EU Proposes Rights Checks for High-Risk AI Systems
By Suad Seferi ·

Key takeaways
- The European Commission has opened a consultation on a proposed Fundamental Rights Impact Assessment template to help organisations evaluate risks before deploying certain high-risk AI systems.
- The assessment obligations under Article 27 of the EU AI Act are scheduled to apply from December 2, 2027. They cover specified public and private-sector deployers, not every organisation using artificial intelligence.
- Public feedback is open until January 15, 2027. Eligible organisations interested in testing the draft template must express interest by December 4, 2026.
The European Commission has opened a public consultation on a proposed assessment template designed to help organisations identify risks to fundamental rights before deploying certain high-risk artificial intelligence systems. Announced on November 16, the initiative aims to make one of the EU AI Act's most consequential requirements easier to implement. Public authorities and certain private organisations will be required to examine how their use of high-risk AI could affect individuals, including their rights to privacy, equality and fair treatment. The consultation focuses on a Fundamental Rights Impact Assessment (FRIA), a requirement established under Article 27 of the AI Act. The Commission is seeking feedback on the draft template until January 15, 2027, while organisations interested in testing it can apply until December 4, 2026. What organisations will need to assess The proposed framework addresses a question that becomes increasingly important as AI enters public services and other sensitive activities: what happens when an automated system influences decisions affecting people's lives? Under Article 27, covered organisations must assess the potential consequences of using certain high-risk AI systems before deploying them. The assessment must describe how the system will be used, who could be affected, what risks it creates and which safeguards are available. Organisations will need to document the expected duration and frequency of use, identify potentially affected groups, explain human oversight arrangements and establish measures for addressing identified risks. The requirements are particularly relevant where AI systems influence decisions involving access to essential services or financial opportunities. For example, a financial institution using a high-risk AI system to assess creditworthiness may need to consider whether the system could produce discriminatory outcomes or unfairly disadvantage particular groups. A public authority deploying AI in a sensitive decision-making process may similarly need to examine whether individuals could face unequal treatment or have difficulty challenging an automated assessment. The obligation does not apply to every organisation using AI. Article 27 covers specified deployers and use cases, including public bodies, certain private providers of public services and operators of particular high-risk systems involving creditworthiness and insurance. Brussels wants to simplify compliance The draft template was prepared by the European AI Office with input from the European Union Agency for Fundamental Rights. It provides a common set of questions intended to guide organisations through the assessment process. The Commission has also published a draft explanatory notice clarifying which organisations are covered, when assessments are required and how responsibilities differ between AI system providers and deployers. That distinction is important. A company developing an AI system is responsible for meeting the obligations applicable to its product. An organisation using that system may face additional responsibilities because the risks depend on how and where the technology is deployed. The same AI system could have different consequences depending on the institution operating it, the individuals affected and the decisions it supports. Under the proposed approach, organisations would use the finalised template to prepare their assessments and notify the relevant market surveillance authorities of the results. The Commission also plans to make the adopted template available through an automated tool on the AI Act Single Information Platform. This could help smaller organisations complete assessments without developing their own documentation processes from the beginning. However, a standard template cannot replace the underlying evaluation. Organisations will still need to identify risks specific to their systems and demonstrate how those risks are addressed. The requirements begin in December 2027 The fundamental rights assessment obligation is scheduled to apply from December 2, 2027. That date forms part of the revised implementation timetable for the EU AI Act's high-risk AI requirements. The Commission is developing the assessment template ahead of the deadline to give organisations time to understand their responsibilities. The consultation also includes an opportunity for eligible organisations to test the proposed questionnaire using their own AI systems. Participants will receive support from the European AI Office, including an introductory meeting, assistance during testing and a final discussion of the results. Participation is voluntary and does not alter an organisation's legal obligations. The Commission may select participants to ensure representation across different sectors, use cases, organisation sizes and EU member states. Why this matters beyond the European Union Although the AI Act is EU legislation, its effects can extend to organisations operating outside the bloc. Companies in the Western Balkans increasingly develop software, provide outsourced technology services and supply digital solutions to European customers. Where their activities fall within the AI Act's territorial scope, they may also face relevant regulatory obligations. For organisations in North Macedonia, Albania, Serbia, Kosovo, Montenegro and Bosnia and Herzegovina, the new assessment framework provides an example of the documentation and accountability expectations associated with regulated AI deployment. However, the consultation does not create a new legal obligation for all Balkan organisations. Applicability depends on the nature of the AI system, the organisation's role and its connection to the EU market. The proposed template may also serve as a practical reference for regional public institutions considering AI in sensitive administrative processes, even where EU requirements do not directly apply. Assessing risks before implementation…
Frequently asked questions
What is a Fundamental Rights Impact Assessment?
A Fundamental Rights Impact Assessment (FRIA) evaluates how deploying a high-risk AI system could affect people's fundamental rights. It examines potential risks, affected individuals, human oversight arrangements and measures for addressing harmful outcomes.
Which organisations must conduct these assessments?
Article 27 covers specified deployers of high-risk AI systems, including public bodies, certain private entities providing public services and organisations operating particular high-risk systems involving creditworthiness or life and health insurance.
When will the assessments become mandatory?
The relevant requirements are scheduled to apply from December 2, 2027. The Commission is consulting on the assessment template before that date.
Does the EU AI Act apply to companies in the Western Balkans?
It can apply to organisations outside the EU when their activities fall within the regulation's territorial scope. However, the requirements do not automatically apply to every company or public institution in the Western Balkans.