre and the AI Security Institute to develop an approach for responding to serious incidents involving such systems.
\nThe work will examine how organisations can identify, investigate and contain incidents where autonomous AI behaviour contributes to security failures.
\nExisting cybersecurity procedures are largely designed around malicious actors, compromised infrastructure and conventional software vulnerabilities.
\nAutonomous systems introduce additional considerations, including the permissions granted to agents, the instructions they follow and the mechanisms available to interrupt their actions.
\nThe government has not yet published a completed incident-response framework.
\nThe UK is also examining whether its Cyber Security and Resilience Bill should more explicitly address risks associated with advanced artificial intelligence.
\nPotential measures include stronger reporting requirements for serious AI incidents and near misses, improved information sharing and additional protections for essential services.
\nHealthcare, energy and water infrastructure are among the sectors where disruptions to digital systems could have significant consequences.
\nThese proposals remain under consideration and should not be confused with requirements already established in law.
\nSeparately, the UK's AI Security Institute continues to evaluate advanced AI models, including through testing arrangements that provide access to models before their public release.
\nIndependent evaluation has become an increasingly important part of the government's approach to AI safety, particularly as models gain capabilities that allow them to interact directly with software systems and external tools.
\nThe government's findings demonstrate the potential value of using advanced AI to inspect large software repositories and identify weaknesses that may otherwise remain undetected.
\nThey also illustrate an important distinction between discovering a vulnerability and improving the security of an entire organisation.
\nAI-generated findings still require technical validation. Security teams must determine whether identified weaknesses are exploitable, assess their severity and implement appropriate fixes.
\nA large number of reported vulnerabilities does not necessarily indicate a proportional improvement in cybersecurity. The significance depends on the quality of the findings, the risks involved and whether the problems are successfully resolved.
\nThe earlier UK pilot provides useful evidence of practical results, including documented vulnerabilities and remediation. The broader programme will require similarly transparent reporting to establish whether AI-assisted detection can consistently outperform or complement existing security methods.
\nThe British government is moving beyond limited experiments by extending AI-assisted vulnerability detection and developing new procedures for security incidents involving autonomous systems.
\nDetails of the wider programme's budget, deployment timetable and performance targets have not yet been fully disclosed.
\nThe next measure of success will be whether AI-assisted security assessments produce lasting improvements across government infrastructure, including faster remediation and reduced exposure to serious vulnerabilities.
\nFor now, the UK's findings offer a concrete example of AI being deployed to address existing cybersecurity problems, even as governments prepare for a new generation of risks created by increasingly capable autonomous systems.
","featured_image_id":371,"intro_image_id":371,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":52,"ordering":0,"published_at":"2026-10-12 20:45:19","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-12 20:45:19","updated_at":"2026-10-12 20:45:19","key_takeaways":"[\"1,352 vulnerabilities identified: Two UK government AI cybersecurity pilots identified and addressed vulnerabilities across 548 code repositories in 18 departments.\",\"AI agents introduce new security challenges: British authorities are developing procedures for responding to incidents involving autonomous AI systems.\",\"Cybersecurity rules could expand: The UK is considering additional safeguards under its Cyber Security and Resilience Bill, including improved incident reporting.\"]","faq":"[{\"q\":\"How many cybersecurity vulnerabilities did the UK identify using AI?\",\"a\":\"The UK government reported that two pilot programmes identified and addressed 1,352 vulnerabilities across 548 code repositories in 18 government departments.\"},{\"q\":\"How is artificial intelligence being used in UK cybersecurity?\",\"a\":\"Advanced AI models are being used to analyse software repositories, identify potential security weaknesses and support technical teams investigating vulnerabilities.\"},{\"q\":\"What are the cybersecurity risks of autonomous AI agents?\",\"a\":\"Autonomous agents can interact with external systems and perform tasks with limited human supervision. If improperly configured or controlled, they may access sensitive information, execute unintended actions or create new security vulnerabilities.\"},{\"q\":\"Is the UK introducing new laws to regulate AI cybersecurity risks?\",\"a\":\"The government is considering additional measures under its Cyber Security and Resilience Bill. Possible changes include improved reporting of serious AI incidents and stronger protections for essential services. These proposals are not yet enacted.\"}]","sources":"[{\"title\":\"UK government announcement, October 12, 2026\",\"url\":\"https://www.gov.uk/government/news/britain-steps-up-national-defences-against-growing-ai-risks\"},{\"title\":\"UK Parliament, ministerial statement\",\"url\":\"https://questions-statements.parliament.uk/written-statements/detail/2026-10-12/hlws365\"},{\"title\":\"Government case study on AI-powered cyber defence\",\"url\":\"https://www.gov.uk/government/case-studies/when-ai-leaves-the-lab-testing-frontier-models-in-government-cyber-defence\"},{\"title\":\"- Government guidance on AI and vulnerability risks\",\"url\":\"https://www.gov.uk/guidance/ai-open-code-and-vulnerability-risk-in-the-public-sector\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/7abc2ec47b0ba09b1fd2f479fd354b2e.webp","featured_image_alt":"National Cyber Security Centre","featured_image_width":1600,"featured_image_height":1067,"intro_image_path":"/api/uploads/2026/10/7abc2ec47b0ba09b1fd2f479fd354b2e.webp","intro_image_width":1600,"intro_image_height":1067,"seo_title":"UK Uses AI to Find 1,352 Cybersecurity Flaws","meta_description":"Britain expands AI-powered cyber defence after identifying 1,352 vulnerabilities across government systems and plans new protections against AI agents.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":163,"title":"ArtCraft Is Challenging the Software Giants","slug":"artcraft-is-challenging-the-software-giants","short_description":null,"content":"A developer has used artificial intelligence to build free alternatives to some of the world's most established software applications. After targeting Adobe's creative tools, the project has expanded to Microsoft Word, Excel and PowerPoint.
\nThe applications are far from finished. But the development raises a question that goes beyond software pricing: what happens when customers who have spent years depending on a company's products discover that building alternatives is becoming easier?
\nBrandon Thomas, an experienced software developer behind the ArtCraft project, has attracted attention for creating open-source alternatives to applications such as Adobe Photoshop, Illustrator and Premiere Pro.
\nUsing Anthropic's Claude to accelerate development, Thomas has helped create a growing collection of applications written in Rust, designed to reproduce familiar professional workflows without proprietary subscription requirements.
\nThe project has now expanded to Microsoft Office with three applications:
\nWordCraft, an alternative to Microsoft Word.
\nGridCraft, a spreadsheet application modelled on Excel.
\nDeckCraft, a presentation tool inspired by PowerPoint.
\nThe source code is publicly available on GitHub. ArtCraft's website currently lists 12 applications spanning image editing, video production, documents, spreadsheets, presentations, audio and computer-aided design.
\n
Thomas is not someone who discovered programming through a few AI prompts. Reporting by Tom's Hardware identifies him as a developer with approximately 15 years of software engineering experience. That distinction matters. AI has accelerated the development process, but the project should not be mistaken for evidence that engineering experience has become unnecessary.
\nThe applications are also designed to operate locally, with support for several operating systems. Their architecture includes command-line controls and Model Context Protocol (MCP) interfaces, allowing AI agents to interact directly with the software.
\nThat last feature deserves attention. ArtCraft is trying to reproduce familiar applications while making them accessible to a different kind of user: software agents capable of carrying out tasks on someone's behalf.
\nAI Balkans also tested a selection of the ArtCraft applications.
\nOur initial experience was remarkably positive. The applications we tried worked smoothly, demonstrating that these projects are more than attractive interface concepts or screenshots circulating on social media.
\nThis was an initial hands-on assessment, not an exhaustive technical evaluation. We did not establish full compatibility with Adobe or Microsoft products, nor did we perform a comprehensive assessment of security, accessibility, performance under heavy workloads or professional reliability.
\nThose distinctions are important.
\nArtCraft's own documentation acknowledges substantial limitations. WordCraft's maintainers estimate that the application is approximately 60% of the way towards fully replacing Word for real work. GridCraft is estimated at around 50%, while DeckCraft is approximately 55% ready by the project's own assessments.
\nThese are developer estimates rather than independently verified benchmarks.
\nA presentation application that can create slides is not necessarily ready to manage a complex corporate presentation. A spreadsheet that supports formulas still needs to preserve calculations, formatting and large workbooks reliably. Professional image editing demands accurate file handling, colour management and consistent results.
\nArtCraft is still alpha software, and its developers say so.
\nYet there is a substantial difference between an unfinished project that barely functions and one that already performs useful work while development continues.
\nOur experience suggests these applications deserve serious attention, even if recommending them as complete replacements would be premature.
\nFor years, companies such as Adobe and Microsoft have operated from positions of considerable market strength.
\nTheir products are deeply embedded in education, business, government and creative industries. Files, training programmes, established workflows and collaboration practices make switching difficult, even when users become dissatisfied.
\nThis creates a degree of protection that newer software developers rarely enjoy.
\nIt also creates a risk.
\nIn 2025, Adobe changed its Creative Cloud individual plans, introducing Creative Cloud Pro and revised pricing in affected markets. Microsoft subsequently announced changes to commercial Microsoft 365 subscription prices, effective July 2026, alongside additional capabilities.
\n\nBoth companies presented their changes as part of continuing investment in product development and customer value. There are legitimate costs associated with maintaining complex software, providing security updates, operating cloud infrastructure and supporting enterprise customers.
\nNevertheless, customers judge those decisions according to their own needs and budgets.
\nAdobe's subscription practices have also faced regulatory scrutiny. In March 2026, the company announced a settlement with the US Department of Justice concerning allegations about subscription disclosures and cancellation practices. Adobe denied wrongdoing while agreeing to a $75 million payment and $75 million in services for eligible customers.
\nNone of this establishes that ArtCraft users have abandoned Adobe or Microsoft in significant numbers. There is no reliable evidence yet of such a migration.
\nWhat it does establish is the commercial environment in which alternatives are emerging.
\nFor an established software company, changing a subscription policy may be relatively straightforward. Recovering a customer's confidence can be much harder.
\nAnd if dissatisfaction encourages developers to create alternatives, reversing the original decision does not make those alternatives disappear.
\nTheir source code remains available. Other developers can contribute, modify the applications or build competing projects. Users who have begun testing different software may discover that they no longer need every feature of their previous subscriptions.
\nThe competitive environment has already changed, even if the market leader retains most of its customers.
\nOpen-source alternatives to commercial software are not new.
\nLibreOffice has provided a mature alternative to Microsoft Office for years. GIMP, Inkscape, Blender and numerous other projects demonstrate that powerful software does not necessarily require the same commercial model as proprietary applications.
\nArtCraft did not invent this approach.
\nWhat makes the current development interesting is the speed at which AI-assisted software engineering can help experienced developers assemble, test and improve substantial applications.
\nTasks that previously required considerable manual programming can now be accelerated through coding agents. A developer can delegate parts of implementation, testing and documentation while concentrating on architecture, product direction and unresolved technical problems.
\nThis does not eliminate the difficult parts of software engineering. Generated code still needs inspection, security review, integration and maintenance. The larger an application becomes, the more consequential those responsibilities are.
\nIt does, however, change the economics of experimentation.
\nA small team can now attempt projects that might previously have required significantly more development resources. Even when most experiments fail, the number of potential competitors can increase.
\nFor established companies, the challenge is no longer limited to other large corporations with comparable engineering budgets.
\nIt also includes experienced independent developers, open-source communities and small teams equipped with increasingly capable AI development tools.
\nThe advantage of scale remains substantial. But the cost of attempting to challenge that scale is changing.
\nFor freelancers, creative studios, startups and smaller organisations across the Balkans, software subscriptions represent recurring operating costs.
\nA free alternative that handles everyday tasks reliably could make a practical financial difference. Locally running applications may also appeal to organisations seeking greater control over their files and working environments.
\nThe opportunity comes with important qualifications.
\nOrganisations exchanging complex Word documents, Excel workbooks, PowerPoint presentations or Adobe project files need dependable compatibility. Businesses handling confidential material need confidence in software security and maintenance. Local language support, including Macedonian and Albanian text handling, should be tested rather than assumed.
\nThe sensible approach is to experiment with alternatives on non-critical work before considering any wider migration.
\nFor the region's developers, the larger lesson may be more important than the applications themselves. AI-assisted engineering creates opportunities to develop specialised software for smaller markets, local languages and workflows that major international vendors have little commercial incentive to prioritise.
\nThe opportunity is not simply to reproduce another company's interface. It is to build useful software around needs that remain poorly served.
\nAdobe and Microsoft have advantages that ArtCraft cannot reproduce quickly: decades of development, established customer relationships, technical support, extensive integrations and mature enterprise infrastructure.
\nThere are also unresolved questions about long-term maintenance, code quality, file compatibility and the intellectual-property boundaries of software designed to resemble established commercial applications. ArtCraft describes its work as clean-room reimplementation, but that description alone does not settle every potential legal question.
\nThe project may grow into a sustainable alternative. Some applications may progress much further than others. Development could also slow as the difficulty of maintaining a dozen complex products becomes apparent.
\nThe next meaningful evidence will come from sustained development, independent testing, improvements in file compatibility and actual adoption, rather than GitHub attention alone.
\nStill, ArtCraft offers an important lesson to the software industry.
\nEstablished companies can change prices, revise policies and introduce new features. Those decisions remain within their control.
\nWhat they cannot fully control is what their customers and competitors learn to build in response.
\nAnd once those alternatives exist, winning users back may require more than restoring the conditions that persuaded them to look elsewhere.
","featured_image_id":368,"intro_image_id":368,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":8,"view_count":53,"ordering":0,"published_at":"2026-10-12 15:51:22","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-12 15:51:22","updated_at":"2026-10-12 15:51:22","key_takeaways":"[\"ArtCraft is challenging established software business models. Brandon Thomas and contributors have developed free, open-source alternatives to Adobe and Microsoft applications with substantial AI assistance.\",\"AI Balkans' initial testing was positive, but the apps are unfinished. The project shows functional promise while acknowledging major remaining compatibility and reliability work.\",\"Changing a business policy cannot automatically restore lost trust. As AI reduces some software development barriers, dissatisfied users and independent developers have more opportunities to explore alternatives.\"]","faq":"[{\"q\":\"Who created ArtCraft's Adobe and Microsoft alternatives?\",\"a\":\"Brandon Thomas is the developer most closely associated with the project. He has used Anthropic's Claude in developing the applications, while public repositories allow contributions from others.\"},{\"q\":\"Are WordCraft, GridCraft and DeckCraft free?\",\"a\":\"Yes. They are publicly available as open-source software through ArtCraft's GitHub repositories, with permissive licensing.\"},{\"q\":\"Can these applications replace Adobe Creative Cloud or Microsoft 365?\",\"a\":\"Not comprehensively at present. They remain early-stage products, with important limitations involving compatibility, feature depth and reliability. They may already be useful for selected tasks.\"},{\"q\":\"Does ArtCraft mean large software companies are losing customers?\",\"a\":\"Not necessarily. There is no verified evidence that ArtCraft has caused substantial customer losses for Adobe or Microsoft. Its significance lies in demonstrating how the economics of creating competing software may be changing.\"}]","sources":"[{\"title\":\"ArtCraft official application directory\",\"url\":\"https://getartcraft.com/apps\"},{\"title\":\"Tom's Hardware, 11 October 2026\",\"url\":\"https://www.tomshardware.com/software/microsoft-office/creator-of-free-adobe-clones-unveils-open-source-microsoft-office-excel-powerpoint-and-word-replications-autodesk-tools-and-other-subscription-based-services-may-come-next\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/1b29d4650b242a8dafafb0395b604b2d.webp","featured_image_alt":"Artcraft","featured_image_width":1200,"featured_image_height":630,"intro_image_path":"/api/uploads/2026/10/1b29d4650b242a8dafafb0395b604b2d.webp","intro_image_width":1200,"intro_image_height":630,"seo_title":"What Adobe and Microsoft Can’t Undo","meta_description":"ArtCraft's AI-built Adobe and Microsoft alternatives raise questions about software subscriptions, customer trust and the cost of competition.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":161,"title":"Most Americans Say AI Is Developing Too Fast","slug":"most-americans-say-ai-is-developing-too-fast","short_description":null,"content":"Nearly two-thirds of Americans believe artificial intelligence is developing too quickly, according to a new national survey released on October 8 by the AP-NORC Center for Public Affairs Research.
\nThe findings show that 64% of U.S. adults consider the current pace of AI development too fast, compared with just 8% who believe it is moving too slowly. Approximately one-quarter say development is progressing at about the right pace.
\nThe survey, conducted between September 24 and 28, 2026, involved 2,140 adults across the United States. It comes as technology companies continue introducing increasingly capable AI systems into workplaces, education, information services and other areas of daily life.
\nBeyond concerns about speed, the research identifies two priorities shared by large majorities of respondents: keeping AI under human control and protecting workers from its potential consequences.
\nAlthough attitudes differ between political groups, concern about the pace of AI development is widespread.
\nAmong Democrats, 76% believe AI is advancing too quickly. The figure is lower among Republicans, at 54%, but still represents a majority.
\nThe findings suggest that unease about AI development extends across political affiliations, even when respondents disagree about the broader role of government in regulating emerging technologies.
\nThere is also considerable agreement about several policy objectives.
\nRoughly eight in ten respondents consider protecting American workers and keeping AI under human control important priorities for the U.S. government.
\nThe survey also found broad support across political groups for technological innovation, indicating that concern about AI does not necessarily translate into opposition to its development.
\nDifferences become more pronounced when respondents are asked about environmental protection and competition with other countries.
\nThe emphasis on employment protection reflects an increasingly important question in the public discussion about artificial intelligence: how technological progress should be managed when it affects existing jobs.
\nAI systems are becoming capable of performing tasks previously carried out by employees, particularly in areas involving information processing, customer communication, administrative work and software development.
\nThe AP-NORC findings do not measure actual job losses or predict how employment will change. They measure public priorities at a time when businesses and governments are deciding how to introduce increasingly autonomous technologies.
\nHuman oversight is another central concern.
\nAs AI applications move beyond generating content toward carrying out tasks and supporting decisions, questions about responsibility become more significant.
\nFor public institutions and companies, this includes determining when automated systems should require human approval and who remains accountable when mistakes occur.
\nThe survey suggests that many Americans want these considerations to remain central to AI policy.
\nThe research also examined public confidence in how political leaders are addressing artificial intelligence.
\nOnly 31% of respondents approve of President Donald Trump's handling of AI.
\nAsked which major political party they trust more to manage the technology, 23% selected the Democratic Party and 21% the Republican Party. Another 39% said neither party could be trusted to do a good job.
\nThese findings point to limited public confidence in the ability of political institutions to address the challenges associated with AI development.
\nThey also suggest that respondents' concerns extend beyond the companies building AI systems to the institutions responsible for establishing rules and safeguards.
\nThe survey examined whether Americans are using AI to obtain information ahead of the November 2026 midterm elections.
\nTwenty-one percent reported that they had used or planned to use AI to learn about a political candidate or an issue appearing on the ballot.
\nThe figures were lower for other election-related activities.
\nTwelve percent had used or planned to use AI to find information about voting procedures, while 10% reported the same for voter registration information. Another 10% said they had used or planned to use AI to help decide whom to vote for.
\nBecause these figures combine existing use with intended future use, they should not be interpreted as the proportion of Americans already relying on AI for electoral decisions.
\nNevertheless, the findings indicate that AI is becoming part of how some Americans approach political information, even as concerns about its reliability and oversight persist.
\nThe survey reflects American public opinion and cannot be treated as evidence of attitudes in Europe or the Western Balkans.
\nHowever, the policy questions it raises are relevant to countries considering how to regulate AI and introduce it into public services, education and business.
\nFor Western Balkan economies, where AI adoption is developing alongside efforts to strengthen digital infrastructure and public-sector capacity, questions about skills, employment and human accountability are likely to remain important.
\nThe findings also illustrate why measuring AI adoption alone provides an incomplete picture of public attitudes.
\nPeople may use AI tools while remaining concerned about how quickly the technology is advancing and whether sufficient safeguards are in place.
\nFor policymakers, the challenge is to distinguish public support for technological innovation from public confidence in the institutions overseeing its development.
\nThe AP-NORC survey provides evidence of that distinction in the United States. Whether similar attitudes exist across the Western Balkans would require comparable regional research.
","featured_image_id":366,"intro_image_id":366,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":29,"ordering":0,"published_at":"2026-10-09 08:40:58","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-09 08:40:58","updated_at":"2026-10-09 08:40:58","key_takeaways":"[\"64% of U.S. adults believe AI is developing too quickly, including 76% of Democrats and 54% of Republicans.\",\"Around eight in ten Americans want the government to prioritize keeping AI under human control and protecting workers.\",\"Confidence in political leadership on AI is limited, with 39% saying neither major U.S. political party can be trusted to handle the technology well.\"]","faq":"[{\"q\":\"How many Americans believe AI is developing too quickly?\",\"a\":\"According to AP-NORC, 64% of U.S. adults believe AI is advancing too fast, while 8% say it is advancing too slowly.\"},{\"q\":\"How many people participated in the survey?\",\"a\":\"The survey included 2,140 U.S. adults interviewed between September 24 and 28, 2026. Its overall margin of sampling error is ±2.9 percentage points.\"},{\"q\":\"What concerns do Americans have about AI?\",\"a\":\"The survey identifies strong public support for protecting workers and keeping AI under human control. It also reveals differences over environmental concerns and international competition.\"},{\"q\":\"Does the survey represent attitudes in the Western Balkans?\",\"a\":\"No. The findings represent U.S. public opinion. Comparable polling would be necessary to determine whether people in Western Balkan countries hold similar views.\"}]","sources":"[{\"title\":\"AP-NORC Center for Public Affairs Research, Most Adults Think AI Is Developing Too Fast, October 8, 2026\",\"url\":\"https://apnorc.org/projects/most-adults-think-ai-is-developing-too-fast/\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/e15424ff39580ada3b0b7f9718e618f3.webp","featured_image_alt":"AI Developing Too Fast","featured_image_width":1200,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/e15424ff39580ada3b0b7f9718e618f3.webp","intro_image_width":1200,"intro_image_height":900,"seo_title":"64% of Americans Say AI Is Developing Too Fast","meta_description":"A new AP-NORC survey finds 64% of Americans believe AI is moving too fast, with strong support for worker protection and human oversight.","robots":"index","canonical_url":"americans-ai-developing-too-fast-survey","og_title":"","og_description":""},{"id":158,"title":"Google Unveils Gemini Agent and New Medical AI Findings","slug":"google-unveils-gemini-agent-and-new-medical-ai-findings","short_description":null,"content":"Google announced a new AI agent designed to carry out tasks across workplace applications on October 8, alongside findings from a clinical study in which its medical AI system interacted with real patients before their appointments.
\nThe announcements cover two areas where AI is moving into practical use: everyday business operations and medical consultations. Although the technologies serve different purposes, both raise questions about how much responsibility AI systems should be given and how their work should be supervised.
\nAt its Gemini at Work 2026 event, Google Cloud introduced the Gemini agent, a system designed to complete tasks across business applications rather than simply respond to individual requests.
\nAccording to Google Cloud's announcement, the agent can access organizational information, plan activities, use connected tools and deliver completed work through applications employees already use.
\nThe system is designed to operate across Gmail, Google Drive, Docs, Sheets, Slides, Chat and Calendar. It can also connect to external services, including Microsoft 365, Slack, Salesforce and development environments.
\nFor example, an employee could ask the agent to research a market, prepare a financial analysis and produce a presentation. Instead of managing each application separately, the agent would coordinate the different steps.
\nGoogle also introduced the concept of coworker agents. These are AI agents with dedicated organizational identities, including email addresses, calendars and access permissions. Teams can assign them ongoing responsibilities and interact with them through shared documents or messaging channels.
\nThe company says organizations will be able to control what information these agents can access and review records of their actions.
\nGiving AI systems permission to execute workplace tasks introduces challenges beyond those associated with conventional chatbots.
\nAn agent that prepares a document has limited operational consequences. An agent connected to corporate databases, financial systems or internal communications may be able to take actions affecting an organization's operations.
\nGoogle says its architecture addresses these concerns through separate agent identities, role-based permissions, audit logs and administrative policies.
\nThe system can also select different AI models for different tasks, including models from Google's Gemini family and Anthropic's Claude family. This approach is intended to balance performance and operating costs.
\nFor businesses considering AI agents, the announcement shifts attention toward practical questions about access permissions, oversight and the reliability of automatically executed tasks.
\nOn the same day, Google announced results from a clinical study conducted with Beth Israel Deaconess Medical Center and published in The Lancet.
\nThe research involved 98 patients who interacted with Google's Articulate Medical Intelligence Explorer (AMIE), an experimental conversational AI system, before urgent primary care appointments.
\nUnlike earlier evaluations using simulated consultations, the study examined interactions involving actual patients in a clinical environment.
\nPhysicians supervised the AI conversations in real time. According to Google's summary of the research, none required interruption under the study's predefined safety criteria.
\nThe company reported that AMIE's lists of possible diagnoses included the physicians' final diagnoses in 90% of cases.
\nClinicians also said the AI-generated summaries helped them prepare for patient visits in 75% of cases and influenced their approach to care in more than half of the consultations.
\nThese results suggest that conversational AI may help physicians collect information before appointments and organize relevant details. However, they do not establish that the system can independently diagnose patients or safely replace medical professionals.
\nThe study represents an early investigation of patient-facing AI in a supervised clinical setting.
\nIts relatively small sample, the involvement of supervising physicians and the controlled research conditions limit how widely the findings can be applied.
\nA diagnostic match rate also does not establish that an AI system makes correct clinical decisions in every situation. Medical assessments involve more than generating possible diagnoses, including examination, testing, treatment decisions and follow-up care.
\nGoogle acknowledges that larger clinical trials are needed before the technology's safety and effectiveness can be assessed at scale.
\nThe potential value is therefore more specific than replacing a doctor. AI may be able to help gather patient histories, prepare clinical summaries and support communication between patients and healthcare professionals.
\nThe workplace and healthcare announcements show Google developing AI systems for environments where accuracy, permissions and accountability matter.
\nIn business settings, organizations will need to decide which tasks agents can complete independently and which actions require human approval.
\nHealthcare presents additional requirements involving patient safety, privacy, clinical validation and professional responsibility.
\nFor organizations in the Western Balkans, these developments provide a useful reference point for future AI adoption. The underlying questions are relevant to local businesses, hospitals and public institutions, even where resources and technical infrastructure differ substantially from those of Google's early enterprise customers and clinical research partners.
\nThe next tests will take place outside product demonstrations and limited research settings. For Gemini, organizations will need to establish whether autonomous workflows provide measurable benefits under real operating conditions. For AMIE, larger clinical studies must determine whether the early findings translate into safe and useful care.
","featured_image_id":362,"intro_image_id":362,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":30,"ordering":0,"published_at":"2026-10-09 08:15:56","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-09 08:15:56","updated_at":"2026-10-09 08:15:56","key_takeaways":"[\"Google introduced a Gemini workplace agent capable of coordinating tasks across business applications and connected systems.\",\"A clinical study involving 98 patients found promising results for Google's experimental AMIE system under physician supervision.\",\"Both developments raise important questions about human oversight, reliability and accountability in AI deployment.\"]","faq":"[{\"q\":\"What is Google's new Gemini agent?\",\"a\":\"An AI system designed to plan and execute multi-step workplace tasks using connected applications and organizational information.\"},{\"q\":\"Did Google's medical AI diagnose patients with 90% accuracy?\",\"a\":\"Not exactly. Google reported that AMIE's differential diagnoses included doctors' final diagnoses in 90% of cases. This is not equivalent to independent diagnostic accuracy.\"},{\"q\":\"Can AMIE replace doctors?\",\"a\":\"No. The system remains experimental, and the published study used physician supervision. Further clinical trials are needed.\"},{\"q\":\"Why do these announcements matter?\",\"a\":\"They demonstrate how AI development is moving toward systems that undertake work directly in business environments and assist with specific clinical activities.\"}]","sources":"[{\"title\":\"Google Cloud, Gemini at Work 2026, October 8\",\"url\":\"https://cloud.google.com/blog/products/ai-machine-learning/welcome-to-gemini-at-work-2026\"},{\"title\":\"Google, AMIE clinical study summary, October 8\",\"url\":\"https://blog.google/innovation-and-ai/technology/health/amie-clinical-study-lancet/\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/62faf2606a5020595a7d583e1a75d7c5.webp","featured_image_alt":"Google Gemini At Work","featured_image_width":1600,"featured_image_height":788,"intro_image_path":"/api/uploads/2026/10/62faf2606a5020595a7d583e1a75d7c5.webp","intro_image_width":1600,"intro_image_height":788,"seo_title":"Google Launches Gemini Agent and Reports Medical AI Study","meta_description":"Google introduces a workplace AI agent and reports results from a clinical study involving 98 patients and its experimental medical AI system.","robots":"index","canonical_url":"google-gemini-agent-medical-ai-study","og_title":"","og_description":""},{"id":157,"title":"Mistral Launches Large 4 as Europe Pushes Open AI","slug":"mistral-launches-large-4-as-europe-pushes-open-ai","short_description":null,"content":"French AI company Mistral has launched the public preview of Mistral Large 4, its largest and most capable model to date, as the European company continues to position itself as a major open-weight alternative to US and Chinese AI developers.
\nThe model was released on October 6 and is available through Mistral Studio and its API. Mistral says the open weights will be released later this month.
\nMistral Large 4 is a multimodal Mixture-of-Experts model with around 1 trillion total parameters and 49 billion active parameters. It supports a context window of up to one million tokens and can work across text, documents, images, coding and agentic workflows.
\nThe launch matters beyond model specifications.
\nMistral has increasingly tied its product strategy to a broader European argument around open models, infrastructure and technological sovereignty. Large 4 strengthens that position at a time when European governments and companies are looking for alternatives to relying entirely on AI systems developed outside the continent.
\nMistral says Large 4 is its strongest model so far for coding, multimodal understanding and agentic tasks.
\nThe company also claims the model is competitive with leading open models globally and performs particularly well in enterprise domains such as cybersecurity, finance and law. These performance claims come from Mistral’s own evaluations and will need broader independent testing as the model becomes more widely available.
\nThe open-weight release expected later this month will be particularly important.
\nOpen-weight models allow organisations to run and adapt models in their own environments rather than relying only on hosted APIs. For companies, governments and research institutions dealing with sensitive data, this can provide greater control over deployment, infrastructure and data handling.
\nThat is one reason open models have become a central part of Europe’s AI debate.
\nMistral has made open-weight development one of its defining positions, while also offering commercial cloud services and enterprise products.
\nLarge 4 arrives as Europe attempts to strengthen several parts of its AI ecosystem at the same time.
\nThe European Union is expanding AI Factory infrastructure around EuroHPC supercomputers, while national governments are increasing investment in sovereign computing capacity, research and AI services.
\nMistral itself has also been expanding its European footprint.
\nIn September, the company announced a new German hub in Munich and said it was increasing its work with European industrial customers. Earlier in the same month, Mistral announced a €3 billion Series D funding round that valued the company at more than €21 billion.
\nThe company has increasingly presented its strategy around three connected elements: open models, European infrastructure and enterprise deployment.
\nLarge 4 fits directly into that approach.
\nFor organisations in the Western Balkans, the significance is less about whether Mistral Large 4 wins a benchmark against another frontier model.
\nThe more practical question is whether increasingly capable European open-weight models can become realistic options for companies, universities and public institutions that want greater control over where their AI systems run and how data is processed.
\nThe availability of open weights can make local or regional deployment possible, but model size also matters.
\nA trillion-parameter model requires substantial computing resources even when only a fraction of its parameters are active during inference. That means Large 4 is unlikely to be something most smaller organisations simply download and operate on ordinary infrastructure.
\nInstead, access may depend on cloud providers, specialised data centres, European AI Factories or other shared computing infrastructure.
\nThis makes the development of models such as Large 4 closely linked to the infrastructure question Europe is already trying to solve.
\nMistral Large 4 is currently available as a public preview through Mistral’s hosted services.
\nAccording to the company, the model’s weights are expected to be released before the end of October.
\nThat release will provide a clearer picture of how developers and organisations can deploy, customise and evaluate the model outside Mistral’s own platform.
\nFor Europe, Large 4 is another test of whether the continent can build AI systems that are not only technically competitive, but also usable across its own research, business and public-sector ecosystem.
","featured_image_id":361,"intro_image_id":361,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":35,"ordering":0,"published_at":"2026-10-07 09:00:36","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-07 09:00:36","updated_at":"2026-10-07 09:00:36","key_takeaways":"[\"Mistral released the public preview of Mistral Large 4 on October 6.\",\"The model has around 1 trillion total parameters and 49 billion active parameters.\",\"It supports multimodal inputs, agentic workflows and a context window of up to one million tokens.\",\"Mistral says open weights will be released later in October.\",\"The launch strengthens Mistral’s position in Europe’s push for open and more sovereign AI infrastructure.\"]","faq":"[{\"q\":\"What is Mistral Large 4?\",\"a\":\"Mistral Large 4 is Mistral’s latest general-purpose multimodal AI model and its largest model released so far.\"},{\"q\":\"Is Mistral Large 4 open source?\",\"a\":\"The model is currently available in public preview. Mistral says the open weights will be released later in October 2026.\"},{\"q\":\"How large is the model?\",\"a\":\"Mistral describes Large 4 as having around 1 trillion total parameters, with 49 billion active parameters using a Mixture-of-Experts architecture.\"},{\"q\":\"What can Mistral Large 4 do?\",\"a\":\"It is designed for text generation, coding, multimodal understanding, document processing and agentic workflows.\"},{\"q\":\"Why does it matter for Europe?\",\"a\":\"Mistral is one of Europe’s leading AI model developers, and Large 4 adds another high-capability open-weight model to Europe’s effort to reduce dependence on external AI providers.\"}]","sources":"[{\"title\":\"Mistral Large 4 announcement\",\"url\":\"https://mistral.ai/news/mistral-large-4/\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/f5859a24aaf0e093b88e318587a04820.webp","featured_image_alt":"Mistral","featured_image_width":1600,"featured_image_height":914,"intro_image_path":"/api/uploads/2026/10/f5859a24aaf0e093b88e318587a04820.webp","intro_image_width":1600,"intro_image_height":914,"seo_title":"Mistral Launches Large 4 as Europe Pushes Open AI","meta_description":"Mistral launches Large 4, a trillion-parameter multimodal model, strengthening Europe’s push for open-weight and sovereign AI.","robots":"index","canonical_url":"mistral-launches-large-4-europe-open-ai","og_title":"","og_description":""},{"id":153,"title":"OpenAI Brings Visual Ads to ChatGPT","slug":"openai-brings-visual-ads-to-chatgpt","short_description":null,"content":"OpenAI is introducing a new visual advertising format inside ChatGPT, extending its growing advertising business into one of the platform’s most prominent features: image generation.
\nThe company announced on October 5 that it will begin testing visual ads later this month in the United States with an initial group of advertisers. The ads will appear during image generation and will remain separate from the image itself, according to OpenAI.
\nRather than relying only on text placements, the new format is designed to show products, services and experiences visually while users are exploring ideas or making purchasing decisions.
\nOpenAI says the ads will be clearly labelled and will not influence the answers ChatGPT provides.
\nThe announcement is another step in ChatGPT’s rapid development from an AI assistant into a broader consumer platform where search, shopping, recommendations and advertising increasingly overlap.
\nAdvertising inside ChatGPT is no longer an experiment at the margins of OpenAI’s business.
\nThe company began testing ads with Free and Go users in the United States in February 2026. Since then, it has expanded the advertising programme into additional markets and introduced tools allowing companies and agencies to buy and manage campaigns.
\nBy August, OpenAI said ChatGPT Ads had reached a $1 billion annualised revenue run rate less than 200 days after launch. It also opened advertising access across markets including Europe, India, the Middle East and North Africa.
\nThe latest announcement says ChatGPT now reaches 1.2 billion people each week, giving OpenAI an audience comparable in scale to some of the world's largest consumer internet platforms.
\nThat scale changes the significance of advertising inside an AI assistant.
\nTraditional online advertising typically follows users while they browse websites, social networks or search results. ChatGPT operates differently. People often arrive with an explicit task: comparing products, planning a trip, choosing software, researching a purchase or generating an idea.
\nAdvertising is therefore appearing much closer to the point where decisions are being formed.
\nThe first test of the new format will focus specifically on image generation.
\nOpenAI says advertisers will be able to use images to demonstrate products, show how they might be used or present experiences associated with them. The sponsored content will remain separate from the image generated for the user.
\nThe timing is notable.
\nChatGPT has recently expanded its shopping capabilities, including virtual try-on features for clothing and accessories and the ability to save products inside the platform.
\nTaken together, these developments point toward a version of ChatGPT where product discovery, visualisation, recommendations and transactions become increasingly connected.
\nFor brands, that could make conversational AI a new marketing channel. For users, it makes the distinction between an AI recommendation and a paid placement increasingly important.
\nOpenAI has repeatedly emphasised that advertisements operate separately from ChatGPT's responses.
\nAccording to the company's advertising policy, advertisers cannot pay to influence, rank or alter the answers generated by ChatGPT. OpenAI also says conversations remain private from advertisers and that user data is not sold to them.
\nAds currently appear primarily for users on the Free and Go plans. Plus, Pro, Business, Enterprise and Edu subscriptions remain advertising-free. Users identified as being under 18 are also excluded from advertising.
\nOpenAI additionally restricts advertising around sensitive conversations and categories such as political content, mental health, self-harm, weapons and other high-risk areas.
\nThese safeguards will become increasingly important as advertising becomes more integrated into the ChatGPT experience.
\nFor years, Google dominated online advertising partly because users revealed commercial intent through search queries.
\nAI assistants could create another valuable form of intent.
\nSomeone asking an assistant to choose a laptop, prepare a holiday itinerary, redesign a living room or compare business software may already be close to making a purchase. That makes the interaction commercially valuable even before the user visits an online shop.
\nOpenAI has been steadily building towards that opportunity.
\nIn September, the company introduced sponsored agents that allow users to enter a conversation with a business after interacting with an advertisement. It has also been developing advertising tools, integrations and measurement systems around ChatGPT.
\nThe new visual format takes that strategy one step further.
\nThe immediate rollout remains limited. Testing starts later in October in the United States, and OpenAI has not announced when visual ads will become available to users in Europe or the Western Balkans.
\nBut the direction is becoming clearer: ChatGPT is no longer being developed only as a subscription and software platform. Advertising is becoming a significant part of its consumer business, placing OpenAI increasingly into competition with the companies that have dominated digital advertising for the past two decades.
","featured_image_id":355,"intro_image_id":355,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":16,"ordering":0,"published_at":"2026-10-05 21:08:17","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-05 21:08:17","updated_at":"2026-10-05 21:08:17","key_takeaways":"[\"OpenAI will begin testing a new visual advertising format in ChatGPT later in October 2026, initially in the United States.\",\"Visual ads will first appear during image generation but will remain clearly labelled and separate from generated images.\",\"OpenAI says ChatGPT now reaches 1.2 billion people each week, while its advertising business has already reached a $1 billion annualised revenue run rate.\"]","faq":"[{\"q\":\"Will ads influence ChatGPT's answers?\",\"a\":\"OpenAI says no. Advertising operates separately from the system generating ChatGPT responses, and advertisers cannot pay to change or rank answers.\"},{\"q\":\"Who sees ads in ChatGPT?\",\"a\":\"Ads can currently appear for users on the Free and Go plans. Plus, Pro, Business, Enterprise and Edu plans remain ad-free.\"},{\"q\":\"Where will the new visual ads appear?\",\"a\":\"OpenAI says the first tests will take place during image generation in ChatGPT, with advertisements shown separately from the generated image.\"},{\"q\":\"Are visual ChatGPT ads available in Europe?\",\"a\":\"Not yet as part of this initial test. OpenAI says the visual format will begin testing in the United States later in October. Its broader advertising platform is already to advertisers in multiple European markets.\"}]","sources":"[{\"title\":\"OpenAI, Building advertising for the way people use AI, October 5, 2026.\",\"url\":\"https://openai.com/index/new-chatgpt-ads-format-and-measurement\"},{\"title\":\"OpenAI Help Center, Ads in ChatGPT.\",\"url\":\"https://help.openai.com/en/articles/20001047-ads-in-chatgpt\"}]","category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/01c37ab0f346bf7bc6fccfc186e89315.webp","featured_image_alt":"OpenAI Adds","featured_image_width":1600,"featured_image_height":800,"intro_image_path":"/api/uploads/2026/10/01c37ab0f346bf7bc6fccfc186e89315.webp","intro_image_width":1600,"intro_image_height":800,"seo_title":"OpenAI Brings Visual Ads to ChatGPT","meta_description":"OpenAI is testing visual ads in ChatGPT image generation as its advertising business grows and the platform reaches 1.2 billion weekly users.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":149,"title":"Open-Weight AI Models Are Reaching Advanced Cyber Capabilities","slug":"open-weight-ai-models-are-reaching-advanced-cyber-capabilities","short_description":"","content":"Advanced cybersecurity capabilities that only months ago were largely confined to tightly controlled frontier AI systems are beginning to appear in models that anyone can download.
\nAnthropic said on September 29 that GLM-5.3, an open-weight model developed by Chinese AI company Z.ai, can autonomously build sophisticated end-to-end cyber exploits and has safeguards that the company found relatively easy to bypass.
\nThe findings add to growing evidence that the gap between restricted frontier AI systems and openly available models is narrowing in cybersecurity.
\nGLM-5.3 was released by Z.ai in August, with its model weights subsequently made publicly available.
\nAnthropic tested the model as part of its Frontier Red Team research and reported that simple techniques could bypass GLM-5.3's safeguards between 64% and 100% of the time in its simulated evaluations. By comparison, Anthropic said the same attacks were unsuccessful against the safeguarded Claude models included in its testing.
\nThe results do not mean GLM-5.3 is the world's most capable AI model for cybersecurity.
\nAn independent assessment published earlier by the U.S. National Institute of Standards and Technology's Center for AI Standards and Innovation found that GLM-5.3 still performs below the strongest U.S. frontier models.
\nBut NIST reached another important conclusion: GLM-5.3 is currently the most cyber-capable open-weight model it has evaluated.
\nThe distinction matters because some of the most powerful cybersecurity capabilities developed by U.S. AI companies are not freely available.
\nAnthropic, for example, has previously restricted access to some advanced cyber capabilities through trusted-access programmes designed for vetted security organisations.
\nOpen-weight models work differently. Once their weights are publicly released, organisations and individuals can download them, modify them and run them on their own infrastructure.
\nThat makes conventional API restrictions much harder to enforce.
\nAccording to Anthropic, GLM-5.3 demonstrated capabilities similar to those shown five months earlier by Claude Mythos Preview, a restricted Anthropic system capable of autonomously developing sophisticated exploits.
\nNIST's benchmark results provide additional context.
\nGLM-5.3 solved 40.4% of tasks in SEC-Bench Pro and achieved 61.1% on ExploitBench, while the strongest U.S. models evaluated by NIST performed considerably better. Across its combined cyber capability measure, NIST estimated that GLM-5.3 trails the current U.S. frontier by approximately four months.
\nFour months may sound substantial in ordinary software development.
\nIn frontier AI, it is not.
\nThe more important change is that capabilities that previously required access to restricted systems are appearing in models whose weights can be downloaded publicly.
\nThe same capabilities are not inherently malicious.
\nModels capable of identifying vulnerabilities and developing exploits can help security teams discover weaknesses before attackers do.
\nAnthropic explicitly acknowledged this dual use, arguing that defenders should also have access to highly capable AI systems as attackers adopt increasingly powerful tools.
\nThe challenge is that defensive and offensive cybersecurity often rely on many of the same technical skills.
\nA system that can find a vulnerability for a security researcher may also help someone exploit it.
\nThat makes cybersecurity one of the clearest areas where the debate around open AI models becomes more complicated than a simple choice between open and closed development.
\nFor organisations in the Western Balkans, the immediate issue is less about which country currently leads AI benchmarks and more about the falling cost of sophisticated cyber capability.
\nBusinesses, public institutions and civil society organisations across the region are adopting AI tools while often working with limited cybersecurity staff and budgets.
\nIf advanced vulnerability discovery and exploit development become available through downloadable models, the technical barrier for both attackers and defenders falls.
\nThat puts more pressure on ordinary security practices: patching known vulnerabilities, restricting system access, monitoring unusual activity, separating sensitive infrastructure and ensuring that AI agents do not receive unnecessary permissions.
\nThe development also complicates AI governance.
\nRestrictions imposed by individual AI providers can reduce misuse through hosted services, but they cannot easily control models once weights have been released and distributed.
\nGLM-5.3 is unlikely to be the last model to raise that problem.
\nThe question now is how quickly openly available models will close the remaining cybersecurity capability gap with the most advanced restricted systems.
","featured_image_id":329,"intro_image_id":329,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":6,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":37,"ordering":0,"published_at":"2026-10-01 08:26:14","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-01 08:26:14","updated_at":"2026-10-01 08:26:14","key_takeaways":"[\"NIST describes GLM-5.3 as the most cyber-capable open-weight AI model it has evaluated so far.\",\"Anthropic says simple techniques bypassed GLM-5.3's cyber safeguards in 64% to 100% of its simulated tests.\",\"The broader issue is that advanced cyber capabilities are moving from restricted frontier systems into downloadable models.\"]","faq":"[{\"q\":\"What is GLM-5.3?\",\"a\":\"GLM-5.3 is an artificial intelligence model developed by Chinese AI company Z.ai, formerly known as Zhipu AI. Its model weights have been publicly released.\"},{\"q\":\"Is GLM-5.3 more capable than U.S. frontier AI models?\",\"a\":\"No. NIST found that it remains below the strongest U.S. models on its cybersecurity benchmarks, estimating a gap of about four months on its combined capability measure.\"},{\"q\":\"Why is its open-weight release important?\",\"a\":\"Open-weight models can be downloaded, modified and operated independently. This makes provider-level restrictions and API safeguards harder to enforce once the model has been distributed.\"},{\"q\":\"Can these capabilities also help cybersecurity teams?\",\"a\":\"Yes. The same capabilities used to discover and exploit vulnerabilities can help defenders identify weaknesses, test systems and fix security problems.\"}]","sources":null,"category_name":"Global AI","category_slug":"global-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/d412ad44c1474ab547b909cea142eac0.webp","featured_image_alt":"GLM-5.3 logo and text beside layered, glowing geometric shapes on a dark background.","featured_image_width":1600,"featured_image_height":800,"intro_image_path":"/api/uploads/2026/10/d412ad44c1474ab547b909cea142eac0.webp","intro_image_width":1600,"intro_image_height":800,"seo_title":"Open-Weight AI Models Reach Advanced Cyber Capabilities","meta_description":"Anthropic and NIST say Z.ai’s GLM-5.3 brings advanced cyber capabilities to an openly downloadable AI model.","robots":"index","canonical_url":"","og_title":"","og_description":""}]},{"category":{"id":3,"name":"Insights","slug":"insights"},"posts":[{"id":42,"title":"The AI Gold Rush Meets a Wall of Costs and Caution","slug":"the-ai-gold-rush-meets-a-wall-of-costs-and-caution","short_description":"","content":"The artificial intelligence boom that has powered markets, tech valuations, and corporate strategy for two years is now facing a harder question: how much of this surge is real, and how much is speculation?
\nWhat once looked like a race to dominate the future is increasingly looking like a test of discipline. Companies are still investing heavily in AI, but the mood has shifted. Costs are rising fast, returns are proving harder to measure, and some of the biggest names in tech are beginning to slow down, trim usage, or rethink how much they are willing to spend.
\nThat shift matters because bubbles rarely form around worthless ideas. They form when a powerful idea attracts so much capital that financial expectations move ahead of business reality. AI is clearly real. It is already changing software development, customer support, data analysis, and enterprise workflows. But the question now is whether the money flowing into AI is running ahead of what companies can actually earn from it.
\nOne of the clearest signs of pressure is cost. AI systems are expensive to build and even more expensive to run at scale. That has already forced some companies to become more selective. Microsoft reportedly reduced or cut back internal AI usage after the bill grew too large. Uber, meanwhile, reportedly used up its full-year AI budget in just a few months. In both cases, the issue was not whether the technology was useful, but whether the spending could be defended in practical business terms.
\nThat is the heart of the current slowdown. Many firms adopted AI quickly during the first wave of enthusiasm. But once usage spread, the bills arrived. What looked like a low-friction productivity boost often turned into a recurring expense with no immediate guarantee of return. Executives are now asking tougher questions: Does this tool improve margins? Does it reduce headcount, save time, or generate revenue? If the answer is unclear, the spending gets cut.
\nAnthropic adds another important layer to the story. Rather than simply chasing growth, the company has publicly urged a temporary pause in frontier AI development so policymakers and society can catch up with the risks. That is not a sign of retreat so much as a sign that even inside the industry, there is growing recognition that speed alone is not enough. Safety, governance, and accountability are becoming part of the conversation.
\nThere is also a broader market shift. Investors who once treated every AI-related company as a sure winner are becoming more cautious. Some have trimmed exposure to leading AI names after strong gains, suggesting that the market is starting to separate durable businesses from speculative bets. That is often how a boom loses momentum: not with a sudden collapse, but with a gradual loss of conviction.
\nThe infrastructure behind AI is also under pressure. Data centers, chips, cloud capacity, and power consumption all require enormous capital. Big Tech has poured billions into that buildout, but the size of the investment is now raising hard questions about sustainability. Energy demand is rising. Profitability is uncertain. And in some cases, the cost of staying in the race may be approaching the limit of what even the largest companies want to absorb.
\nThis is why comparisons to the dot-com era keep resurfacing. The point is not that AI is a fad or a fake revolution. It is that real revolutions often go through speculative excess before the market finds a more realistic balance. The internet changed the world, but many internet companies failed when the hype outpaced the fundamentals. AI may be following a similar path.
\nIf so, the likely outcome is not the collapse of AI itself, but a correction in expectations. The strongest companies, with real products and proven business models, are likely to survive and even strengthen. The weaker firms, the overvalued startups, and the projects built mostly on narrative may not.
\nThat would not mean the AI story is over. It would mean the market is maturing. And that may be the real sign of where the AI boom stands today: still powerful, still transformative, but no longer untouchable.
","featured_image_id":97,"intro_image_id":97,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":3,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":9,"ordering":0,"published_at":"2026-06-27 10:10:03","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-27 10:10:03","updated_at":"2026-06-27 10:10:03","key_takeaways":null,"faq":null,"sources":null,"category_name":"Insights","category_slug":"insights","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/2344470b78292bf8022e9b053d1cc747.webp","featured_image_alt":"A silhouetted man faces glowing AI dashboards showing declines of 9.2% and 4.8% and a fluctuating blue line chart.","featured_image_width":1536,"featured_image_height":1024,"intro_image_path":"/api/uploads/2026/06/2344470b78292bf8022e9b053d1cc747.webp","intro_image_width":1536,"intro_image_height":1024,"seo_title":"The AI Gold Rush Meets a Wall of Costs and Caution","meta_description":"The artificial intelligence boom that has powered markets, tech valuations, and corporate strategy for two years is now facing a harder question: how…","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":38,"title":"The Balkans’ Biggest AI Opportunity Is Industry, Not Chatbots","slug":"the-balkans-biggest-ai-opportunity-is-industry-not-chatbots","short_description":"","content":"The artificial intelligence conversation in the Balkans is often dominated by public-sector experiments, language models, education and regulation. But the region’s biggest economic AI opportunity may be much more practical: industry, manufacturing and small business productivity.
\nFor many Balkan economies, the central challenge is not whether they can create the next global AI platform. The more immediate question is whether AI can help companies become more efficient, competitive and export-ready. This is where applied AI could have the strongest impact.
\nManufacturing, logistics, agriculture, energy, tourism and business services are all areas where AI can create measurable value. Companies can use AI for predictive maintenance, quality control, demand forecasting, route optimization, document processing, customer support and energy efficiency. These are not futuristic ideas. They are practical tools that can reduce costs and improve performance.
\nThis matters because the Western Balkans have many small and medium-sized enterprises that struggle with limited financing, skills shortages and productivity gaps. AI will not solve all of these problems, but it can help companies do more with fewer resources if adopted correctly.
\nThe recent regional focus on AI for industry and manufacturing shows that international organizations are beginning to recognize this opportunity. Industrial AI is different from consumer AI. It is less about generating text and more about improving processes, machines, supply chains and decisions.
\nFor Balkan businesses, this shift could be valuable. Many companies do not need a custom large language model. They need better inventory planning, faster customer communication, automated paperwork, improved production monitoring or smarter use of existing data. These use cases are easier to justify financially because they connect directly to business outcomes.
\nHowever, adoption will not happen automatically. Many SMEs in the region still lack digital maturity. Some do not have clean data, modern software systems or staff trained to use AI tools. Before companies can adopt advanced AI, they often need basic digitalization: structured databases, cloud systems, cybersecurity practices and process documentation.
\nThis is why AI policy in the Balkans should not focus only on high-level strategy. It should include practical support for companies. Chambers of commerce, innovation funds, universities and technology providers can help SMEs identify realistic AI use cases and avoid wasting money on tools they do not need.
\nThere is also a need for local AI integrators. Many businesses will not build AI systems internally. They will need regional technology companies that can understand their sector, adapt existing tools and implement solutions in local languages and business environments. This creates an opportunity for Balkan software firms to move from outsourcing toward higher-value AI consulting and product development.
\nAnother opportunity is regional specialization. Countries do not need to compete in exactly the same way. Serbia may lead in AI infrastructure and research, Albania in public-sector AI, North Macedonia in AI factory access and language technology, Kosovo in youth-driven digital services, Montenegro in public administration and tourism use cases, and Bosnia and Herzegovina in industrial and SME transformation. Together, these strengths could form a regional AI market.
\nThe biggest risk is that AI becomes a buzzword rather than a productivity tool. Businesses may feel pressure to “use AI” without understanding why. This can lead to shallow adoption, poor implementation and disappointment. The better approach is to start with business problems, not technology trends.
\nFor example, a factory should not begin by asking which AI model is popular. It should ask where production delays happen, where quality issues appear, where energy is wasted and where human teams spend time on repetitive tasks. AI becomes useful only when it is connected to a clear operational problem.
\nThe Balkans’ AI future may therefore be less glamorous than global headlines suggest, but more economically important. The region does not need every company to become an AI company. It needs thousands of ordinary companies to become more productive through smart, responsible AI adoption.
\nIf governments, chambers, universities and tech firms work together, industrial AI could become one of the strongest drivers of regional competitiveness. The countries that succeed will be those that move beyond AI hype and help businesses apply the technology where it produces real value.
\nFor the Balkans, the next AI leap may not come from a chatbot. It may come from a factory floor, a logistics company, a hospital back office, a tourism platform or a small business using data better than before.
","featured_image_id":77,"intro_image_id":77,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":3,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":9,"ordering":0,"published_at":"2026-06-25 18:18:20","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 18:18:20","updated_at":"2026-06-25 18:18:20","key_takeaways":null,"faq":null,"sources":null,"category_name":"Insights","category_slug":"insights","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/45781f3915068724308236aa0674f3c1.webp","featured_image_alt":"A humanoid robot operates a glowing blue console amid neon purple servers and data displays.","featured_image_width":1600,"featured_image_height":780,"intro_image_path":"/api/uploads/2026/06/45781f3915068724308236aa0674f3c1.webp","intro_image_width":1600,"intro_image_height":780,"seo_title":"The Balkans’ Biggest AI Opportunity Is Industry, Not Chatbots","meta_description":"The artificial intelligence conversation in the Balkans is often dominated by public-sector experiments, language models, education and regulation.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":37,"title":"AI Regulation Becomes the Balkans’ Next EU Test","slug":"ai-regulation-becomes-the-balkans-next-eu-test","short_description":"","content":"Artificial intelligence is quickly becoming a new test of European integration for the Western Balkans. As the European Union moves ahead with the AI Act and a broader AI policy framework, countries in the region are being pushed to think not only about innovation, but also about regulation, responsibility and trust.
\nThis is a major shift. For years, digital transformation in the Balkans was mostly discussed through e-government, broadband access, startup development and public-sector modernization. Now, AI is forcing governments to answer deeper questions: Who is responsible when an algorithm makes a harmful recommendation? How should high-risk AI systems be supervised? Can public institutions use AI without weakening citizens’ rights? How can smaller economies regulate technology without blocking innovation?
\nThe first structured regional dialogue on artificial intelligence among the Western Balkans Six shows that the region understands the need for coordination. AI policy leads from the six economies met in Podgorica to discuss responsible and sustainable AI adoption, shared priorities and common challenges. This matters because no Balkan country can solve AI governance alone.
\nEU alignment will be one of the strongest drivers of AI regulation in the region. Countries that want closer integration with the EU will need to adapt their legal systems to European standards. That includes risk-based regulation, transparency obligations, human oversight, data protection and safeguards for high-risk AI systems.
\nFor businesses, this creates both pressure and opportunity. Companies building AI tools for European clients will increasingly need to understand compliance. Those that prepare early may gain a competitive advantage. Those that ignore regulation may find it harder to sell into EU markets or work with regulated industries such as healthcare, finance, education and public administration.
\nFor governments, the challenge is more complex. Public institutions want to use AI to improve services, reduce administrative delays and analyze data more efficiently. But public-sector AI is sensitive because it can affect access to benefits, permits, healthcare, education and other essential services. In these areas, mistakes are not only technical failures; they can become rights violations.
\nThis is why AI regulation in the Balkans cannot simply copy European documents without building local capacity. Laws are necessary, but they are not enough. Countries need regulators who understand AI systems, public officials who can manage digital projects responsibly, and courts or oversight bodies that can evaluate algorithmic harm.
\nThe region also needs regulatory sandboxes. These controlled testing environments can help startups, public institutions and researchers experiment with AI while staying within clear safety rules. For smaller markets, sandboxes may be especially useful because they allow innovation without immediately imposing heavy compliance burdens on early-stage companies.
\nAnother important issue is public trust. Many Balkan citizens already have low confidence in institutions. If AI is introduced into public services without transparency, it could increase suspicion rather than improve efficiency. Citizens need to know when AI is being used, what data it relies on and how decisions can be challenged.
\nThe Balkans also face a risk of regulatory fragmentation. If each country develops different AI rules, companies and institutions may struggle to scale across the region. A regional AI policy roadmap could help create more consistency, especially in areas such as public procurement, education, data governance and innovation support.
\nAt the same time, regulation must not become a barrier that only large companies can afford. If compliance costs are too high, local startups and small businesses may avoid AI development or depend entirely on foreign platforms. The region needs a balanced approach: strong enough to protect citizens, but flexible enough to support innovation.
\nAI regulation is therefore becoming more than a legal topic. It is becoming part of the region’s economic strategy, EU integration path and democratic development. Countries that manage it well will be better positioned to attract investment, protect citizens and build trustworthy digital services.
\nThe next stage for the Balkans will be practical implementation. The region does not only need AI strategies and conferences. It needs working institutions, clear rules, trained regulators and responsible adoption by companies and governments.
\nIf the Western Balkans can align with European AI standards while preserving space for local innovation, the region could turn regulation into an advantage. Instead of being late to AI, it could become a region known for practical, trustworthy and EU-compatible AI adoption.
","featured_image_id":75,"intro_image_id":75,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":3,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":1,"show_table_of_contents":1,"read_time_minutes":4,"view_count":15,"ordering":0,"published_at":"2026-06-25 18:15:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 18:15:47","updated_at":"2026-06-25 18:34:06","key_takeaways":null,"faq":null,"sources":null,"category_name":"Insights","category_slug":"insights","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/e357e22a7b081def9e3675d7580849e5.webp","featured_image_alt":"AI text centered in a ring of EU stars between two blue circuit-patterned human profiles.","featured_image_width":1600,"featured_image_height":995,"intro_image_path":"/api/uploads/2026/06/e357e22a7b081def9e3675d7580849e5.webp","intro_image_width":1600,"intro_image_height":995,"seo_title":"AI Regulation Becomes the Balkans’ Next EU Test | AI Balkans","meta_description":"Artificial intelligence is quickly becoming a new test of European integration for the Western Balkans.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":36,"title":"Balkans Shift From AI Hype to Real Infrastructure","slug":"balkans-shift-from-ai-hype-to-real-infrastructure","short_description":"","content":"Artificial intelligence in the Balkans is entering a new phase. Until recently, much of the regional conversation around AI focused on individual tools, public experiments, isolated startups and global trends imported from larger markets. In 2026, the conversation is becoming more serious. The region is beginning to build the foundations needed for long-term AI adoption: infrastructure, policy coordination, public-sector readiness and regional cooperation.
\nThis shift is important because AI development does not depend only on creativity or software talent. It also depends on computing power, access to data, legal clarity, institutional capacity and practical use cases. Without these foundations, countries can use foreign AI tools, but they cannot fully participate in the development of AI systems that reflect local languages, economies and public needs.
\nThe Western Balkans are now moving closer to this infrastructure-based model. Serbia and North Macedonia are becoming connected to Europe’s AI Factory and high-performance computing ecosystem. These initiatives are designed to give local companies, researchers and institutions access to AI-optimized supercomputing resources, technical expertise and European innovation networks.
\nFor smaller economies, this is a major opportunity. Building advanced AI infrastructure independently is expensive and difficult. By connecting with European systems, Balkan countries can reduce the gap between local ambition and technical capacity. This gives startups, universities and public institutions a chance to experiment with more advanced AI solutions without having to rely entirely on commercial foreign platforms.
\nThe rise of AI infrastructure also changes the type of AI projects the region can realistically pursue. Instead of limiting AI adoption to chatbots, translation tools or basic automation, countries can begin exploring more complex use cases: medical imaging, smart manufacturing, energy optimization, local-language models, public data analysis, agricultural forecasting and industrial automation.
\nHowever, infrastructure alone will not create an AI economy. The region also needs people who can use it. This means engineers, researchers, data scientists, cybersecurity experts, public-sector managers and business leaders who understand what AI can and cannot do. Without skills development, even the best infrastructure will remain underused.
\nAnother key challenge is access. AI infrastructure must not become available only to a small circle of institutions or large companies. If the goal is regional transformation, startups, small businesses, universities and civil society organizations must be able to participate. Otherwise, the benefits of AI will remain concentrated.
\nThe Balkans also need stronger data ecosystems. AI systems depend on high-quality, well-organized and legally usable data. Many public institutions in the region still struggle with fragmented databases, outdated systems and limited interoperability. This creates a bottleneck for AI adoption, especially in public services and healthcare.
\nStill, the direction is clear. The region is moving beyond simply discussing AI and beginning to connect itself to the infrastructure that serious AI development requires. This is a major change from the earlier stage of digital transformation, where the main goal was to put services online or support basic software outsourcing.
\nFor the Balkans, the next competitive advantage may not come from trying to build global AI giants immediately. It may come from becoming a strong applied-AI region: a place where AI is used to solve practical problems in government, education, industry, language technology and public services.
\nThe real test will be whether countries can turn infrastructure into accessible innovation. If AI factories, supercomputing links and regional platforms become tools that local innovators actually use, the Balkans could move from AI consumers to AI contributors. If not, the region risks building impressive structures without enough practical impact.
\nThe current moment is therefore critical. The Balkans have talent, digital ambition and growing European support. What they need now is execution: open access, practical training, strong governance and real projects that citizens and businesses can feel.
","featured_image_id":74,"intro_image_id":74,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":3,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":11,"ordering":0,"published_at":"2026-06-25 18:13:48","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 18:13:48","updated_at":"2026-06-25 18:13:48","key_takeaways":null,"faq":null,"sources":null,"category_name":"Insights","category_slug":"insights","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/3ea4d909641a88314630b8a663d7fcdf.webp","featured_image_alt":"A stylized miniature city of servers, buildings and cloud symbols linked by cables and glowing AI nodes.","featured_image_width":1600,"featured_image_height":914,"intro_image_path":"/api/uploads/2026/06/3ea4d909641a88314630b8a663d7fcdf.webp","intro_image_width":1600,"intro_image_height":914,"seo_title":"Balkans Shift From AI Hype to Real Infrastructure","meta_description":"Artificial intelligence in the Balkans is entering a new phase.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":18,"title":"Who controls the machines that will control the world?","slug":"who-controls-the-machines-that-will-control-the-world","short_description":"Artificial General Intelligence (AGI) has become one of those terms that seems to mean everything and nothing at the same time. In Silicon Valley, it is…","content":"Artificial General Intelligence (AGI) has become one of those terms that seems to mean everything and nothing at the same time. In Silicon Valley, it is marketed as the inevitable next step in technological evolution. In policy circles, it is framed as a looming national security issue. For ethicists, it raises difficult questions about consciousness, rights, and responsibility.
\nBut the more immediate question is much simpler: who gets to control it?
\nThat question runs through the reflections of Ben Goertzel, the AI researcher who popularized AGI decades before it entered mainstream discourse and now leads SingularityNET and the Artificial Superintelligence Alliance. His argument is not about a single technological milestone. It is that AGI could reshape the architecture of power itself.
\nThe debate is often reduced to whether machines will become conscious or smarter than humans. Goertzel’s concern is that, long before that point, societies may already have handed over critical layers of knowledge production and governance to a small group of actors.
\nHistory offers familiar warnings. The internet was expected to decentralize information, social media to flatten hierarchies. Instead, both consolidated power inside a handful of corporations. Now, there is no guarantee AI will behave differently.
\nYet not everyone agrees that technology should be the center of the debate. Dr. Seth Dobrin, a geneticist-turned-AI strategist, argues that public discourse begins with a category error: attributing agency to systems that merely generate fluent language.
\n“We’re applying human moral vocabulary to mathematics. A model is software: weights, matrices, and optimization targets,” he argues in an interview via email for AGI Ethics News. In his view, AI should be treated like any engineered system: verified, predictable, and governed through accountability, not anthropomorphism.
\nPower and the illusion of inevitability
\nThat disagreement exposes a deeper divide. Goertzel explores whether future systems might exhibit some form of subjective experience. Dobrin insists the real issue is not machine being, but human responsibility.
\n“You test a tool. You empathize with a being. Confuse the two, and you end up doing neither well”
\nDespite their differences, both converge on a core point: the danger is not machine autonomy, but human concentration of control.
\nGoertzel describes a “sovereign tail” risk in which advanced AI becomes a strategic asset controlled by a few states or hyperscale corporations, forming the infrastructure of economic and informational life.
\n“The deeper danger then inverts: not concentration, but uncoordinated proliferation of forkable systems. The answer is the same: build intelligence as a decentralized network whose power lives in the collective whole.” Goerzel noted in an interview via email for AGI Ethics News.
\nDobrin is more direct. “AGI doesn’t hold power. Companies and states hold power. The technology is a lever; the hands on the lever are human,” he says. The real question is whether a small number of corporations will end up governing global information infrastructure.
\nBoth reject AGI as an autonomous historical force. Instead, it becomes an amplifier of existing power structures shaped by design, ownership, and deployment choices.
\nDecentralization, knowledge, and control
\nGoertzel challenges the assumption that AGI must be built on centralized compute. He instead envisions intelligence emerging across distributed networks rather than being locked inside corporate infrastructure.
\n“The geopolitically safest AGI is one nobody can own,” he argues, because intelligence should be “an event the network performs, not an object at rest anyone can capture.”
\nDobrin reaches a similar conclusion through governance logic. He advocates smaller, specialized models operating on sovereign data and embedded in local systems. The geopolitics of AI, he says, “will be decided by procurement choices, not prophecy.”
\nAt stake is a broader question: whether intelligence can be owned at all, or whether it behaves more like language or culture… something that exists only through distribution. As AI systems increasingly mediate knowledge itself, control over their architecture becomes control over perception. The machine is no longer just carrying knowledge, but it is also shaping it.
\nThe philosophical stakes, though, deepen with consciousness.
\nGoertzel, influenced by Alfred North Whitehead, leans toward pan-experientialism — the idea that experience may be a fundamental feature of organized matter. The question is not whether AI is conscious, but what kinds of internal states its architecture enables.
\nDobrin rejects this framing. “You test a tool. You empathize with a being. Confuse the two, and you end up doing neither well,” he says. AI systems, regardless of complexity, cannot hold rights because they cannot hold responsibility. Only humans and institutions can.
\nDespite this disagreement, both converge on a key concern: opacity.
\nGoertzel argues that ethical debate requires visibility into how systems operate. Black-box models encourage projection, while interpretable systems allow partial understanding of internal reasoning.
\nDobrin echoes this in engineering terms: “You’re entitled to systems that show their work mathematically. Proof, not prediction.”
\nAs systems scale, however, they often become less interpretable, intensifying the governance challenge.
\nA future shaped by choices, not inevitability
\nAI is no longer just distributing information; it is mediating and synthesizing it before humans encounter it. It is actively shaping knowledge itself.
\nGoertzel proposes governance through decentralized, constitutional-style AI ecosystems where transparency and accountability are embedded at the architectural level.
\nDobrin reduces governance to a simpler principle: responsibility never transfers to the machine. “A system can mediate a decision, but a human or institution must own it,” he says. Once “the algorithm decided” becomes acceptable, accountability collapses.
\nIn a statement for AGI Ethics News, innovation strategist Dan Herman points out that over 90 percent of AI startup funding is concentrated in the US and China. For smaller economies, the challenge is not dominance but coordination – building ecosystems capable of collective participation in AI infrastructure.
\nSeen together, Goertzel and Dobrin offer complementary diagnoses. One focuses on the architecture of intelligence, the other on institutional accountability, and both reject technological inevitability.
\nThe central question of AGI may never be whether machines become human-like. It is whether humans allow the systems they build to reproduce the oldest pattern in political history: the concentration of power in fewer and fewer hands.
\nEthical concerns addressed in this article:
\nIn the last thirty years, we haven’t gotten new gadgets and toys, we’ve been actually brought into a completely different reality.
\nFirst the internet. Then broadband. Then smartphones. Then social media. And now artificial intelligence. One wave after another, faster than the last. We’ve learned how to use them quickly, how to scroll, how to download content, how to find information, but we haven’t learned how to live with them properly. That part is missing.
\nThis is the fundamental problem. Our adaptation is consumerist, not social. We know how to be users, but we don’t know how to protect ourselves or be more productive.
\nIndustrial revolutions and technological changes in the past lasted for decades and centuries, and they really took time to be fully used. Laws, regulations, rules, norms, education, public health, all of this was regulated after the new machines were invented. Although it was slow, humanity eventually caught up with the technology. With digital technology, this is drastically different. The gap between the rapid technological revolution and social adaptation continues to grow. By the time schools start teaching something about social media literacy, the platforms have already changed three or four times.
\nLook how long it took for serious debates about smartphones in the classroom? It’s only in the last few years that some countries have started to regulate or limit smartphones due to problems with attention and mental health in young people. But children have had them for more than two decades. Regulation comes late. Education comes late. Protection comes late.
\nThe economy and the market, of course, arrive very early.
\nWe adapt when money is involved. New apps? Immediately. New phones? At 2 years old. New AI tools? Overnight. And ethical frameworks? Critical thinking and training? Psychological coping strategies? ... It’s still loading.
\nAlways connected, but not present
\nMobile phones have not only changed communication, they have also replaced presence. Conversations are now competing with notifications. Face-to-face contact is being interrupted by multitasking across screens. Silence is uncomfortable unless you have a digital device in your hand. Researches show that constant screen exposure weakens the ability to read facial expressions and emotional cues. Empathy diminishes, attention spans fade. Even parents who use their phones around their children are disrupting emotional connection. It’s not a small change in behavior, but a deeply transformative one with serious consequences in our daily socialization.
\nWe talk more than ever. Messages, comments, reactions. And depth and intimacy in communication? Not really.
\nSocial media has only magnified and multiplied the problems. Community used to mean shared physical space and stable long-term interaction. Now it often means just an audience. Feed. Performance. You don’t belong, you just participate and post. And the logic and algorithms of the platforms reward attention and visibility, but not understanding. Likes instead of listening. Metrics instead of understanding.
\nThese communication systems are not neutral tools. They are designed to keep us there, scrolling, clicking, checking, reacting. Notifications, threads, endless feeds and scrolling. Little dopamine hits. Over and over.
\nThen anxiety happens. Sleep is disrupted. Self-esteem is tied to numbers and online validation. Especially among young people. There is strong evidence linking excessive social media use to depression, loneliness, and mood instability, and yet we still treat it as normal behavior because everyone is online.
\nWe have not built social norms for digital balance. We have not found answers to critical public health questions, and we have only normalized excessive use.
\nArtificial Intelligence: Outsourcing thinking
\nArtificial Intelligence is a game-changer. Previous technologies have expanded our memory and accelerated communication. AI is expanding thinking itself. Writing, summarizing, analyzing, making decisions. Practical, yes. Powerful, absolutely. But also very risky if not done properly and carefully.
\nWhen thinking becomes optional, it simultaneously weakens it. Students who use AI often skip the cognitive stage that results in understanding. Adults who use AI without proper training show more stress and lower self-esteem. The tool that is supposed to help ends up being addictive.
\nI'm not saying that AI is bad. It's not. It can support learning, creativity, accessibility. But without educational frameworks, it becomes a machine for addiction to technology. And economic addiction to buying premium versions of AI-based services.
\nWe externalize memory and cognitive activities to algorithms and the cloud. We externalize reasoning. Where does it all end?
\nWe have adapted to rapid technological innovation as consumers and users, but not as societies, communities, and citizens. We know how to install applications, but do not how algorithms shape our perception. We accept terms of service without reading them. We exchange data for any convenience and time saving. We measure ourselves through the metrics of digital platforms. Meanwhile, regulation is slow, fragmented, often reactive. Educational systems teach software skills, but rarely teach attention control, digital ethics, or AI literacy. Psychological protection is almost nonexistent.
\nWhat would real adaptation mean?
\nIf we were actually adapting, socially, not just consumer and commercial-wise three things would look drastically different.
\nFirst, education. Digital literacy would mean more than using tools. It would involve critical analysis of media, awareness of algorithms, ethics of artificial intelligence, attention control. Learning how to think with technology without letting it think for us.
\nSecond, regulation. The design of platforms would be scrutinized in the same way that we scrutinize food safety or pharmaceuticals. Child protection would be proactive, not reactive. Data use would be transparent.
\nThird, mental health protection. We would treat problematic digital use as a public health issue. Promoting offline interaction. Rebuilding shared spaces and physical interaction. Normalizing disconnection from digital devices as a healthy behavior. Right now, we are doing the opposite. Constant connectivity is the norm while disconnection is an aberration.
\nSo here we are. Technologically advanced, socially unprepared and fragmented, personally isolated while being stressed and anxious. Cognitively assisted, mentally exhausted. Connected but alone.
\nFinally, the point is not to reject technology or AI. That would be naive and impossible. The point is to stop encouraging and imposing its use without adaptation and protection. Using tools is becoming easier. Building rules, norms, protections, and skills for them, that's the hard part. We're skipping the hard part, at least for now.
\nBut unless we develop educational, regulatory, and psychological frameworks that match the speed of innovation, technology will continue to shape human behavior according to market logic and profit, not human well-being. And that's the real risk. Not the machines themselves, but our failure to grow with them.
","featured_image_id":42,"intro_image_id":42,"featured_image_credit":null,"featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":3,"author_id":6,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":6,"view_count":19,"ordering":0,"published_at":"2026-06-21 10:53:00","publish_down":null,"created_by":4,"updated_by":1,"created_at":"2026-06-21 10:53:33","updated_at":"2026-06-26 15:13:03","key_takeaways":null,"faq":null,"sources":null,"category_name":"Insights","category_slug":"insights","author_name":"Sead Dzigal","author_slug":"sead-dzigal","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/5752db18c97718f817db7e36e9655dca.webp","featured_image_alt":"A humanoid robot reaches toward a tray of burgers in a futuristic restaurant.","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/06/5752db18c97718f817db7e36e9655dca.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Fast Technology, Slow Societies","meta_description":"In the last thirty years, we haven’t gotten new gadgets and toys, we’ve been actually brought into a completely different reality. First the…","robots":"index","canonical_url":"","og_title":"","og_description":""}]},{"category":{"id":20,"name":"Policy & Regulation","slug":"policy-regulation"},"posts":[{"id":178,"title":"Sweden Blocks ChatGPT and Claude on Government Work Computers","slug":"sweden-blocks-chatgpt-and-claude-on-government-work-computers","short_description":null,"content":"Sweden's Government Offices are restricting access to ChatGPT and Claude on employees' work computers from October 26, directing staff toward approved artificial intelligence services over concerns about information security.
\nThe decision, first reported by Swedish public broadcaster Sveriges Radio on October 23, affects the Government Offices (Regeringskansliet), the administrative organisation supporting the prime minister and government ministries.
\nEmployees will no longer be able to access the public web interfaces of OpenAI's ChatGPT and Anthropic's Claude from government-issued computers. Microsoft Copilot remains available as an approved alternative.
\nHowever, the restriction does not amount to a complete ban on either AI model. Some employees will retain access through an approved AI platform operating within the government's controlled environment.
\nThe Swedish Government Offices want employees to use AI services that have been formally reviewed and approved for official work.
\nAccording to Sveriges Radio, the decision follows internal concerns about staff using AI platforms without appropriate agreements or security controls.
\nMikael Wettercrantz, the Government Offices' director of digitalisation, warned in an internal communication that using AI services without agreements could have consequences for information security.
\nThe concern is particularly relevant to government institutions handling confidential documents, internal correspondence, policy proposals and potentially sensitive personal information.
\nWhen employees enter information into publicly accessible AI services, the organisation must understand how that information is processed, stored and protected.
\nWithout appropriate contractual arrangements and technical safeguards, institutions may have limited control over where information goes and who can access it.
\nThe Government Offices have reviewed Microsoft Copilot and strengthened its security configuration. However, officials have declined to disclose the technical details of those changes, citing security considerations.
\nThe decision does not establish that Copilot is inherently more secure than ChatGPT or Claude. It reflects the Government Offices' assessment of the services and configurations it has approved for workplace use.
\nAn important detail distinguishes Sweden's decision from an outright prohibition.
\nIn a statement provided to Swedish broadcaster SVT, the Government Offices confirmed that employees would retain access to ChatGPT and Claude language models through Intric, an approved AI platform.
\nIntric is a Swedish technology company specialising in AI infrastructure for public institutions and organisations handling sensitive information.
\nIn July 2026, the company announced that it had won a contract to supply its AI platform to the Swedish Government Offices.
\nThe platform allows organisations to develop and operate AI assistants using their own information under defined governance arrangements.
\nIntric says its infrastructure operates in Sweden, supports deployment within customers' own environments and provides institutions with greater control over their data.
\nThe arrangement illustrates the distinction between accessing an AI model through a public website and using the same underlying technology through an institutionally managed service.
\nIn Sweden's case, the restriction focuses on the access method and the organisation's ability to control information handling, rather than rejecting the technology itself.
\nThe Government Offices are not the first Swedish public institution to limit access to AI services.
\nAccording to Sveriges Radio's investigation, the Swedish Tax Agency (Skatteverket) and the National Board of Institutional Care (SiS) have already introduced similar restrictions.
\nThe Swedish Enforcement Authority (Kronofogden) is also reviewing the issue before deciding whether to introduce additional controls.
\nEarlier concerns emerged at the Swedish Transport Administration, where an internal audit identified widespread and potentially risky use of several AI services.
\nThe audit raised concerns about both the sensitivity of information employees entered into AI tools and how they subsequently handled AI-generated responses.
\nThese findings suggest that the problem extends beyond selecting an approved chatbot. Public institutions must also consider employee behaviour, access permissions and the reliability of information produced by AI systems.
\nSweden's Agency for Digital Government, Digg, has already issued guidance recommending that public authorities keep generative AI use under organisational supervision.
\nIts recommendations emphasise protecting confidential information, controlling access to institutional data and establishing clear rules about which information employees may process using AI services.
\nSweden's decision comes as public administrations across Europe are exploring how generative AI can support administrative work without exposing sensitive information or weakening accountability.
\nThe Swedish approach demonstrates one possible response: restricting uncontrolled access while maintaining AI capabilities through approved services.
\nFor government IT departments, this requires more than blocking websites. Institutions need clear policies governing permitted tools, data classification, supplier agreements, access controls and staff responsibilities.
\nThere is also a practical distinction between the security of an AI model and the security of the environment in which that model operates.
\nAn organisation may use the same underlying language model through different services, each with its own contractual arrangements, permissions and information-handling controls.
\nThis is particularly relevant to public institutions considering AI adoption without having established formal procurement procedures or internal usage policies.
\nThe Swedish decision raises a broader question for administrations elsewhere in Europe: how can employees benefit from increasingly capable AI systems while institutions retain effective control over sensitive information?
\nFor Sweden, the immediate change takes effect on October 26. What remains less clear is how the new restrictions will affect everyday government work and whether additional Swedish authorities will adopt comparable measures.
","featured_image_id":424,"intro_image_id":424,"featured_image_credit":"Meet Sweden, the AI incubator. Getty Images","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":15,"ordering":0,"published_at":"2026-10-26 08:33:46","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-26 08:33:46","updated_at":"2026-10-26 08:33:46","key_takeaways":"[\"Sweden's Government Offices are blocking direct browser access to ChatGPT and Claude on work computers from October 26, 2026. The restriction does not apply to the general public.\",\"Microsoft Copilot remains an approved workplace service, while access to ChatGPT and Claude models continues through the Intric platform. The decision therefore concerns how AI services are accessed and governed, rather than a complete prohibition of the models.\",\"Other Swedish authorities have introduced similar restrictions or are considering them, reflecting broader concerns about confidentiality, employee AI use and information-security controls.\"]","faq":"[{\"q\":\"Has Sweden banned ChatGPT and Claude?\",\"a\":\"No. Sweden's Government Offices are restricting direct access to the public web interfaces of ChatGPT and Claude on employees' work computers. The restriction is not a nationwide ban.\"},{\"q\":\"Why are the Swedish Government Offices blocking these AI services?\",\"a\":\"The stated reason is information security. Officials want employees to use approved services with appropriate contractual arrangements and security controls.\"},{\"q\":\"Can Swedish government employees still use AI?\",\"a\":\"Yes. Microsoft Copilot remains approved, and employees can access ChatGPT and Claude language models through Intric, an AI platform used by the Government Offices.\"},{\"q\":\"Are other Swedish government agencies introducing similar restrictions?\",\"a\":\"Yes. The Swedish Tax Agency and National Board of Institutional Care have introduced restrictions, while the Swedish Enforcement Authority is considering additional measures.\"}]","sources":"[{\"title\":\"Sveriges Radio, original investigation, 23 October 2026\",\"url\":\"https://www.sverigesradio.se/artikel/chat-gpt-och-claude-forbjuder-regeringskansliet\"},{\"title\":\"SVT Nyheter, government confirmation, updated 24 October\",\"url\":\"https://www.svt.se/nyheter/inrikes/ai-tjanster-sparras-pa-regeringskansliet\"},{\"title\":\"Swedish Agency for Digital Government, guidance on confidentiality and generative AI\",\"url\":\"https://www.digg.se/kunskap-och-stod/regler-och-rekommendationer/regler-och-rekommendationer/sakerstall-sekretess-vid-anvandning-av-generativ-ai\"}]","category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/38e2e86534210d0213839c4518c9443c.webp","featured_image_alt":"Sweden","featured_image_width":1440,"featured_image_height":960,"intro_image_path":"/api/uploads/2026/10/38e2e86534210d0213839c4518c9443c.webp","intro_image_width":1440,"intro_image_height":960,"seo_title":"Sweden Blocks ChatGPT and Claude on Government PCs","meta_description":"Sweden restricts ChatGPT and Claude on government work computers from October 26, citing security concerns. Approved AI services remain available.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":162,"title":"Europe’s AI Factories Face Access Limits","slug":"europe-s-ai-factories-face-access-limits","short_description":null,"content":"A review of European regulations and computing access programmes shows that all six Western Balkan economies have a route to Europe's AI infrastructure. High demand is already restricting access to some of its most powerful systems.
\nEurope's AI Factories promise startups, researchers and businesses access to supercomputing resources that would otherwise be prohibitively expensive. The Western Balkans are included in that ambition, but an examination of European Commission documents and EuroHPC access rules reveals an important qualification: eligibility does not guarantee available computing capacity.
\nAll six Western Balkan economies are associated with the EU's Horizon Europe research programme. Under Article 16 of the EuroHPC Regulation, users established in countries associated with Horizon Europe or the Digital Europe Programme are entitled to access the European Union's share of computing time on EuroHPC supercomputers acquired after 2020, subject to the applicable access conditions.
\nThat includes organisations based in Albania, Bosnia and Herzegovina, Kosovo, Montenegro, North Macedonia and Serbia.
\nThe opportunity is substantial. European programmes offer qualifying companies free access to GPU resources for AI development, testing and research.
\nBut EuroHPC's own notices show that demand is already affecting availability. Some computing systems are temporarily unavailable for new allocations, raising practical questions about how smaller companies can obtain the resources they need.
\nMuch of the regional attention has focused on North Macedonia and Serbia, which were selected in October 2025 to host AI Factory Antennas.
\nNorth Macedonia's VEZILKA initiative and Serbia's SAIFA are intended to connect local organisations with the European AI Factory network, including technical expertise, computing infrastructure and related services.
\nNeither country, however, has exclusive access to the network.
\nThe European Commission's list of Horizon Europe associated countries includes all six Western Balkan economies. Article 16(4) of Regulation (EU) 2021/1173 provides an access entitlement to the Union's share of computing time on EuroHPC systems acquired after 2020.
\nThe distinction is particularly relevant for Kosovo.
\nAlthough Kosovo is not named in one of EuroHPC's general AI Factory eligibility lists, it is officially associated with Horizon Europe. The broader regulation therefore provides a legal basis for access to qualifying computing resources.
\nThis does not remove the need to satisfy individual programme requirements. Applicants must still meet the conditions of the relevant call, including technical suitability and the proposed use of the infrastructure.
\nIt does mean that the absence of a national AI Factory or antenna should not automatically be understood as exclusion from European computing resources.
\nEuroHPC offers several access routes for industrial innovation, depending on the size and complexity of a project.
\nThe Playground programme is aimed at startups, small businesses and organisations beginning to work with advanced computing. It allows smaller AI experiments to receive allocations lasting one, two or three months.
\nApplications are processed on a first-come, first-served basis. EuroHPC targets access within two working days after successful eligibility and technical assessments.
\nThe Fast Lane programme serves companies with more demanding requirements, allowing applications for up to 50,000 GPU hours.
\nFor projects requiring greater capacity, the Large Scale programme offers allocations exceeding 50,000 GPU hours, with project durations of three, six or twelve months.
\nQualifying research and innovation activities can receive access without computing fees. Commercial access outside those conditions may be charged.
\nThese arrangements create a realistic opportunity for regional developers working on language models, scientific applications and other computationally demanding AI systems.
\nA company in Albania developing an Albanian-language model, for example, does not necessarily need to purchase its own high-performance hardware or wait for Albania to establish an AI Factory.
\nIt can explore the relevant European access programmes and apply for computing time.
\nWhether its application succeeds depends on the project requirements and available resources.
\nThe clearest limitation appears in EuroHPC's published access notices.
\nOn its Large Scale access page, EuroHPC states that the MareNostrum and LUMI-G resources are unavailable for new allocations because of exceptionally high demand.
\nIts Fast Lane programme also identifies the MareNostrum5 ACC partition as unavailable for new Fast Lane allocations.
\nApplicants are advised to select alternative systems when their preferred computing resources cannot be accessed.
\nThese restrictions do not mean that Europe's entire AI Factory network has reached capacity. Other systems remain available through the relevant programmes.
\nThey do demonstrate that access is constrained by actual hardware availability, even when an organisation meets the formal eligibility requirements.
\nFor smaller businesses, choosing an alternative system may require adjustments to software environments, project schedules or computing requirements.
\nEuroHPC also publishes limits on the maximum GPU hours that can be allocated on individual systems over specified periods. Large projects must take those constraints into account when planning their applications.
\nThe result is a system in which free access is possible but cannot be treated as unlimited or guaranteed.
\nThe AI Factory Antennas in North Macedonia and Serbia introduce another important dimension.
\nTheir purpose is to bring the services of established European AI Factories closer to national research and business ecosystems.
\nThis can include connecting companies with technical expertise and supporting their use of European AI infrastructure.
\nFor a startup with limited experience in high-performance computing, such assistance could be valuable.
\nHowever, the existence of these antennas does not mean that companies in North Macedonia and Serbia automatically receive preferential GPU allocations.
\nNor does it establish that organisations in Albania, Bosnia and Herzegovina, Kosovo or Montenegro receive less computing time.
\nThe public documents examined do not provide sufficient country-by-country allocation figures to support either conclusion.
\nWhat can be established is that the European network combines centralised computing infrastructure with a developing set of national and regional support structures.
\nThe ability of smaller companies to use that infrastructure will depend partly on how effectively those structures connect applicants with available resources.
\nThere is a difference between announcing access to infrastructure and demonstrating that organisations are using it.
\nEuroHPC publishes the application procedures, eligibility criteria and technical arrangements governing its programmes.
\nThe material reviewed for this article does not establish how many Western Balkan organisations have successfully obtained AI Factory computing allocations, or how GPU hours have been distributed among the six economies.
\nThose figures would provide a clearer picture of the programme's practical reach.
\nA country could be fully eligible under European rules while relatively few of its businesses make use of the available resources. Reasons might include limited awareness, technical capacity, project readiness or infrastructure availability.
\nWithout application and allocation data, none of those explanations can be established as the principal cause.
\nThe documented capacity restrictions are nevertheless relevant. They demonstrate that formal access and actual resource availability are different parts of the same system.
\nThe European Commission has presented AI Factories as an important part of Europe's effort to develop its own AI capabilities and support innovation.
\nFor the Western Balkans, the programme provides access to infrastructure on a scale that smaller national economies would struggle to finance independently.
\nThe legal framework offers a route for organisations across all six economies, while the antennas in North Macedonia and Serbia are intended to bring support services closer to local users.
\nThe next question is measurable: how many regional startups and research institutions will actually receive computing time, and what will they produce with it?
\nEuroHPC's published restrictions already show that demand can limit access to particular systems.
\nFor the Western Balkans, the real value of Europe's AI Factories will be determined by successful applications, usable computing capacity and completed projects, rather than the number of infrastructure partnerships announced.
","featured_image_id":367,"intro_image_id":367,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":7,"view_count":39,"ordering":0,"published_at":"2026-10-11 09:24:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-11 09:24:38","updated_at":"2026-10-11 09:25:21","key_takeaways":"[\"All six Western Balkan economies are associated with Horizon Europe, providing a legal route to qualifying EuroHPC computing resources.\",\"Eligible startups can apply for free computing time through programmes such as Playground, Fast Lane and Large Scale.\",\"EuroHPC has reported high-demand restrictions on certain computing systems, showing that formal access does not guarantee capacity.\"]","faq":"[{\"q\":\"Can companies from all six Western Balkan economies access AI Factories?\",\"a\":\"The EuroHPC legal framework provides an access route for users in Horizon Europe associated countries on systems acquired after 2020. Individual application requirements and restrictions still apply.\"},{\"q\":\"Is access completely free?\",\"a\":\"Qualifying research and SME innovation activities can use computing resources without access fees. Other commercial uses may be chargeable.\"},{\"q\":\"Does Kosovo qualify?\",\"a\":\"Kosovo is associated with Horizon Europe and falls within the country category covered by Article 16(4) for eligible EuroHPC systems. Specific calls may apply additional conditions.\"},{\"q\":\"Why are some supercomputers unavailable?\",\"a\":\"EuroHPC reports exceptionally high demand for selected resources, including LUMI-G and MareNostrum-related partitions.\"}]","sources":"[{\"title\":\"EuroHPC Regulation, Articles 16–18\",\"url\":\"https://eur-lex.europa.eu/eli/reg/2021/1173/2026-01-20/eng\"},{\"title\":\"European Commission, Horizon Europe associated countries\",\"url\":\"https://research-and-innovation.ec.europa.eu/news/all-research-and-innovation-news/updates-association-third-countries-horizon-europe-2021-12-21_en\"},{\"title\":\"EuroHPC, Playground Access\",\"url\":\"https://www.eurohpc-ju.europa.eu/playground-access-ai-factories_en\"},{\"title\":\"EuroHPC, Fast Lane Access\",\"url\":\"https://www.eurohpc-ju.europa.eu/fast-lane-access-ai-factories_en\"},{\"title\":\"EuroHPC, Large Scale Access\",\"url\":\"https://www.eurohpc-ju.europa.eu/large-scale-access-ai-factories_en\"},{\"title\":\"European Commission, AI Factories\",\"url\":\"https://digital-strategy.ec.europa.eu/en/policies/ai-factories\"}]","category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/31034a2a772706065f08a3397f88a8cb.webp","featured_image_alt":"AI Antenna","featured_image_width":1200,"featured_image_height":800,"intro_image_path":"/api/uploads/2026/10/31034a2a772706065f08a3397f88a8cb.webp","intro_image_width":1200,"intro_image_height":800,"seo_title":"Europe’s AI Factories Face Access Limits","meta_description":"All six Western Balkan economies have routes to European AI computing resources, but capacity restrictions are already limiting availability.","robots":"index","canonical_url":"eu-ai-factories-western-balkans-access-limits","og_title":"","og_description":""},{"id":105,"title":"EU Delays Key AI Act Rules Until 2027 and 2028","slug":"eu-delays-key-ai-act-rules-until-2027-and-2028","short_description":"","content":"The European Union has pushed back some of the most demanding parts of its AI Act, giving companies more than a year of additional time to comply with rules covering high-risk artificial intelligence systems.
\nUnder the EU's new AI Omnibus regulation, rules for high-risk AI used in areas such as employment, education, critical infrastructure, biometrics and migration will now apply from 2 December 2027.
\nAI systems built into regulated physical products, including machinery and other equipment covered by EU product-safety legislation, have been given an even longer transition period. Those requirements move to 2 August 2028.
\nThe change is substantial.
\nMany of those obligations had been expected to start applying in August 2026 under the original AI Act timetable.
\nBut it would be wrong to describe the move as a delay of the AI Act itself.
\nLarge parts of Europe's AI rulebook are already in force.
\nThe timing can be confusing because the EU has been implementing the AI Act in stages.
\nRules banning certain unacceptable AI practices began applying earlier. Obligations for general-purpose AI models followed, while transparency requirements covering areas such as AI-generated content began applying on 2 August 2026.
\nThose rules have not disappeared because the high-risk timetable changed.
\nThe delay specifically affects important parts of the regime governing systems classified as high-risk under Annex III and AI systems covered through regulated products under Annex I.
\nThat means an AI company could already face obligations under one part of the Act while having until late 2027 to meet another.
\nFor businesses trying to understand whether they are compliant, that distinction matters more than the headline.
\nSome of the systems receiving additional time are among the most sensitive uses of AI covered by European law.
\nThey include certain applications involving biometrics, critical infrastructure, education, employment, migration, asylum and border control.
\nFor example, an AI system used to evaluate job applicants or influence decisions about access to education may fall into the high-risk framework depending on how it is used.
\nThose systems face requirements covering areas such as risk management, data governance, technical documentation, record keeping, human oversight, accuracy and cybersecurity.
\nThe new deadline for these Annex III systems is 2 December 2027.
\nThe Commission says the additional time is partly intended to solve a practical problem: companies were being asked to comply before some of the technical standards needed to demonstrate compliance were ready.
\nEuropean standards organisations CEN and CENELEC have been working on harmonised standards for the AI Act, but that process did not progress according to the original schedule.
\nThose standards are voluntary, but they matter.
\n![\"EU](\"/api/uploads/2026/07/8b45ee0395ff9bbea1d0c30c3f13e7d6.webp\")
A company following an approved harmonised standard can gain a presumption that its system complies with the corresponding legal requirements. Without those standards, both companies and regulators have less certainty about what compliance should look like in practice.
The EU chose to move the deadline rather than force the high-risk regime into effect before that framework was sufficiently mature.
\nA separate category receives even more time.
\nHigh-risk AI systems that are safety components of products already governed by EU harmonisation legislation will not face the relevant AI Act requirements until 2 August 2028.
\nThis can include AI incorporated into products such as machinery and other regulated equipment.
\nThe reason for the longer transition is that these systems sit at the intersection of AI regulation and existing product-safety law.
\nManufacturers may need to adjust testing, conformity assessments, documentation and certification processes across both regimes.
\nThe EU is effectively giving that part of the market another year before the requirements become fully applicable.
\nThe AI Omnibus does more than change dates.
\nThe legislation is part of a wider attempt by Brussels to simplify the digital regulatory framework following sustained complaints from European businesses about the growing cost and complexity of compliance.
\nAmong the changes, simplified obligations previously aimed primarily at small and medium-sized enterprises are being extended to some small mid-cap companies.
\nThe Omnibus also removes the requirement for providers to register certain systems that are exempt from the high-risk classification in the EU's central database.
\nA planned requirement for a harmonised post-market monitoring plan has also been removed, giving companies more flexibility in how they structure monitoring after deployment.
\nAnother change concerns AI literacy.
\nThe previous direct AI-literacy obligation on companies has been replaced by a softer approach centred more heavily on encouragement and measures promoted by the Commission and EU member states.
\nThese are not merely adjustments to dates.
\nThey represent a broader shift in how Brussels is trying to balance AI regulation with pressure to make Europe easier to build and invest in.
\nThe simplification package also adds restrictions.
\nThe Omnibus introduces a prohibition targeting AI systems designed to generate non-consensual sexually explicit or intimate material, including so-called nudification applications, as well as systems producing child sexual abuse material.
\nIt also allows processing of certain sensitive categories of personal data when necessary to detect and correct bias in AI systems, subject to safeguards.
\nThe direction is therefore not straightforward deregulation.
\nEurope is easing some administrative requirements and buying companies additional implementation time while retaining, and in some areas expanding, controls over uses it considers particularly harmful.
\nFor companies, the risk created by the delay is assuming that nothing needs to happen until 2027.
\nThat would be a mistake.
\nWhether an organisation has obligations today depends on what AI it provides, where it is used and which part of the Act applies.
\nA company offering a general-purpose AI model faces a different timetable from an employer deploying an automated recruitment system. A manufacturer integrating AI into regulated machinery has another timetable again.
\nTransparency requirements can already apply even when the corresponding high-risk obligations do not.
\nThe practical starting point is therefore classification.
\nCompanies need to understand whether they are acting as a provider, deployer, importer or distributor under the Act, what type of system they are dealing with, and which application date applies to it.
\nThe extra time is most valuable to organisations that use it to build documentation, governance and technical controls before compliance becomes mandatory.
\nThe deadlines also matter outside EU member states.
\nTechnology companies in North Macedonia, Serbia, Albania, Bosnia and Herzegovina, Montenegro and Kosovo increasingly develop software for European customers or work as part of EU-facing supply chains.
\nThe AI Act can therefore become relevant even when the developer itself is based outside the Union.
\nFor Western Balkan companies planning AI products for the European market, the additional time creates breathing room.
\nIt should not be mistaken for an exemption.
\nCountries in the region are also progressively aligning digital policy and regulation with the European Union, making the AI Act important beyond companies that are directly caught by its territorial scope.
\nThe new timetable gives businesses a clearer date to work toward.
\n2 December 2027 is now the major deadline for many high-risk AI applications involving people and public services.
\n2 August 2028 is the corresponding date for high-risk AI embedded in regulated products.
\nBetween now and then, the details will matter as much as the dates. European standards still need to mature, regulators need to develop enforcement practice, and companies need to establish which of their systems actually fall inside the high-risk regime.
\nThe AI Act is still coming.
\nFor some of its toughest requirements, Europe has simply decided it needs more time to make the rules work.
","featured_image_id":214,"intro_image_id":214,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":7,"view_count":51,"ordering":0,"published_at":"2026-08-12 14:02:01","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-12 14:02:01","updated_at":"2026-08-12 14:02:01","key_takeaways":null,"faq":null,"sources":null,"category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/ef5edfa73c5032cd29c7500382090100.webp","featured_image_alt":"AI, EU Flag and robot","featured_image_width":1600,"featured_image_height":1030,"intro_image_path":"/api/uploads/2026/07/ef5edfa73c5032cd29c7500382090100.webp","intro_image_width":1600,"intro_image_height":1030,"seo_title":"EU Delays Key AI Act Rules Until 2027 and 2028","meta_description":"The EU has delayed major high-risk AI Act rules until December 2027 and August 2028. Here is what changed and which rules still apply.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":94,"title":"Europe’s AI Act Now Has Investigators","slug":"europe-s-ai-act-now-has-investigators","short_description":"","content":"For years, Europe’s approach to artificial intelligence was discussed mainly through draft laws, risk categories, voluntary codes and future deadlines.
\nThat phase is ending.
\nFrom August 2, 2026, the European Commission’s AI Office and national authorities begin exercising key enforcement powers under the EU AI Act. Brussels is also strengthening the teams responsible for investigating possible violations, examining powerful general-purpose AI models and responding when AI systems create serious risks.
\nThe AI Act is no longer only a rulebook. It is becoming an enforcement system.
\nThe European Commission’s AI Office was established to coordinate implementation of the AI Act across the European Union.
\nIts most important direct responsibility concerns general-purpose AI models, including the foundation models behind many chatbots, coding assistants, content generators and AI agents.
\nUntil now, most public attention has focused on what companies must disclose and which systems fall into different risk categories. The harder questions begin now.
\nWho decides whether a model presents a systemic risk? What evidence must a company provide? How will regulators test systems that are constantly updated? And what happens when a model behaves differently in real-world use than it did during internal testing?
\nThe AI Office includes teams focused on regulation, compliance and AI safety. Their work will include examining possible infringements, assessing model risks and supporting enforcement across the EU.
\nThe Commission has also appointed an independent scientific panel to assist regulators with model evaluation and systemic-risk assessment.
\nThat matters because enforcing AI rules cannot depend only on lawyers reviewing company documents. Regulators will increasingly need technical experts capable of testing models, examining their capabilities and challenging claims made by developers.
\nThe EU’s attention is not limited to visible deepfakes or chatbots that fail to identify themselves.
\nRegulators are also concerned about broader risks linked to advanced models, including their potential use in cyberattacks, manipulation, illegal content generation and other activities that may affect public safety or fundamental rights.
\nThe European Commission has direct authority over certain general-purpose AI providers, while national authorities remain responsible for many AI systems deployed within individual member states.
\nThis shared structure reflects how modern AI systems operate.
\nA model may be developed by a company in the United States, hosted in another country, integrated into software by a European business and then used by citizens across several jurisdictions.
\nNo single national regulator can easily supervise that entire chain.
\nEurope is therefore building a system in which central institutions examine the most powerful models, while national authorities supervise how AI is used in workplaces, public services, healthcare, education, policing and other sensitive areas.
\nThe AI Office also operates a whistleblower mechanism through which people connected to AI providers can report suspected violations.
\nEmployees, contractors and business partners may have access to internal evaluations, safety failures or company decisions that never become public.
\nExternal researchers and journalists usually see only the final product. Insiders may know whether a company ignored a serious risk, concealed a failed test or released a model before safety concerns were resolved.
\nA confidential reporting channel gives regulators another way to identify problems before they grow into larger incidents.
\nIts success, however, will depend on trust. People must believe that reports will be treated seriously, their identities will be protected and the AI Office has the capacity to investigate the information it receives.
\nPassing a law is easier than enforcing it against some of the world’s largest technology companies.
\nMajor AI providers have strong legal teams, large technical departments and substantial financial resources. Their systems are also complex enough that regulators may struggle to independently reproduce company evaluations or determine whether a safety measure is genuinely effective.
\nEurope must avoid two opposite failures.
\nThe first would be weak enforcement, where companies submit large amounts of documentation but face little technical scrutiny.
\nThe second would be enforcement that becomes so slow and bureaucratic that smaller European developers carry a heavier burden than the global corporations the law was designed to oversee.
\nThe credibility of the AI Act will depend on whether regulators can distinguish between real risk and minor administrative mistakes.
\nA startup that makes an honest reporting error should not be treated in the same way as a powerful provider that hides a serious capability or repeatedly ignores safety obligations.
\nProportionate enforcement will be one of the biggest challenges.
\nWestern Balkan countries are outside the European Union, but they will not remain outside the influence of this enforcement system.
\nRegional companies provide software, outsourcing, data services, technical support and digital products to clients across the EU. Some operate AI tools that can be accessed by European users. Others integrate models developed by global providers into services used by banks, schools, public institutions and businesses.
\nThe question will no longer be only whether an organisation uses AI.
\nCompanies may increasingly need to know which model they use, where data is processed, how outputs are reviewed, what risks were identified and who is responsible when the system fails.
\nThis will matter particularly for organisations working in regulated sectors or participating in European projects.
\nIt will also matter for governments in the region. Candidate countries seeking closer alignment with the EU will face growing pressure to develop compatible laws, institutions and oversight mechanisms.
\nCopying the language of the AI Act will be the easy part.
\nFinding qualified inspectors, technical auditors, legal experts and independent researchers will be much harder.
\nThe Balkans already face shortages of specialised digital expertise in many public institutions. AI enforcement will require even more advanced knowledge, combining law, cybersecurity, data protection and machine-learning evaluation.
\nWithout that capacity, national laws may exist mostly on paper.
\nThe new AI disclosure labels will be the most visible change for ordinary users.
\nPeople may notice a warning under a synthetic video, a notice attached to AI-generated content or a chatbot explaining that it is not human.
\nBut the deeper change will happen away from the screen.
\nEurope is building institutions that can request information, investigate companies, examine advanced models and impose consequences when obligations are ignored.
\nWhether those institutions will be fast enough, technically capable enough and independent enough remains uncertain.
\nThe AI Act has already given Europe one of the world’s most ambitious legal frameworks for artificial intelligence.
\nFrom August 2, the more difficult phase begins: proving that the rules can actually be enforced.
","featured_image_id":126,"intro_image_id":126,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":1,"show_table_of_contents":0,"read_time_minutes":6,"view_count":50,"ordering":0,"published_at":"2026-08-01 11:12:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-01 11:12:23","updated_at":"2026-10-01 18:06:40","key_takeaways":null,"faq":null,"sources":null,"category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/2604f5de8c500cc8eec9eb5ede7801c1.webp","featured_image_alt":"The European Union flag waves against a blue, clouded sky.","featured_image_width":1600,"featured_image_height":936,"intro_image_path":"/api/uploads/2026/07/2604f5de8c500cc8eec9eb5ede7801c1.webp","intro_image_width":1600,"intro_image_height":936,"seo_title":"Europe’s AI Act Now Has Investigators","meta_description":"Europe moves from AI regulation to enforcement as the EU AI Office gains powers to investigate companies, advanced models and systemic risks.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":90,"title":"Europe’s AI Disclosure Rules Begin on August 2","slug":"europe-s-ai-disclosure-rules-begin-on-august-2","short_description":"","content":"The new requirements will not place a warning on everything created with AI. They will, however, make it harder for chatbots, deepfakes and automated public communication to quietly pass as human or authentic.
\nFrom August 2, some of Europe’s most visible uses of artificial intelligence will have to become more transparent.
\nPeople in the European Union must be informed when they are communicating directly with certain AI systems or encountering particular forms of AI-generated and manipulated content. The obligations come from Article 50 of the EU AI Act and enter into application on August 2, 2026.
\nThe European Commission published its final implementation guidelines on July 27, only days before the rules take effect. They clarify how the requirements apply to chatbots, synthetic images and audio, deepfakes and AI-generated text dealing with matters of public interest.
\n![\"AI,](\"/api/uploads/2026/07/ef5edfa73c5032cd29c7500382090100.webp\")
The change does not mean that every email corrected by an AI tool, every edited photograph or every article drafted with some AI assistance must carry a large “AI-generated” label.
\nThe rules are more specific than that.
\nThey target situations in which people may reasonably believe that artificial content, communication or behaviour is authentic.
\nCompanies and public institutions using AI systems that communicate directly with individuals will generally need to disclose that the person is interacting with a machine.
\nThe notice should be clear and provided early in the interaction, unless it is already obvious from the circumstances that the system is artificial.
\nThis affects customer-support chatbots, digital assistants and other interactive AI services. It also challenges a common design practice in which organisations give automated systems human names, photographs and conversational personalities while leaving users to discover for themselves that no person is present.
\nThe issue is not whether an AI assistant sounds friendly. It is whether the user understands who, or what, is answering.
\nThe rules also require disclosure when AI is used to create or manipulate images, audio or video that could falsely appear authentic.
\nThis includes deepfakes representing real or plausible people, places, objects, organisations and events.
\nThe disclosure must be understandable to the audience. Machine-readable metadata may help platforms and technical systems identify synthetic material, but hidden information inside a file will not always be enough for the person watching or listening.
\nThere are more flexible provisions for artistic, fictional, satirical and creative works. That does not create a broad exemption for synthetic political communication, fabricated institutional announcements or manipulated footage presented as evidence of a real event.
\nThis distinction matters far beyond Brussels.
\nAI-generated campaign photographs, synthetic public messages and manipulated political content are already appearing across the Western Balkans, often without consistent labelling or even a basic explanation of how the material was produced.
\nA creator may consider the fiction obvious. A citizen encountering a cropped image or reposted video days later may not.
\nOne of the more sensitive requirements concerns AI-generated or manipulated text published to inform the public about matters of public interest.
\nThis can include communication relating to elections, public administration, health, justice, law enforcement, public safety and other issues that shape public debate.
\nSuch material may require disclosure when it is generated or substantially manipulated by AI and published without meaningful human review or editorial responsibility.
\nThis is particularly relevant to news organisations, government institutions, political parties and communications agencies experimenting with automated publishing.
\nThe regulation is not designed to punish a journalist for using transcription, translation, spelling correction or another standard editing tool. The Commission’s guidelines explicitly recognise exceptions such as standard editing.
\nThe harder question is what happens when AI produces most of a public statement, news article or institutional announcement and a person merely approves it without properly checking the facts, sources or meaning.
\nA human click on the publish button does not automatically transform automated content into responsible communication.
\nThe real test is whether someone with the authority and competence to intervene has reviewed the content, can change or reject it and accepts responsibility for what is published.
\nProviders of generative AI systems are expected to make synthetic content detectable, including through machine-readable marking.
\nThe organisations and professionals deploying those systems may carry separate responsibilities to disclose deepfakes or certain AI-generated public-interest publications to their audiences.
\nThe Commission has also introduced a voluntary Code of Practice covering the marking and labelling of AI-generated content. Joining the code is optional, but the underlying Article 50 transparency obligations are not. Companies that choose another compliance method must be able to demonstrate that it is adequate.
\nThe EU has additionally prepared optional icons that organisations may use when labelling artificial content.
\nStill, no icon can solve the entire problem.
\nA label may tell the audience that AI was involved. It does not reveal whether the content is accurate, whether the sources were checked or whether anyone is prepared to take responsibility for mistakes.
\nThe AI Act applies within the European Union, but its influence will extend into neighbouring markets.
\nTechnology companies in the Western Balkans that provide AI products or services to EU users may fall within its reach. Regional businesses working with European clients will also increasingly encounter contractual demands for compliant marking, disclosure and documentation.
\nFor candidate countries, the rules are also likely to shape future legislation and institutional practice.
\nBut the more immediate pressure may come from the public rather than regulators.
\nMedia outlets that publish automated reporting without proper review, institutions that use synthetic photographs without explanation and political actors that circulate manipulated material may not face identical legal consequences across the region.
\nThey will still face the same question from their audiences: why was this presented as real?
\nEurope’s new rules will not end deceptive AI content on August 2. They will not prevent every deepfake or force every AI-assisted sentence to carry a warning.
\nWhat they do is establish a clearer principle.
\nArtificial intelligence can support public communication, but it should not quietly impersonate a person, an event, an institution or a newsroom.
","featured_image_id":125,"intro_image_id":125,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":6,"view_count":27,"ordering":0,"published_at":"2026-07-29 08:27:28","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-29 08:27:28","updated_at":"2026-07-29 08:27:28","key_takeaways":null,"faq":null,"sources":null,"category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/8b45ee0395ff9bbea1d0c30c3f13e7d6.webp","featured_image_alt":"EU Artificial Intelligence Act, with servers in the background, and European stars.","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/07/8b45ee0395ff9bbea1d0c30c3f13e7d6.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Europe’s AI Disclosure Rules Begin on August 2","meta_description":"EU rules covering AI chatbots, deepfakes and certain AI-generated public-interest content take effect on August 2, 2026.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":73,"title":"China Considers Limits on Overseas Access to Advanced AI Models","slug":"china-considers-limits-on-overseas-access-to-advanced-ai-models","short_description":"","content":"China is reportedly considering new restrictions on overseas access to some of its most advanced artificial intelligence models, a move that could reshape how Chinese AI companies offer frontier AI technology to customers around the world.
\nAccording to a report by The Wall Street Journal, Chinese authorities are discussing whether tighter controls should apply to the international availability of high-end AI models developed by companies including Alibaba, ByteDance and Z.ai. The discussions are ongoing, and no formal policy has been announced.
\nIf the proposals move forward, they would mark a notable change in China's approach to artificial intelligence.
\nUntil now, Chinese technology companies have steadily expanded the global reach of their AI services, positioning themselves as competitive alternatives to Western providers. Lower costs, increasingly capable models and open access have helped attract developers, startups and businesses looking beyond the dominant platforms from OpenAI, Google and Anthropic.
\nThe reported discussions suggest Beijing may now be weighing a different priority: whether its most advanced AI systems should be treated as strategic assets rather than products freely available on the global market.
\nThe timing is significant.
\nOver the past two years, governments have become increasingly involved in shaping who can build, access and export advanced AI technologies. The United States has tightened restrictions on advanced AI chips and high-performance computing exports, while several Western AI companies have introduced regional limitations for certain services in response to regulatory or security concerns.
\nIf China follows a similar path, it would signal that access to frontier AI is becoming another area where technology and geopolitics are closely intertwined.
\nFor businesses, the implications extend beyond China itself.
\nCompanies building products on top of AI models often make decisions based on performance, pricing and reliability. But if governments begin influencing where advanced models can be deployed, those decisions may also become subject to political and regulatory considerations.
\nThat is particularly relevant for startups and software companies across the Balkans, where many teams rely on internationally available AI models to develop products, automate workflows and experiment with new applications. As the number of capable AI providers grows, long-term access to those platforms could become just as important as benchmark performance or pricing.
\nAt this stage, there is no indication that China has reached a final decision. The discussions could evolve, change direction or end without new regulations. Even so, the reports reflect a broader shift taking place across the global AI industry.
\nThe conversation is no longer focused only on which company builds the most capable model. Increasingly, it is also about who controls access to that technology—and under what conditions.
","featured_image_id":155,"intro_image_id":155,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":30,"ordering":0,"published_at":"2026-07-09 12:00:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-09 12:00:10","updated_at":"2026-07-09 12:01:12","key_takeaways":null,"faq":null,"sources":null,"category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/e190d8008d34e5819c806c5c84b1c94b.webp","featured_image_alt":"China AI and flag in the background","featured_image_width":1280,"featured_image_height":720,"intro_image_path":"/api/uploads/2026/07/e190d8008d34e5819c806c5c84b1c94b.webp","intro_image_width":1280,"intro_image_height":720,"seo_title":"China Weighs Limits on Overseas Access to AI Models","meta_description":"China is weighing limits on overseas access to advanced AI models, signalling a new phase in global AI policy and competition.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":71,"title":"AI Enters Western Balkan Governments Before Rules Are Ready","slug":"ai-enters-western-balkan-governments-before-rules-are-ready","short_description":"","content":"Artificial intelligence is already being used inside public institutions across the Western Balkans, even as most governments are still developing the legal and institutional frameworks needed to regulate its use.
\nAn investigation by Radio Free Europe/Radio Liberty (RFE/RL) found that public authorities in Bosnia and Herzegovina, Montenegro, Serbia and Kosovo have begun using generative AI tools, including ChatGPT, Microsoft Copilot and Google Gemini, for everyday administrative work ranging from drafting documents and translating texts to researching information and supporting internal processes.
\nThe findings suggest that AI adoption is moving faster than governance across much of the region. While public institutions are experimenting with new technologies to improve efficiency, clear policies defining how these systems should be used—and what information should never be shared with them—often remain absent.
\nFor governments across the Western Balkans, artificial intelligence has already moved beyond pilot discussions and digital transformation strategies.
\n![\"wb.png\"](\"/api/uploads/2026/07/ffd47f35b4a8354d2378a996e4b88451.webp\")
According to RFE/RL's reporting, a growing number of public institutions acknowledge that employees are using AI-powered tools as part of their daily work. In many cases, these systems assist with administrative tasks, drafting official correspondence, proofreading documents, summarising lengthy reports and translating content.
At the same time, the investigation found that many institutions do not maintain formal records of AI use, making it difficult to assess how widely these technologies have been adopted or whether consistent internal practices exist across government bodies.
\nThe lack of documentation also makes oversight more challenging as AI becomes increasingly integrated into routine public administration.
\nThe investigation highlights significant differences across the region.
\nSerbia appears to be among the most advanced in applying AI within public administration. AI-powered virtual assistants are already available through several government services, including the eGovernment portal, the Serbian Business Registers Agency and the Tax Administration. Authorities are also preparing legislation intended to align the country's regulatory framework with the European Union's AI Act while building on Serbia's existing National AI Strategy.
\nMontenegro is developing its Artificial Intelligence Development Strategy for the 2026–2030 period. The government is also working with the World Bank on a roadmap aimed at introducing AI more systematically into public administration as part of broader digital transformation reforms linked to the country's EU accession process.
\nIn Bosnia and Herzegovina, AI adoption is taking place without a dedicated legal framework governing the technology in public administration. The country's complex institutional structure presents additional challenges for creating a unified national approach to AI governance.
\nKosovo continues expanding digital public services, but dedicated legislation regulating the use of artificial intelligence by state institutions has yet to be introduced. While AI is recognised as part of broader digital transformation efforts, governance mechanisms remain at an early stage.
\nAlthough North Macedonia was not part of the RFE/RL investigation, similar discussions are taking place across the country as public institutions, universities and policymakers increasingly examine how AI can be introduced responsibly while remaining aligned with emerging European standards.
\nThe investigation identifies a common concern shared across the region: governance has not kept pace with adoption.
\n![\"westernbalkans.webp\"](\"/api/uploads/2026/07/7cdaafd065aeab4ab7759f52ea3384ac.webp\")
Experts interviewed by RFE/RL point to several unresolved issues, including the protection of personal data, the handling of confidential government information and the reliability of AI-generated content used in official work.
Unlike traditional software, generative AI systems can produce inaccurate or fabricated information while presenting it with confidence. That makes human verification essential whenever AI contributes to official documents, legal analysis or public communication.
\nQuestions also remain over accountability. If AI-generated information influences a government decision or contains sensitive data, responsibility ultimately rests with the institution and the public official—not the AI system itself.
\nThe regional debate comes as the European Union begins implementing the AI Act, the world's first comprehensive legal framework for artificial intelligence.
\nAlthough most Western Balkan countries are not yet subject to the regulation, EU candidate countries are expected to gradually align their legislation with European standards as part of the accession process.
\nFor governments across the region, that means AI policy is becoming more than a technology issue. It is increasingly linked to public administration reform, cybersecurity, transparency and public trust.
\nThe challenge is no longer whether artificial intelligence will be used inside government institutions.
\nThat transition has already begun.
\nThe next phase will depend on whether legislation, internal policies and public-sector training can develop quickly enough to ensure AI is deployed safely, responsibly and with clear accountability.
\nSource: Radio Free Europe/Radio Liberty (RFE/RL) investigation
","featured_image_id":150,"intro_image_id":150,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":41,"ordering":0,"published_at":"2026-07-08 09:46:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-08 09:42:48","updated_at":"2026-08-01 11:29:40","key_takeaways":null,"faq":null,"sources":null,"category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/ea6e63682d1ae702676287f0b16d11de.webp","featured_image_alt":"The Balkans Connecting","featured_image_width":1448,"featured_image_height":1086,"intro_image_path":"/api/uploads/2026/07/ea6e63682d1ae702676287f0b16d11de.webp","intro_image_width":1448,"intro_image_height":1086,"seo_title":"AI Enters Western Balkan Governments Before Rules Are Ready","meta_description":"Public institutions across the Western Balkans are already using AI tools, while governance, legal frameworks and internal rules continue to lag behind.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":66,"title":"UN Calls for Ban on Autonomous AI Weapons","slug":"un-calls-for-ban-on-autonomous-ai-weapons","short_description":"","content":"The United Nations has renewed its call for a global ban on lethal autonomous weapons, warning that artificial intelligence must not be allowed to decide who lives and who dies.
\nUN Secretary-General António Guterres said systems capable of selecting and attacking targets without meaningful human control should be prohibited through a legally binding international treaty. His message comes as governments face growing pressure to define clear limits for the military use of AI.
\nAutonomous weapons, often described as “killer robots,” are no longer a distant policy debate. AI is already being used in surveillance, targeting support, drone systems, battlefield analysis, and military decision-making. The concern is not only whether these systems can be accurate, but whether they should ever be given authority over lethal force.
\nFor the UN, the answer is clear: human judgment must remain central in any decision to use deadly force.
\nThe debate is becoming more urgent as artificial intelligence moves faster than international regulation. Humanitarian groups and arms-control advocates have warned that without binding rules, countries could enter a new arms race built around increasingly autonomous military systems.
\nThe legal question is equally serious. If an AI-enabled weapon makes a fatal mistake, responsibility becomes difficult to assign. Is the commander accountable? The state? The developer? The company that trained the model? Existing laws were not designed for systems that can act with high levels of autonomy.
\nFor Europe and the Western Balkans, this is not a remote issue. Countries in the region are aligning their digital and security policies with European and international frameworks. Any future treaty on autonomous weapons would shape defense policy, AI governance, procurement standards, and public debate across the region.
\nThe UN’s position puts one principle at the center of global AI regulation: machines may assist human decisions, but they must not replace human responsibility.
\nAs AI becomes part of modern warfare, the question is no longer whether governments will use it. The question is where the world draws the line.
","featured_image_id":141,"intro_image_id":141,"featured_image_credit":"UN Secretary-General António Guterres","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":2,"view_count":16,"ordering":0,"published_at":"2026-07-07 07:05:37","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-07 07:05:37","updated_at":"2026-07-07 07:05:37","key_takeaways":null,"faq":null,"sources":null,"category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/06b165b96ad7384efe0f9bd92ba23cdb.webp","featured_image_alt":"A suited man speaks at a podium for the Global Dialogue on AI Governance, with a UN flag behind him.","featured_image_width":1200,"featured_image_height":800,"intro_image_path":"/api/uploads/2026/07/06b165b96ad7384efe0f9bd92ba23cdb.webp","intro_image_width":1200,"intro_image_height":800,"seo_title":"UN Calls for Ban on Autonomous AI Weapons","meta_description":"The UN is urging a global ban on lethal autonomous weapons, warning that AI must never decide life-and-death matters without human control.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":65,"title":"UN Opens First Global AI Governance Dialogue in Geneva","slug":"un-opens-first-global-ai-governance-dialogue-in-geneva","short_description":"","content":"GENEVA — The United Nations opened its first Global Dialogue on AI Governance on Monday, bringing governments, technology companies, researchers and civil society into a new global debate over who should shape the rules for artificial intelligence.
\nThe talks come as AI systems move deeper into public services, education, healthcare, media, security and the economy, while many governments are still trying to catch up with the speed of the technology.
\n![\"\"](\"/api/uploads/2026/07/9dd0706c56901bb37d5dbc5bea1b5bc5.webp\")
The message from the UN was clear: AI governance should not be written only by powerful countries and major technology companies.
\n“AI is advancing at runaway speed. The question is whether we will govern it together, or let it govern us,” UN Secretary-General António Guterres said. “For the first time, the AI Dialogue gives every country a seat at the table.”
\nThe Dialogue was mandated by the UN General Assembly as a recurring platform where member states and stakeholders can exchange national experiences, discuss risks and build common approaches to AI governance.
\nIn Geneva, discussions are focusing on AI safety, accountability, human oversight, international cooperation and the widening gap between countries with advanced AI capacity and those still building digital infrastructure.
\nThat divide is becoming one of the central issues in global AI policy. While some countries are investing heavily in compute power, AI labs and regulation, others are still working on basic digital public infrastructure, skills development and institutional capacity.
\nAccording to UNESCO, the process follows months of consultations involving governments, civil society, academia, the private sector and the technical community. More than 1,500 written submissions were received from organisations and individuals across regional groups.
\nThe UN says AI is already being governed through national regulations, technical standards, procurement frameworks and bilateral agreements, but not equally. Governance frameworks have largely been shaped by countries with advanced AI sectors, while many countries most exposed to AI’s consequences have had less influence over how those rules are designed.
\nThe Dialogue also follows the preliminary report of the Independent International Scientific Panel on Artificial Intelligence, a 40-member expert group co-chaired by Yoshua Bengio and Maria Ressa. The panel warned that current safeguards are struggling to keep pace with the growth of AI capabilities.
\nFor the Balkans, the Geneva talks matter because the region is still defining its place in the AI economy.
\nCountries across Southeast Europe face the same questions now being discussed at UN level: how to regulate AI, how to protect citizens, how to support innovation, and how to avoid becoming only consumers of systems developed elsewhere.
\nThe challenge is not only legal. It is also practical.
\nSmaller economies need stronger digital infrastructure, local expertise, public-sector capacity, regional cooperation and language inclusion if AI is to support development rather than deepen dependency on foreign platforms.
\nFor Balkan governments, universities, civil society and AI communities, the signal from Geneva is clear: global AI governance is now being shaped, and smaller regions cannot afford to remain outside the conversation.
\nThe UN process will not immediately create binding global AI rules. But it could influence national AI strategies, procurement standards, safety expectations and future international cooperation.
\nFor the Balkans, that makes participation important not only as a diplomatic exercise, but as a strategic question: whether the region helps shape the future of AI governance, or simply adapts to rules written elsewhere.
","featured_image_id":139,"intro_image_id":139,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":21,"ordering":0,"published_at":"2026-07-06 11:27:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-06 11:27:52","updated_at":"2026-07-06 11:30:53","key_takeaways":null,"faq":null,"sources":null,"category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/d83d365befe43c1b3cfb96a0ffbbe704.webp","featured_image_alt":"United Nations Global Dialogue on AI Governance","featured_image_width":1420,"featured_image_height":799,"intro_image_path":"/api/uploads/2026/07/d83d365befe43c1b3cfb96a0ffbbe704.webp","intro_image_width":1420,"intro_image_height":799,"seo_title":"UN Opens First Global AI Governance Dialogue in Geneva","meta_description":"The UN has opened its first Global Dialogue on AI Governance, calling for safer, more inclusive AI rules as smaller regions seek a stronger voice.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":58,"title":"EU Unveils Tech Sovereignty Package to Boost AI, Chips and Cloud","slug":"eu-unveils-tech-sovereignty-package-to-boost-ai-chips-and-cloud","short_description":"","content":"BRUSSELS — The European Commission has adopted a new Tech Sovereignty Package aimed at strengthening Europe’s digital autonomy, with measures focused on artificial intelligence, semiconductors, cloud infrastructure, open-source software and energy systems.
\nThe package, adopted on 3 June 2026, marks a major shift in the EU’s digital agenda: from regulating technology to building more of the infrastructure behind it. According to the Commission, Europe’s tech sovereignty means the ability to develop and control key technologies, data and infrastructure while reducing reliance on non-EU providers.
\nThe initiative includes two legislative proposals, Chips Act 2.0 and the Cloud and AI Development Act, alongside an EU Open Source Strategy and a Strategic Roadmap for Digitalisation and AI in Energy.
The Commission says the EU currently relies on non-EU countries for more than 80% of key digital products, services, infrastructure and intellectual property, a dependency it sees as a risk to competitiveness, security and long-term resilience.
\nArtificial intelligence is one of the central areas of the package. The EU’s wider AI policy already includes the AI Act, the AI Continent Action Plan, AI Factories and measures to improve access to computing infrastructure, data, skills and adoption across strategic sectors.
\nThe proposed Cloud and AI Development Act is intended to support more sovereign and competitive European AI and cloud solutions. The Commission says it will create better conditions for businesses, researchers and public administrations to invest in AI and cloud technologies.
\nSemiconductors are another priority. Through Chips Act 2.0, the EU wants to strengthen Europe’s chip ecosystem by supporting research, design and manufacturing capacity while improving supply chain resilience.
\nThe open-source strategy is also presented as part of Europe’s sovereignty agenda, with the Commission aiming to support the full lifecycle of open-source technologies from research and development to market adoption.
\nFor the Western Balkans, the package matters because EU digital policy increasingly shapes the region’s own regulatory and innovation agenda. Countries working toward EU alignment are likely to feel the effects of Europe’s approach to AI infrastructure, cloud services, cybersecurity, data governance and digital skills.
\nThe package also sends a broader signal to Balkan governments, universities and startups: future AI competitiveness will depend not only on regulation, but on access to computing power, trusted data, cloud infrastructure and skilled talent.
\n\n\n\n\n\n\n\n\n\nEurope’s latest move shows that the AI race is no longer only about models and applications. It is also about who controls the chips, cloud systems, data infrastructure and energy capacity behind them.
","featured_image_id":126,"intro_image_id":126,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":20,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":40,"ordering":0,"published_at":"2026-07-02 22:35:13","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-02 22:35:13","updated_at":"2026-07-02 22:35:13","key_takeaways":null,"faq":null,"sources":null,"category_name":"Policy & Regulation","category_slug":"policy-regulation","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/2604f5de8c500cc8eec9eb5ede7801c1.webp","featured_image_alt":"The European Union flag waves against a blue, clouded sky.","featured_image_width":1600,"featured_image_height":936,"intro_image_path":"/api/uploads/2026/07/2604f5de8c500cc8eec9eb5ede7801c1.webp","intro_image_width":1600,"intro_image_height":936,"seo_title":"EU Tech Sovereignty Package Targets AI, Chips and Cloud","meta_description":"The European Commission has adopted a Tech Sovereignty Package to strengthen Europe’s AI, cloud, semiconductor and open-source capacity.","robots":"index","canonical_url":"","og_title":"","og_description":""}]}],"topic_pool":[{"id":176,"title":"AI Balkans Is Now Available on Android","slug":"ai-balkans-is-now-available-on-android","short_description":null,"featured_image_id":422,"intro_image_id":422,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":3,"view_count":16,"ordering":0,"published_at":"2026-10-25 09:39:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-25 09:39:27","updated_at":"2026-10-25 10:09:55","category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/52f040254899795e1cb762d1f0ac40d6.webp","featured_image_alt":"Aibalkans android","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/52f040254899795e1cb762d1f0ac40d6.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"AI Balkans Android App Now Available on Google Play","meta_description":"Download the official AI Balkans Android app for regional AI news, Lumi assistant, EU AI Act Tracker and personalised reading.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":172,"title":"Western Balkans AI Roadmap Presented, Awaiting Leaders' Endorsement","slug":"western-balkans-ai-roadmap-presented-awaiting-leaders-endorsement","short_description":null,"featured_image_id":393,"intro_image_id":393,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":5,"view_count":31,"ordering":0,"published_at":"2026-10-18 16:50:41","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-18 16:50:41","updated_at":"2026-10-18 16:50:41","category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/638f26e3b54535227acc12d920ace4a3.webp","featured_image_alt":"WB RCC","featured_image_width":1600,"featured_image_height":1066,"intro_image_path":"/api/uploads/2026/10/638f26e3b54535227acc12d920ace4a3.webp","intro_image_width":1600,"intro_image_height":1066,"seo_title":"Western Balkans AI Roadmap Awaits Endorsement","meta_description":"Six Western Balkan economies joined AI roadmap discussions in Montenegro. The RCC says formal endorsement is expected at the Berlin Process summit.","robots":"index","canonical_url":"western-balkans-ai-roadmap-awaits-endorsement","og_title":"","og_description":""},{"id":159,"title":"Skopje and Pristina Selected for UNDP AI Programme","slug":"skopje-and-pristina-selected-for-undp-ai-programme","short_description":null,"featured_image_id":363,"intro_image_id":363,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":24,"ordering":0,"published_at":"2026-10-09 08:22:38","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-09 08:22:38","updated_at":"2026-10-09 08:22:38","category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/f3d03d10c71f37b55da58b7d71f4b6ce.webp","featured_image_alt":"UNDP","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/f3d03d10c71f37b55da58b7d71f4b6ce.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Skopje and Pristina Join UNDP AI Programme","meta_description":"Skopje and Pristina join UNDP's CEF x AI programme to explore AI applications in water management, urban greenery and transport planning.","robots":"index","canonical_url":"skopje-pristina-undp-ai-programme","og_title":"","og_description":""},{"id":155,"title":"World Bank Warns Balkans Risk Falling Behind on AI","slug":"world-bank-warns-balkans-risk-falling-behind-on-ai","short_description":null,"featured_image_id":359,"intro_image_id":359,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":20,"ordering":0,"published_at":"2026-10-07 08:46:38","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-07 08:46:38","updated_at":"2026-10-07 08:46:38","category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/6e39e23882afcc9eef53183872a0e4b2.webp","featured_image_alt":"World Bank","featured_image_width":673,"featured_image_height":456,"intro_image_path":"/api/uploads/2026/10/6e39e23882afcc9eef53183872a0e4b2.webp","intro_image_width":673,"intro_image_height":456,"seo_title":"World Bank Warns Balkans Risk Falling Behind on AI","meta_description":"World Bank warns limited AI adoption, computing capacity and skills could leave developing economies, including the Western Balkans, further behind.","robots":"index","canonical_url":"world-bank-warns-balkans-risk-falling-behind-on-ai","og_title":"","og_description":""},{"id":141,"title":"Slovenia Deepens AI Ties With US Through New Technology Partnership","slug":"slovenia-deepens-ai-ties-with-us-through-new-technology-partnership","short_description":"","featured_image_id":311,"intro_image_id":311,"featured_image_credit":"Minister Tone Kajzer signing the Joint Statement on AI Opportunity Partnership","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":3,"view_count":54,"ordering":0,"published_at":"2026-09-25 07:33:23","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-25 07:33:23","updated_at":"2026-09-25 07:33:23","category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/50d7277dda2641c1eb80274dbc176a0a.webp","featured_image_alt":"A suited man wearing a Slovenian flag pin signs a document at a formal event, with a flower arrangement in the foreground.","featured_image_width":1600,"featured_image_height":1066,"intro_image_path":"/api/uploads/2026/09/50d7277dda2641c1eb80274dbc176a0a.webp","intro_image_width":1600,"intro_image_height":1066,"seo_title":"Slovenia Signs AI Partnership With US","meta_description":"Slovenia has signed an AI partnership with the US covering semiconductors, computing infrastructure and secure technology supply chains.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":139,"title":"Microsoft Data Shows AI Use Rising Across the Balkans","slug":"microsoft-data-shows-ai-use-rising-across-the-balkans","short_description":"","featured_image_id":305,"intro_image_id":305,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":38,"ordering":0,"published_at":"2026-09-23 12:49:36","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-23 12:49:36","updated_at":"2026-09-23 12:49:36","category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/bff3e3039f68483b42bda67c52531fee.webp","featured_image_alt":"World map of AI user share by economy in June 2026, with rates ranging from 49.4% in Norway to 73.3% in the UAE.","featured_image_width":1600,"featured_image_height":1027,"intro_image_path":"/api/uploads/2026/09/bff3e3039f68483b42bda67c52531fee.webp","intro_image_width":1600,"intro_image_height":1027,"seo_title":"AI Use Is Rising Across the Balkans, Microsoft Data Shows","meta_description":"Microsoft data shows generative AI use rising across the Balkans, led by Bulgaria, Slovenia, Croatia and Serbia.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":131,"title":"Croatia Public Sector AI Use Outpaces Internal Rules","slug":"croatia-public-sector-ai-use-outpaces-internal-rules","short_description":"","featured_image_id":289,"intro_image_id":289,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":37,"ordering":0,"published_at":"2026-09-12 10:41:28","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-12 10:41:28","updated_at":"2026-09-12 10:41:28","category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/c50d80e1b61af15de515e875271d50b0.webp","featured_image_alt":"Officials sign documents at a formal meeting before Croatian, EU, Czech, Hungarian, Polish and Slovak flags.","featured_image_width":1396,"featured_image_height":869,"intro_image_path":"/api/uploads/2026/09/c50d80e1b61af15de515e875271d50b0.webp","intro_image_width":1396,"intro_image_height":869,"seo_title":"Croatia Public Sector AI Use Outpaces Internal Rules","meta_description":"50 Croatian public institutions use or test AI, but 26 have no specific internal rules, according to new research by Gong.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":127,"title":"University of Ljubljana Sets Common Rules for AI Across Campus","slug":"university-of-ljubljana-sets-common-rules-for-ai-across-campus","short_description":"","featured_image_id":278,"intro_image_id":278,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":36,"ordering":0,"published_at":"2026-09-06 12:40:14","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-06 12:40:14","updated_at":"2026-09-06 12:40:14","category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/0edf8cefebe858fde5364b67073c8c80.webp","featured_image_alt":"The University of Ljubljana’s ornate historic main building stands in a sunny courtyard framed by trees.","featured_image_width":1200,"featured_image_height":630,"intro_image_path":"/api/uploads/2026/09/0edf8cefebe858fde5364b67073c8c80.webp","intro_image_width":1200,"intro_image_height":630,"seo_title":"University of Ljubljana Sets Campus-Wide AI Rules","meta_description":"The University of Ljubljana has introduced common AI rules covering students, lecturers, researchers and professional staff.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":125,"title":"Bulgaria Prepares National AI Law as EU Enforcement Begins","slug":"bulgaria-prepares-national-ai-law-as-eu-enforcement-begins","short_description":"","featured_image_id":125,"intro_image_id":125,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":31,"ordering":0,"published_at":"2026-09-06 12:31:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-06 12:31:57","updated_at":"2026-09-06 12:32:02","category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/8b45ee0395ff9bbea1d0c30c3f13e7d6.webp","featured_image_alt":"EU Artificial Intelligence Act, with servers in the background, and European stars.","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/07/8b45ee0395ff9bbea1d0c30c3f13e7d6.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Bulgaria Prepares National AI Law as EU Rules Begin","meta_description":"Bulgaria is finalising national AI legislation as enforcement of key EU AI Act rules begins across member states.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":124,"title":"Romania Begins Building Its First National AI Factory","slug":"romania-begins-building-its-first-national-ai-factory","short_description":"","featured_image_id":275,"intro_image_id":275,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":7,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":25,"ordering":0,"published_at":"2026-09-06 12:25:59","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-06 12:25:59","updated_at":"2026-09-06 12:25:59","category_name":"Balkan AI","category_slug":"balkan-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/3781ed520f70b9990015cdd207d2e83e.webp","featured_image_alt":"A EuroHPC graphic titled “New EuroHPC AI Factories” shows RO AI connected to CZAI, 1HealthAI, Gaia AI, LitAI and NLAIF.","featured_image_width":1200,"featured_image_height":600,"intro_image_path":"/api/uploads/2026/09/3781ed520f70b9990015cdd207d2e83e.webp","intro_image_width":1200,"intro_image_height":600,"seo_title":"Romania Begins Building Its First National AI Factory","meta_description":"Romania has started implementing RO AI Factory, a EuroHPC-backed national AI infrastructure expected to become operational by 2027.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":173,"title":"Solo Entrepreneur Toolkit 2026: 15 Tools You Can Start Using for Free","slug":"solo-entrepreneur-toolkit-2026-15-tools-you-can-start-using-for-free","short_description":null,"featured_image_id":394,"intro_image_id":394,"featured_image_credit":"Jakub Żerdzicki / Unsplash","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":14,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":8,"view_count":21,"ordering":0,"published_at":"2026-10-20 19:19:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-20 19:19:35","updated_at":"2026-10-20 19:20:59","category_name":"Tools & Guides","category_slug":"ai-tools","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/d9f265e308481038b67a862b6f6c0c9d.webp","featured_image_alt":"Laptop Device","featured_image_width":1600,"featured_image_height":1067,"intro_image_path":"/api/uploads/2026/10/d9f265e308481038b67a862b6f6c0c9d.webp","intro_image_width":1600,"intro_image_height":1067,"seo_title":"Solo Entrepreneur Toolkit 2026: 15 Free Tools","meta_description":"Discover 15 free and freemium tools for solo entrepreneurs in 2026. Build websites, manage clients, create content and automate everyday tasks.","robots":"index","canonical_url":"olo-entrepreneur-toolkit-2026-free-tools","og_title":"","og_description":""},{"id":136,"title":"Anthropic Launches Claude Docs and Slides in Beta","slug":"anthropic-launches-claude-docs-and-slides-in-beta","short_description":"","featured_image_id":297,"intro_image_id":297,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":14,"author_id":10,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":29,"ordering":0,"published_at":"2026-09-18 09:33:32","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-18 09:33:32","updated_at":"2026-09-18 09:33:32","category_name":"Tools & Guides","category_slug":"ai-tools","author_name":"Guest Contributor","author_slug":"guest-contributor","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/c2d81cf086c06dad71c59570b4bf90c4.webp","featured_image_alt":"A Claude Docs interface shows a collaborative document, bar chart, highlighted edits, and comment bubbles on a coral background.","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/09/c2d81cf086c06dad71c59570b4bf90c4.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Anthropic Launches Claude Docs and Slides in Beta","meta_description":"Anthropic launches Claude Docs and Slides, bringing collaborative documents and presentations directly into Claude as it expands beyond chat.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":113,"title":"The AI Stack for a One-Person Business in 2026","slug":"the-ai-stack-for-a-one-person-business-in-2026","short_description":"","featured_image_id":76,"intro_image_id":76,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":14,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":11,"view_count":29,"ordering":0,"published_at":"2026-08-23 09:59:43","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-23 09:59:43","updated_at":"2026-08-23 09:59:43","category_name":"Tools & Guides","category_slug":"ai-tools","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/396dde8ff392398dc24d235b0043b9aa.webp","featured_image_alt":"A person types on a laptop as neon binary code streams from a circuit-board display.","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/06/396dde8ff392398dc24d235b0043b9aa.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"The AI Stack for a One-Person Business in 2026","meta_description":"A realistic AI stack for a one-person business in 2026, with current pricing, free options, hidden costs and tools actually worth paying for.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":96,"title":"ACTLAS Maps the EU AI Act as a Knowledge Graph","slug":"actlas-maps-the-eu-ai-act-as-a-knowledge-graph","short_description":"","featured_image_id":227,"intro_image_id":227,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":14,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":2,"view_count":26,"ordering":0,"published_at":"2026-08-02 13:21:41","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-02 13:21:41","updated_at":"2026-08-02 13:21:41","category_name":"Tools & Guides","category_slug":"ai-tools","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/0970d42fdf88fc18906d74e6d9fdd440.webp","featured_image_alt":"ACTLAS Graph Explorer displays a colorful EU AI Act knowledge graph and details for the “intended purpose” definition.","featured_image_width":1280,"featured_image_height":736,"intro_image_path":"/api/uploads/2026/08/0970d42fdf88fc18906d74e6d9fdd440.webp","intro_image_width":1280,"intro_image_height":736,"seo_title":"ACTLAS Maps the EU AI Act as a Knowledge Graph","meta_description":"ACTLAS uses knowledge graphs to connect the EU AI Act, GDPR and ISO 42001, helping teams trace obligations, evidence and regulatory gaps.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":122,"title":"GPT-6 Astra vs Claude Fable 5.1: The Frontier AI Race Moves Beyond Chatbots","slug":"gpt-6-astra-vs-claude-fable-5-1-the-frontier-ai-race-moves-beyond-chatbots","short_description":"","featured_image_id":267,"intro_image_id":267,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":17,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":6,"view_count":38,"ordering":0,"published_at":"2026-09-04 07:41:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-04 07:41:33","updated_at":"2026-09-04 07:41:54","category_name":"Comparisons","category_slug":"comparisons","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/9a8a0b3853f386ce33287c166155f106.webp","featured_image_alt":"ChatGPT logo, crossed swords and the letters “AI” on a tan background.","featured_image_width":1600,"featured_image_height":800,"intro_image_path":"/api/uploads/2026/09/9a8a0b3853f386ce33287c166155f106.webp","intro_image_width":1600,"intro_image_height":800,"seo_title":"GPT-6 Astra vs Claude Fable 5.1: What Separates Them","meta_description":"GPT-6 Astra and Claude Fable 5.1 target coding, research and agentic work. We compare capabilities, pricing, access and safety.","robots":"index","canonical_url":"gpt-6-astra-vs-claude-fable-5-1","og_title":"","og_description":""},{"id":51,"title":"Codex vs Claude Code: Which Coding Agent Wins?","slug":"codex-vs-claude-code-which-coding-agent-wins","short_description":"","featured_image_id":107,"intro_image_id":107,"featured_image_credit":"The Futurum Group","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":17,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":8,"view_count":29,"ordering":0,"published_at":"2026-06-27 15:48:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-27 15:48:24","updated_at":"2026-06-27 16:59:55","category_name":"Comparisons","category_slug":"comparisons","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/ef1b678b5284b20bbb1cfe10315b7528.webp","featured_image_alt":"Blurred hands type code behind blue binary graphics and glowing circuit lines.","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/06/ef1b678b5284b20bbb1cfe10315b7528.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Codex vs Claude Code: Which Coding Agent Wins?","meta_description":"Codex vs Claude Code compared for real developer workflows, codebase understanding, cost, git, testing, and small team productivity.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":20,"title":"AINOW AI Education Platform for Teachers: Online and Offline Access","slug":"ainow-ai-education-platform-for-teachers-online-and-offline-access","short_description":"Free, offline-first AI literacy platform for educators. 300+ prompts, 57 tools, 18-chapter guide. Works on any device. No tracking. Open source. Suad Seferi","featured_image_id":50,"intro_image_id":50,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":15,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":1,"show_table_of_contents":1,"read_time_minutes":5,"view_count":27,"ordering":0,"published_at":"2026-06-22 19:22:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-22 19:26:30","updated_at":"2026-06-25 18:33:05","category_name":"Guides","category_slug":"guides","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/8a11840dbae9d7cfea2207940bf8019f.webp","featured_image_alt":"AINOW Society, Platform for AI Literacy preview","featured_image_width":1600,"featured_image_height":901,"intro_image_path":"/api/uploads/2026/06/8a11840dbae9d7cfea2207940bf8019f.webp","intro_image_width":1600,"intro_image_height":901,"seo_title":"AINOW AI Education Platform for Teachers: Online and Offline Access","meta_description":"Free, offline-first AI literacy platform for educators. 300+ prompts, 57 tools, 18-chapter guide. Works on any device. No tracking. Suad Seferi","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":39,"title":"The Western Balkans Are No Longer a Digital Afterthought","slug":"the-western-balkans-are-no-longer-a-digital-afterthought","short_description":"","featured_image_id":78,"intro_image_id":78,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":5,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":6,"view_count":17,"ordering":0,"published_at":"2026-06-25 18:26:41","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-25 18:26:41","updated_at":"2026-06-25 18:26:41","category_name":"Reports","category_slug":"reports","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/e4af3774eea0ae53ee5d9886bd2e7f9e.webp","featured_image_alt":"A hand touches a glowing AI interface surrounded by blue circuit lines and digital overlays.","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/06/e4af3774eea0ae53ee5d9886bd2e7f9e.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"The Western Balkans Are No Longer a Digital Afterthought","meta_description":"For years, the Western Balkans have been treated as a secondary market by global technology companies.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":13,"title":"Macedonian AI Data 2026: A Strategy for Digital Sovereignty","slug":"macedonian-ai-data-2026-a-strategy-for-digital-sovereignty","short_description":"Macedonian language data for artificial intelligence reached 3.53 billion words in January 2026. This figure sounds like a success. We must look past the total volume. High-quality filtering reduces this amount to 1.47 billion words of trainable text. We are building on a digital bottleneck. Your AI performance depends on the quality and variety of these words.","featured_image_id":59,"intro_image_id":59,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":5,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":71,"ordering":0,"published_at":"2026-06-20 08:39:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-20 08:39:15","updated_at":"2026-06-25 18:28:09","category_name":"Reports","category_slug":"reports","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/0f19ae59b6aa0f858e6c60fd27d23646.webp","featured_image_alt":"AI Data neural network style","featured_image_width":1200,"featured_image_height":800,"intro_image_path":"/api/uploads/2026/06/0f19ae59b6aa0f858e6c60fd27d23646.webp","intro_image_width":1200,"intro_image_height":800,"seo_title":"Macedonian AI Data 2026: Strategy for Digital Sovereignty","meta_description":"North Macedonia faces a 100x data gap compared to English AI. Learn why a National Corpus and strategic curation are vital for accurate local AI models.","robots":"index","canonical_url":"https://www.aibalkans.com/macedonian-ai-data-2026-a-strategy-for-digital-sovereignty","og_title":"","og_description":""},{"id":144,"title":"AI Summaries Can Change What People Remember","slug":"ai-summaries-can-change-what-people-remember","short_description":"","featured_image_id":316,"intro_image_id":316,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":2,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":40,"ordering":0,"published_at":"2026-09-26 09:53:21","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-26 09:53:21","updated_at":"2026-09-26 09:53:21","category_name":"Research","category_slug":"research","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/755257602e24c00cbbd22c5c72c6f250.webp","featured_image_alt":"A woman looks at a laptop beside the AI Balkans logo and floating graphics of a video and an AI text summary.","featured_image_width":1448,"featured_image_height":1086,"intro_image_path":"/api/uploads/2026/09/755257602e24c00cbbd22c5c72c6f250.webp","intro_image_width":1448,"intro_image_height":1086,"seo_title":"AI Summaries Can Change What People Remember | AI Balkans","meta_description":"AI-generated summaries containing incorrect information can alter what people later remember about an event, according to new research from Georgetown…","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":46,"title":"DeepL Review: How Well Does It Fit Balkan Languages?","slug":"deepl-review-how-well-does-it-fit-balkan-languages","short_description":"","featured_image_id":101,"intro_image_id":101,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":16,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":6,"view_count":9,"ordering":0,"published_at":"2026-06-27 15:19:01","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-27 15:19:01","updated_at":"2026-06-27 15:19:01","category_name":"Reviews","category_slug":"reviews","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/5699c819aedfb071cc5118139c337392.webp","featured_image_alt":"DeepL Review","featured_image_width":1600,"featured_image_height":792,"intro_image_path":"/api/uploads/2026/06/5699c819aedfb071cc5118139c337392.webp","intro_image_width":1600,"intro_image_height":792,"seo_title":"DeepL Review: How Well Does It Fit Balkan Languages?","meta_description":"A practical review of DeepL for Balkan language translation, including Macedonian, Albanian, Serbian, Croatian, Bosnian, Slovenian, and Romanian.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":45,"title":"Perplexity Review: Is AI Search Ready for Balkan Researchers and Journalists?","slug":"perplexity-review","short_description":"","featured_image_id":100,"intro_image_id":100,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":16,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":6,"view_count":16,"ordering":0,"published_at":"2026-06-27 15:15:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-27 15:15:38","updated_at":"2026-06-28 10:12:00","category_name":"Reviews","category_slug":"reviews","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/73c32d5d8ed84a64ce056658cdb21793.webp","featured_image_alt":"The white Perplexity logo and name appear over a black background with flowing dotted lines.","featured_image_width":1400,"featured_image_height":713,"intro_image_path":"/api/uploads/2026/06/73c32d5d8ed84a64ce056658cdb21793.webp","intro_image_width":1400,"intro_image_height":713,"seo_title":"Perplexity Review: Is AI Search Ready for Balkan Researchers and Journalists?","meta_description":"A practical AI Balkans review of Perplexity for journalists, researchers, analysts, students, NGOs, and policy teams working with AI information.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":44,"title":"SMEs Are Using AI, But Real Integration Is Still Uneven","slug":"smes-are-using-ai-but-real-integration-is-still-uneven","short_description":"","featured_image_id":99,"intro_image_id":99,"featured_image_credit":"","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":4,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":6,"view_count":7,"ordering":0,"published_at":"2026-06-27 15:12:40","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-27 15:12:40","updated_at":"2026-06-28 11:34:18","category_name":"Surveys","category_slug":"surveys","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/b980e695e4054d7bd602dce04259f1e8.webp","featured_image_alt":"A woman works on a laptop before a large blue screen displaying an AI brain graphic and data.","featured_image_width":1400,"featured_image_height":932,"intro_image_path":"/api/uploads/2026/06/b980e695e4054d7bd602dce04259f1e8.webp","intro_image_width":1400,"intro_image_height":932,"seo_title":"SMEs Are Using AI, But Real Integration Is Still Uneven","meta_description":"New European survey data shows that SMEs are adopting AI tools, but many still struggle with skills, integration, cybersecurity & practical implementation.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":43,"title":"ChatGPT, Claude, or Gemini?","slug":"chatgpt-claude-or-gemini","short_description":"","featured_image_id":98,"intro_image_id":98,"featured_image_credit":"Tom's Guide","featured_image_overlay":null,"featured_image_overlay_position":null,"category_id":4,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":7,"view_count":16,"ordering":0,"published_at":"2026-06-27 15:05:48","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-06-27 15:05:48","updated_at":"2026-06-28 11:34:21","category_name":"Surveys","category_slug":"surveys","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/06/98cd6b36e36bffe2633314d39e2be8a6.webp","featured_image_alt":"The ChatGPT, Gemini and Claude logos appear on teal, black and coral panels.","featured_image_width":1200,"featured_image_height":675,"intro_image_path":"/api/uploads/2026/06/98cd6b36e36bffe2633314d39e2be8a6.webp","intro_image_width":1200,"intro_image_height":675,"seo_title":"ChatGPT, Claude, or Gemini? | AI Balkans","meta_description":"A practical comparison of ChatGPT, Claude, and Gemini for small businesses, NGOs, agencies, educators, and teams in the Balkans.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":177,"title":"Claude Takes Software Security Beyond Bug Detection","slug":"claude-takes-software-security-beyond-bug-detection","short_description":null,"featured_image_id":423,"intro_image_id":423,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":18,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":6,"view_count":15,"ordering":0,"published_at":"2026-10-26 08:19:37","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-26 08:19:37","updated_at":"2026-10-26 08:19:37","category_name":"Cybersecurity","category_slug":"cybersecurity","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/ab55c0a3fc619bc0af554b7c60012f30.webp","featured_image_alt":"Claude code scan","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/ab55c0a3fc619bc0af554b7c60012f30.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Claude Uses Mathematical Proofs to Verify Software Security","meta_description":"Anthropic says Claude verified security properties across 123 open-source libraries and uncovered bugs missed by traditional testing.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":174,"title":"PHP2Shell Sparks Global Security Alert for Websites","slug":"php2shell-sparks-global-security-alert-for-websites","short_description":null,"featured_image_id":411,"intro_image_id":411,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":18,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":9,"view_count":24,"ordering":0,"published_at":"2026-10-21 18:12:55","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-21 18:12:55","updated_at":"2026-10-21 18:12:55","category_name":"Cybersecurity","category_slug":"cybersecurity","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/18258cdb232fc62815dc9dff161cf97e.webp","featured_image_alt":"PHP Vulnerability","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/10/18258cdb232fc62815dc9dff161cf97e.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"PHP2Shell Sparks Global Security Alert for Websites","meta_description":"Two critical PHP vulnerabilities threaten WordPress and other websites. See Balkan usage data and the PHP updates administrators must install.","robots":"index","canonical_url":"php2shell-php-wordpress-balkans-security","og_title":"","og_description":""},{"id":103,"title":"OpenAI Slows Astra Release as Cyber Capabilities Raise New Safety Concerns","slug":"openai-slows-astra-release-as-cyber-capabilities-raise-new-safety-concerns","short_description":"","featured_image_id":145,"intro_image_id":145,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":18,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":5,"view_count":18,"ordering":0,"published_at":"2026-08-12 13:45:20","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-12 13:45:20","updated_at":"2026-08-12 13:45:20","category_name":"Cybersecurity","category_slug":"cybersecurity","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/7b7f0023b918f429a5ca0f299a7cf40b.webp","featured_image_alt":"OpenAI logo with robo hand","featured_image_width":1200,"featured_image_height":800,"intro_image_path":"/api/uploads/2026/07/7b7f0023b918f429a5ca0f299a7cf40b.webp","intro_image_width":1200,"intro_image_height":800,"seo_title":"OpenAI Slows Astra Release Over AI Cybersecurity Risks","meta_description":"OpenAI is slowing Astra's release after cybersecurity tests raised concerns about increasingly capable autonomous AI agents.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":101,"title":"AI Agents Are Raising New Cybersecurity Risks","slug":"ai-agents-are-raising-new-cybersecurity-risks","short_description":"","featured_image_id":236,"intro_image_id":236,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":18,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":5,"view_count":21,"ordering":0,"published_at":"2026-08-08 09:44:22","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-08 09:44:22","updated_at":"2026-08-08 09:44:22","category_name":"Cybersecurity","category_slug":"cybersecurity","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/68cd562f4f45f344c3b9fa8c5514b844.webp","featured_image_alt":"A hand supports a glowing blue holographic AI interface surrounded by circuitry and digital particles.","featured_image_width":992,"featured_image_height":662,"intro_image_path":"/api/uploads/2026/08/68cd562f4f45f344c3b9fa8c5514b844.webp","intro_image_width":992,"intro_image_height":662,"seo_title":"AI Agents Are Raising New Cybersecurity Risks","meta_description":"AI agents from Meta, OpenAI and Anthropic have breached testing boundaries, raising new questions about cybersecurity and autonomous AI systems.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":93,"title":"When AI Agents Find Their Way Out of the Sandbox","slug":"when-ai-agents-find-their-way-out-of-the-sandbox","short_description":"","featured_image_id":223,"intro_image_id":223,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":18,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":6,"view_count":32,"ordering":0,"published_at":"2026-07-30 07:57:14","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-30 07:57:14","updated_at":"2026-07-30 07:57:14","category_name":"Cybersecurity","category_slug":"cybersecurity","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/2d84aefc4f34968c253c4582757f4098.webp","featured_image_alt":"Green streams of binary digits spiral into a dark vortex on a black background.","featured_image_width":1600,"featured_image_height":1067,"intro_image_path":"/api/uploads/2026/07/2d84aefc4f34968c253c4582757f4098.webp","intro_image_width":1600,"intro_image_height":1067,"seo_title":"When AI Agents Find Their Way Out of the Sandbox","meta_description":"An OpenAI cybersecurity test reached Hugging Face production systems, revealing how weak containment can turn AI capability into real-world risk.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":79,"title":"Grok Build Sparks Privacy Debate After Developer Concerns","slug":"grok-build-sparks-privacy-debate-after-developer-concerns","short_description":"","featured_image_id":197,"intro_image_id":197,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":18,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":20,"ordering":0,"published_at":"2026-07-13 19:58:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-13 19:58:02","updated_at":"2026-07-13 20:04:35","category_name":"Cybersecurity","category_slug":"cybersecurity","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/f709e883334f6f442abea602599dd78f.webp","featured_image_alt":"Grok Build Beta logo displayed over an orange and black gradient background.","featured_image_width":1200,"featured_image_height":630,"intro_image_path":"/api/uploads/2026/07/f709e883334f6f442abea602599dd78f.webp","intro_image_width":1200,"intro_image_height":630,"seo_title":"Grok Build Sparks Privacy Debate After Developer Concerns","meta_description":"xAI responds after a researcher reported that Grok Build uploaded full code repositories, raising questions about privacy and AI coding agents.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":145,"title":"Belgrade Puts Clinical AI in Focus at MIND2026","slug":"belgrade-puts-clinical-ai-in-focus-at-mind2026","short_description":"","featured_image_id":321,"intro_image_id":321,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":19,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":25,"ordering":0,"published_at":"2026-09-27 15:56:59","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-27 15:56:59","updated_at":"2026-09-27 15:56:59","category_name":"Healthcare AI","category_slug":"healthcare-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/fb75e22a9a78ec0ebe7c5dd05eff53c3.webp","featured_image_alt":"Four men sit on a MIND stage before an audience, backed by Telekom Srbija branding and Serbian medtech panel text.","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/09/fb75e22a9a78ec0ebe7c5dd05eff53c3.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"Belgrade Hosts MIND2026 Congress on AI in Medicine","meta_description":"MIND2026 brings doctors and researchers to Belgrade to examine how AI is being used in diagnostics, treatment and clinical healthcare.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":140,"title":"Claude Helps Researchers Discover New Enzyme System","slug":"claude-helps-researchers-discover-new-enzyme-system","short_description":"","featured_image_id":310,"intro_image_id":310,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":19,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":37,"ordering":0,"published_at":"2026-09-24 07:38:13","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-24 07:38:13","updated_at":"2026-09-24 07:38:13","category_name":"Healthcare AI","category_slug":"healthcare-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/b79ea9370cc8a267a3a578f2e25eb5a8.webp","featured_image_alt":"Scientific diagrams show Evo 2, gLM2, and Claude Mythos 5 detecting DNA tandem repeats using sequence and signal plots.","featured_image_width":831,"featured_image_height":723,"intro_image_path":"/api/uploads/2026/09/b79ea9370cc8a267a3a578f2e25eb5a8.webp","intro_image_width":831,"intro_image_height":723,"seo_title":"Claude Helps Researchers Find a New Enzyme System","meta_description":"Anthropic says Claude helped researchers identify a previously uncharacterized enzyme system, offering a new example of AI-assisted scientific discovery.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":115,"title":"Duke Launches AI Training for Nurses Focused on Safety and Clinical Judgment","slug":"duke-launches-ai-training-for-nurses-focused-on-safety-and-clinical-judgment","short_description":"","featured_image_id":255,"intro_image_id":255,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":19,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":24,"ordering":0,"published_at":"2026-08-25 10:10:52","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-25 10:10:52","updated_at":"2026-08-25 10:10:52","category_name":"Healthcare AI","category_slug":"healthcare-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/19113b0df328a320c1e2d5e2041d1439.webp","featured_image_alt":"Duke University logo in dark blue on a white background.","featured_image_width":1200,"featured_image_height":672,"intro_image_path":"/api/uploads/2026/08/19113b0df328a320c1e2d5e2041d1439.webp","intro_image_width":1200,"intro_image_height":672,"seo_title":"Duke Launches AI Training for Nurses Focused on Safety","meta_description":"Duke has launched a three-course AI programme teaching nurses to assess AI tools safely while keeping clinical judgment at the centre.","robots":"index","canonical_url":"duke-ai-training-nurses-clinical-practice","og_title":"","og_description":""},{"id":76,"title":"Can AI Understand a Patient Who Cannot Explain What Is Wrong?","slug":"can-ai-understand-a-patient-who-cannot-explain-what-is-wrong","short_description":"","featured_image_id":186,"intro_image_id":186,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":19,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":9,"view_count":22,"ordering":0,"published_at":"2026-07-11 15:27:51","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-11 15:27:51","updated_at":"2026-07-11 15:27:51","category_name":"Healthcare AI","category_slug":"healthcare-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/b9ca8cd8bbe1696e5b484df6e0faef4a.webp","featured_image_alt":"A doctor and a small robot monitor a patient beside glowing heart, brain, and pulse graphics.","featured_image_width":1600,"featured_image_height":850,"intro_image_path":"/api/uploads/2026/07/b9ca8cd8bbe1696e5b484df6e0faef4a.webp","intro_image_width":1600,"intro_image_height":850,"seo_title":"Can Medical AI Understand Every Patient?","meta_description":"New research finds that communication style can change medical AI triage advice, raising concerns about language, health literacy and unequal care.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":55,"title":"Healthcare AI Is Coming. Are Balkan Hospitals Ready?","slug":"healthcare-ai-is-coming-are-balkan-hospitals-ready","short_description":"","featured_image_id":118,"intro_image_id":118,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":19,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":21,"ordering":0,"published_at":"2026-07-01 10:33:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-01 10:33:39","updated_at":"2026-07-01 10:36:46","category_name":"Healthcare AI","category_slug":"healthcare-ai","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/ea912d3604e8ee5fef4c86fe1db24762.webp","featured_image_alt":"AI and Human hand touching the healthcare","featured_image_width":1440,"featured_image_height":658,"intro_image_path":"/api/uploads/2026/07/ea912d3604e8ee5fef4c86fe1db24762.webp","intro_image_width":1440,"intro_image_height":658,"seo_title":"Healthcare AI Is Coming: Are Balkan Hospitals Ready?","meta_description":"AI is reshaping healthcare worldwide. Balkan hospitals must prepare with better data, governance, infrastructure, and medical AI literacy.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":165,"title":"IMF Warns AI Boom Could Threaten Financial Stability","slug":"imf-warns-ai-boom-could-threaten-financial-stability","short_description":null,"featured_image_id":375,"intro_image_id":375,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":21,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":4,"view_count":25,"ordering":0,"published_at":"2026-10-13 16:44:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-13 16:44:09","updated_at":"2026-10-13 16:46:26","category_name":"Business & Economy","category_slug":"business-economy","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/9ab96bf686067ba017d05b13756e9af1.webp","featured_image_alt":"International Monetary Fund","featured_image_width":1600,"featured_image_height":1068,"intro_image_path":"/api/uploads/2026/10/9ab96bf686067ba017d05b13756e9af1.webp","intro_image_width":1600,"intro_image_height":1068,"seo_title":"IMF Warns AI Boom Could Threaten Financial Stability","meta_description":"The IMF warns that an AI investment correction could threaten global financial stability as technology valuations and borrowing risks increase.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":117,"title":"NVIDIA Reportedly Nears $13 Billion Deal for Hugging Face","slug":"nvidia-reportedly-nears-13-billion-deal-for-hugging-face","short_description":"","featured_image_id":259,"intro_image_id":259,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":21,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":25,"ordering":0,"published_at":"2026-08-28 10:02:31","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-28 10:02:31","updated_at":"2026-08-28 10:02:31","category_name":"Business & Economy","category_slug":"business-economy","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/8e24c0353ccb438957647805f053a229.webp","featured_image_alt":"Nvidia and Huggingface Logo's","featured_image_width":1600,"featured_image_height":900,"intro_image_path":"/api/uploads/2026/08/8e24c0353ccb438957647805f053a229.webp","intro_image_width":1600,"intro_image_height":900,"seo_title":"NVIDIA Reportedly Nears $13B Acquisition of Hugging Face","meta_description":"NVIDIA is reportedly nearing a $13 billion deal for Hugging Face, potentially giving the chipmaker a major role in the open AI ecosystem.","robots":"index","canonical_url":"nvidia-hugging-face-acquisition-13-billion","og_title":"","og_description":""},{"id":67,"title":"Microsoft Cuts 4,800 Jobs as AI Investment Reshapes the Company","slug":"microsoft-cuts-4-800-jobs-as-ai-investment-reshapes-the-company","short_description":"","featured_image_id":142,"intro_image_id":142,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":21,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":19,"ordering":0,"published_at":"2026-07-07 07:13:36","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-07 07:13:29","updated_at":"2026-07-07 07:13:36","category_name":"Business & Economy","category_slug":"business-economy","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/b9eeba58d48347a2bb9e8b483b5e040d.webp","featured_image_alt":"Silhouetted visitors stand before a bright green wall displaying the Xbox logo.","featured_image_width":1200,"featured_image_height":572,"intro_image_path":"/api/uploads/2026/07/b9eeba58d48347a2bb9e8b483b5e040d.webp","intro_image_width":1200,"intro_image_height":572,"seo_title":"Microsoft Cuts 4,800 Jobs as AI Investment Grows","meta_description":"Microsoft is cutting 4,800 jobs as it restructures around AI, with Xbox and commercial sales among the hardest-hit divisions.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":116,"title":"UNICEF and Arm Launch AI Initiative Focused on Children and Emerging Economies","slug":"unicef-and-arm-launch-ai-initiative-focused-on-children-and-emerging-economies","short_description":"","featured_image_id":258,"intro_image_id":258,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":23,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":35,"ordering":0,"published_at":"2026-08-25 10:19:50","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-25 10:19:50","updated_at":"2026-08-25 10:19:50","category_name":"AI & Education","category_slug":"ai-education","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/3c5d54e58c82fa72de509281aa99b2a4.webp","featured_image_alt":"UNICEF logo in white on a bright blue background.","featured_image_width":1200,"featured_image_height":627,"intro_image_path":"/api/uploads/2026/08/3c5d54e58c82fa72de509281aa99b2a4.webp","intro_image_width":1200,"intro_image_height":627,"seo_title":"UNICEF and Arm Expand AI Access for Children Worldwide","meta_description":"UNICEF and Arm's AI in Play initiative aims to bring AI tools, literacy and local innovation to millions of children in emerging economies.","robots":"index","canonical_url":"unicef-arm-ai-in-play-children-emerging-economies","og_title":"","og_description":""},{"id":114,"title":"Schools Are Teaching Students to Question AI, Not Just Use It","slug":"schools-are-teaching-students-to-question-ai-not-just-use-it","short_description":"","featured_image_id":252,"intro_image_id":252,"featured_image_credit":"AP News","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":23,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":27,"ordering":0,"published_at":"2026-08-25 10:01:15","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-25 10:01:15","updated_at":"2026-08-25 10:01:15","category_name":"AI & Education","category_slug":"ai-education","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/79340f5b010279b7f3409f662aab8cd6.webp","featured_image_alt":"A blurred speaker stands beside a projected screen displaying “Ask ChatGPT.”","featured_image_width":1200,"featured_image_height":675,"intro_image_path":"/api/uploads/2026/08/79340f5b010279b7f3409f662aab8cd6.webp","intro_image_width":1200,"intro_image_height":675,"seo_title":"Schools Teach Students to Question AI, Not Just Use It","meta_description":"Schools are moving beyond ChatGPT bans, teaching students to verify AI answers, spot mistakes and know when not to trust a chatbot.","robots":"index","canonical_url":"schools-teaching-students-question-ai","og_title":"","og_description":""},{"id":112,"title":"Anthropic Launches Claude Academy for AI Learning","slug":"anthropic-launches-claude-academy-for-ai-learning","short_description":"","featured_image_id":251,"intro_image_id":251,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":23,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":4,"view_count":51,"ordering":0,"published_at":"2026-08-21 10:21:15","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-08-21 10:21:15","updated_at":"2026-08-21 10:21:15","category_name":"AI & Education","category_slug":"ai-education","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/aea18ff119b87ac6c0be05580790f85e.webp","featured_image_alt":"The Anthropic Academy logo appears in black on a cream background.","featured_image_width":739,"featured_image_height":415,"intro_image_path":"/api/uploads/2026/08/aea18ff119b87ac6c0be05580790f85e.webp","intro_image_width":739,"intro_image_height":415,"seo_title":"Anthropic Launches Claude Academy for AI Learning","meta_description":"Anthropic launches Claude Academy with free courses covering Claude, AI fundamentals, model limitations and responsible AI use.","robots":"index","canonical_url":"anthropic-launches-claude-academy-ai-learning","og_title":"","og_description":""},{"id":75,"title":"AI Is Changing Entry-Level Technology Careers","slug":"ai-is-changing-entry-level-technology-careers","short_description":"","featured_image_id":185,"intro_image_id":185,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":23,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":8,"view_count":26,"ordering":0,"published_at":"2026-07-11 15:24:18","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-11 15:24:18","updated_at":"2026-07-11 15:24:18","category_name":"AI & Education","category_slug":"ai-education","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/da16487178d95c6e64ef539f3284c187.webp","featured_image_alt":"A worried graduate in a cap and gown stands beside a smiling humanoid robot labeled “AI” against an orange background.","featured_image_width":1536,"featured_image_height":804,"intro_image_path":"/api/uploads/2026/07/da16487178d95c6e64ef539f3284c187.webp","intro_image_width":1536,"intro_image_height":804,"seo_title":"AI Is Changing Entry-Level Technology Careers","meta_description":"AI coding agents are changing junior technology work. What does the shift mean for students, developers and software companies across the Balkans?","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":69,"title":"FZF AI Platform Explores AI in Higher Education","slug":"fzf-ai-platform-explores-ai-in-higher-education","short_description":"","featured_image_id":144,"intro_image_id":144,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":23,"author_id":10,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":56,"ordering":0,"published_at":"2026-07-07 20:23:00","publish_down":null,"created_by":8,"updated_by":8,"created_at":"2026-07-07 20:23:06","updated_at":"2026-07-07 20:32:06","category_name":"AI & Education","category_slug":"ai-education","author_name":"Guest Contributor","author_slug":"guest-contributor","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/26f96a55e5cb84174a9b5bc8aed7289c.webp","featured_image_alt":"Ukim Classroom","featured_image_width":1448,"featured_image_height":1086,"intro_image_path":"/api/uploads/2026/07/26f96a55e5cb84174a9b5bc8aed7289c.webp","intro_image_width":1448,"intro_image_height":1086,"seo_title":"FZF AI Platform Explores AI in Higher Education","meta_description":"The Faculty of Philosophy in Skopje is studying how AI can support teaching, learning, and research in the humanities and social sciences.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":68,"title":"UKIM Institute Explores AI as a Tool for Gender Equality Research","slug":"ukim-institute-explores-ai-as-a-tool-for-gender-equality-research","short_description":"","featured_image_id":143,"intro_image_id":143,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":23,"author_id":10,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":3,"view_count":58,"ordering":0,"published_at":"2026-07-07 20:13:00","publish_down":null,"created_by":8,"updated_by":8,"created_at":"2026-07-07 20:11:42","updated_at":"2026-07-07 20:32:17","category_name":"AI & Education","category_slug":"ai-education","author_name":"Guest Contributor","author_slug":"guest-contributor","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/2f44aa8070d977068849995a38f731b7.webp","featured_image_alt":"St. Cyril and Methodic University","featured_image_width":1177,"featured_image_height":749,"intro_image_path":"/api/uploads/2026/07/2f44aa8070d977068849995a38f731b7.webp","intro_image_width":1177,"intro_image_height":749,"seo_title":"How UKIM Is Using AI to Support Gender Equality Research","meta_description":"Discover how UKIM's Institute for Gender Studies is using AI-powered tools, digital learning, and research platforms to support gender equality in academia","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":134,"title":"AI Leaders Question the Race to Go Faster","slug":"ai-leaders-question-the-race-to-go-faster","short_description":"","featured_image_id":295,"intro_image_id":295,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":24,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":8,"view_count":35,"ordering":0,"published_at":"2026-09-14 09:47:43","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-14 09:47:43","updated_at":"2026-09-14 09:47:43","category_name":"Analysis","category_slug":"analysis","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/4b105315c41969fb89d5357599619185.webp","featured_image_alt":"Three men appear in side-by-side close-up portraits against blue and white backgrounds.","featured_image_width":1200,"featured_image_height":800,"intro_image_path":"/api/uploads/2026/09/4b105315c41969fb89d5357599619185.webp","intro_image_width":1200,"intro_image_height":800,"seo_title":"AI Leaders Question the Pace of AI Development","meta_description":"As AI companies debate slowing frontier development, businesses and governments are still racing to adopt more AI. What does that gap mean?","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":74,"title":"The New AI Risk: Building What Big Platforms Will Absorb","slug":"the-new-ai-risk-building-what-big-platforms-will-absorb","short_description":"","featured_image_id":156,"intro_image_id":156,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":24,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":13,"view_count":29,"ordering":0,"published_at":"2026-07-10 08:00:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-09 22:51:23","updated_at":"2026-07-10 09:24:14","category_name":"Analysis","category_slug":"analysis","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/accf3cb3f6cb6f0f6aff1a7cb122d6b1.webp","featured_image_alt":"Glowing blue data panels flow into a large digital dashboard on a dark desk.","featured_image_width":1448,"featured_image_height":1086,"intro_image_path":"/api/uploads/2026/07/accf3cb3f6cb6f0f6aff1a7cb122d6b1.webp","intro_image_width":1448,"intro_image_height":1086,"seo_title":"The New AI Risk: Building What Big Platforms Will Absorb","meta_description":"As OpenAI, Anthropic, Google and Microsoft move toward AI agents, startups and companies face a harder question: what is still worth building?","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":61,"title":"Türkiye’s AI Action Plan Shows What the Balkans Are Missing","slug":"turkiye-s-ai-action-plan-shows-what-the-balkans-are-missing","short_description":"","featured_image_id":134,"intro_image_id":134,"featured_image_credit":"TRT World","featured_image_overlay":"Istanbul","featured_image_overlay_position":"bottom-left","category_id":24,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":5,"view_count":28,"ordering":0,"published_at":"2026-07-04 12:10:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-04 12:10:14","updated_at":"2026-07-04 12:12:54","category_name":"Analysis","category_slug":"analysis","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/112771e22b128dd753e938485cca1fd7.webp","featured_image_alt":"Recep Tayyip Erdogan speaking on Yapay Zeka Summit","featured_image_width":1280,"featured_image_height":854,"intro_image_path":"/api/uploads/2026/07/112771e22b128dd753e938485cca1fd7.webp","intro_image_width":1280,"intro_image_height":854,"seo_title":"Türkiye’s AI Action Plan Shows What the Balkans Are Missing","meta_description":"Türkiye’s 2026–2030 AI Action Plan offers lessons for the Western Balkans on infrastructure, skills, governance and public-sector adoption.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":85,"title":"North Macedonia’s AI Politics Problem Is Not the Technology. It Is the Judgment.","slug":"north-macedonia-s-ai-politics-problem-is-not-the-technology-it-is-the-judgment","short_description":"","featured_image_id":210,"intro_image_id":210,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":25,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":7,"view_count":39,"ordering":0,"published_at":"2026-07-19 10:07:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-19 10:07:37","updated_at":"2026-07-19 10:08:25","category_name":"Opinion","category_slug":"opinion","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/9928838d686696a62203bd11a55218c3.webp","featured_image_alt":"Skopje AI Billboard","featured_image_width":599,"featured_image_height":318,"intro_image_path":"/api/uploads/2026/07/9928838d686696a62203bd11a55218c3.webp","intro_image_width":599,"intro_image_height":318,"seo_title":"North Macedonia's AI Politics Problem Is Judgment","meta_description":"AI-generated billboards, ministry images and robotic political posts reveal a deeper problem with judgment, authenticity and public trust.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":60,"title":"Balkan Governments Are Entering the AI Era, But Not at the Same Speed","slug":"balkan-governments-are-entering-the-ai-era-but-not-at-the-same-speed","short_description":"","featured_image_id":131,"intro_image_id":131,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":25,"author_id":1,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":5,"view_count":18,"ordering":0,"published_at":"2026-07-04 11:58:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-07-04 11:58:12","updated_at":"2026-07-04 12:01:24","category_name":"Opinion","category_slug":"opinion","author_name":"Suad Seferi","author_slug":"suad-seferi","author_photo_id":null,"featured_image_path":"/api/uploads/2026/07/70254818bcd6a21a2f2d8bbf46b8bd70.webp","featured_image_alt":"AI Balkans Countries Map","featured_image_width":1408,"featured_image_height":768,"intro_image_path":"/api/uploads/2026/07/70254818bcd6a21a2f2d8bbf46b8bd70.webp","intro_image_width":1408,"intro_image_height":768,"seo_title":"Balkan Governments Enter the AI Era at Different Speeds","meta_description":"AI is entering Balkan public administration, from Serbia’s strategy to Albania’s AI minister and Montenegro’s new policy framework.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":170,"title":"North Macedonia's AI Future Starts With Infrastructure","slug":"north-macedonias-ai-future-starts-with-infrastructure","short_description":null,"featured_image_id":387,"intro_image_id":387,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":27,"author_id":7,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":0,"read_time_minutes":8,"view_count":79,"ordering":0,"published_at":"2026-10-16 14:28:34","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-10-16 14:28:34","updated_at":"2026-10-16 14:28:34","category_name":"Interviews","category_slug":"interviews","author_name":"Aleksandra Todorovikj","author_slug":"aleksandra-todorovikj","author_photo_id":null,"featured_image_path":"/api/uploads/2026/10/4a2f094d1022e341ac22ece7944d44c6.webp","featured_image_alt":"Gonet Server Room","featured_image_width":1448,"featured_image_height":1086,"intro_image_path":"/api/uploads/2026/10/4a2f094d1022e341ac22ece7944d44c6.webp","intro_image_width":1448,"intro_image_height":1086,"seo_title":"North Macedonia's AI Future Starts With Infrastructure","meta_description":"GoNet's Vladimir Stojkovski discusses AI infrastructure, local data centres, cybersecurity and North Macedonia's digital future.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":130,"title":"Sasho Jovanovski on Startup Revolution AI Summit","slug":"sasho-jovanovski-on-startup-revolution-ai-summit","short_description":"","featured_image_id":284,"intro_image_id":284,"featured_image_credit":"","featured_image_overlay":"","featured_image_overlay_position":null,"category_id":27,"author_id":7,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":7,"view_count":53,"ordering":0,"published_at":"2026-09-09 15:48:26","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-09 15:48:26","updated_at":"2026-09-09 15:48:26","category_name":"Interviews","category_slug":"interviews","author_name":"Aleksandra Todorovikj","author_slug":"aleksandra-todorovikj","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/d35f8ce5ca21b72af6e90819fa1a9e56.webp","featured_image_alt":"Sasho Jovanovski, Founder & CEO, speaks into a microphone onstage beside a screen showing his portrait.","featured_image_width":1600,"featured_image_height":1066,"intro_image_path":"/api/uploads/2026/09/d35f8ce5ca21b72af6e90819fa1a9e56.webp","intro_image_width":1600,"intro_image_height":1066,"seo_title":"Sasho Jovanovski on Startup Revolution AI Summit","meta_description":"Sasho Jovanovski discusses Startup Revolution AI Summit, bringing AI, startups, investors and global opportunities to Skopje.","robots":"index","canonical_url":"","og_title":"","og_description":""},{"id":121,"title":"Vladimir Trajkovikj on AI, Education and the Western Balkans","slug":"vladimir-trajkovikj-on-ai-education-and-the-western-balkans","short_description":"","featured_image_id":265,"intro_image_id":265,"featured_image_credit":"Prof.Dr. Vladimir Trajkovikj","featured_image_overlay":"Prof.Dr. Vladimir Trajkovikj","featured_image_overlay_position":"bottom-left","category_id":27,"author_id":7,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":8,"view_count":73,"ordering":0,"published_at":"2026-09-04 07:44:00","publish_down":null,"created_by":1,"updated_by":1,"created_at":"2026-09-01 15:56:53","updated_at":"2026-09-04 07:44:55","category_name":"Interviews","category_slug":"interviews","author_name":"Aleksandra Todorovikj","author_slug":"aleksandra-todorovikj","author_photo_id":null,"featured_image_path":"/api/uploads/2026/09/cade60a2f859b942caed9b62e84fe12f.webp","featured_image_alt":"A man in a black hoodie sits at a table with a microphone during an interview.","featured_image_width":1365,"featured_image_height":768,"intro_image_path":"/api/uploads/2026/09/cade60a2f859b942caed9b62e84fe12f.webp","intro_image_width":1365,"intro_image_height":768,"seo_title":"Vladimir Trajkovikj on AI, Education and the Western Balkans","meta_description":"Vladimir Trajkovikj discusses AI in education, research, universities and what North Macedonia needs to strengthen its AI capacity.","robots":"index","canonical_url":"vladimir-trajkovikj-ai-education-western-balkansladimir-trajkovikj-ai-education-western-balkans","og_title":"","og_description":""},{"id":98,"title":"Monika on Responsible AI, Women in Tech and the Balkans","slug":"monika-on-responsible-ai-women-in-tech-and-the-balkans","short_description":"","featured_image_id":230,"intro_image_id":230,"featured_image_credit":"","featured_image_overlay":"Monika Rizovska Tanevska","featured_image_overlay_position":"bottom-left","category_id":27,"author_id":7,"status":"published","trashed_at":null,"is_featured":0,"is_breaking":0,"is_editor_pick":0,"show_table_of_contents":1,"read_time_minutes":7,"view_count":110,"ordering":0,"published_at":"2026-08-04 09:26:00","publish_down":null,"created_by":5,"updated_by":5,"created_at":"2026-08-04 09:33:48","updated_at":"2026-08-04 09:38:46","category_name":"Interviews","category_slug":"interviews","author_name":"Aleksandra Todorovikj","author_slug":"aleksandra-todorovikj","author_photo_id":null,"featured_image_path":"/api/uploads/2026/08/79098ff2e7c8d1312618c7472b89cae8.webp","featured_image_alt":"A woman in a white blazer looks toward a sunlit window with blinds.","featured_image_width":583,"featured_image_height":371,"intro_image_path":"/api/uploads/2026/08/79098ff2e7c8d1312618c7472b89cae8.webp","intro_image_width":583,"intro_image_height":371,"seo_title":"Monika on AI, Leadership and Women in Technology","meta_description":"Women in Tech leader Monika discusses responsible AI, leadership, regional collaboration and opportunities for women in technology across the Balkans.","robots":"index","canonical_url":"","og_title":"","og_description":""}],"settings":{"directory_enabled":"","glossary_enabled":"1","tools_enabled":"","cases_enabled":"1","newsletter_enabled":"1","site_name":"AI Balkans","tagline":"The AI Hub for the Balkans.","home_title":"AI Balkans | AI News, Research & Tools for the Region","home_description":"AI Balkans is an independent AI hub by AINOW Society, covering AI news, research, tools, policy, innovation, and developments from the Balkan and beyond.","homepage_author":"Suad Seferi","homepage_published_date":"2026-06-21","contact_email":"","contact_phone":"","facebook_url":"https://www.facebook.com/aibalkans","facebook_app_id":"1532026741761615","instagram_url":"","linkedin_url":"https://www.linkedin.com/company/aibalkans/","twitter_url":"","youtube_url":"","google_analytics_id":"G-JY4KMNF045","footer_note":"© {year} {site}. The AI hub for news, research, and innovation across the Balkans by [AINOW Society](https://www.ainow.mk)","logo_url":"/api/uploads/2026/06/dddb2ea4906fdfdf1af178dd0f07d699.webp","logo_dark_url":"/api/uploads/2026/06/113a1e37601fc22fdda3469615dac9b5.webp","footer_logo_url":"/api/uploads/2026/06/113a1e37601fc22fdda3469615dac9b5.webp","favicon_url":"/api/uploads/2026/06/9cf02dc267fd3cc3af1842d3b513cb6b.webp","social_image_url":"/api/uploads/2026/06/f8fd9dcc005f1239df1845e613dbb22a.webp","theme_ink":"#111111","theme_paper":"#ffffff","theme_accent":"#bb1919","theme_muted":"#f3f4f6","recaptcha_site_key":"6LerWTAtAAAAADk_M6IwlSULuhE7Ug-UNZBz75Xe","maintenance_mode":"","maintenance_message":""},"trending_tags":[{"name":"Anthropic","slug":"anthropic","uses":15},{"name":"North Macedonia","slug":"north-macedonia","uses":15},{"name":"Claude","slug":"claude","uses":14},{"name":"OpenAI","slug":"openai","uses":13},{"name":"ChatGPT","slug":"chatgpt","uses":11},{"name":"Education","slug":"education","uses":11},{"name":"Kosovo","slug":"kosovo","uses":10},{"name":"AI Balkans","slug":"ai-balkans","uses":9}]}}