Sweden Blocks ChatGPT and Claude on Government Work Computers

By Suad Seferi · 26 October 2026 at 09:33

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. The 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. Employees 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. However, 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. Why Sweden Is Restricting AI Access The Swedish Government Offices want employees to use AI services that have been formally reviewed and approved for official work. According to Sveriges Radio, the decision follows internal concerns about staff using AI platforms without appropriate agreements or security controls. Mikael Wettercrantz, the Government Offices' director of digitalisation, warned in an internal communication that using AI services without agreements could have consequences for information security. The concern is particularly relevant to government institutions handling confidential documents, internal correspondence, policy proposals and potentially sensitive personal information. When employees enter information into publicly accessible AI services, the organisation must understand how that information is processed, stored and protected. Without appropriate contractual arrangements and technical safeguards, institutions may have limited control over where information goes and who can access it. The 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. The 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. ChatGPT and Claude Will Remain Available Through Intric An important detail distinguishes Sweden's decision from an outright prohibition. In 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. Intric is a Swedish technology company specialising in AI infrastructure for public institutions and organisations handling sensitive information. In July 2026, the company announced that it had won a contract to supply its AI platform to the Swedish Government Offices. The platform allows organisations to develop and operate AI assistants using their own information under defined governance arrangements. Intric says its infrastructure operates in Sweden, supports deployment within customers' own environments and provides institutions with greater control over their data. The arrangement illustrates the distinction between accessing an AI model through a public website and using the same underlying technology through an institutionally managed service. In 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. Other Swedish Authorities Have Already Introduced Restrictions The Government Offices are not the first Swedish public institution to limit access to AI services. According to Sveriges Radio's investigation, the Swedish Tax Agency (Skatteverket) and the National Board of Institutional Care (SiS) have already introduced similar restrictions. The Swedish Enforcement Authority (Kronofogden) is also reviewing the issue before deciding whether to introduce additional controls. Earlier concerns emerged at the Swedish Transport Administration, where an internal audit identified widespread and potentially risky use of several AI services. The audit raised concerns about both the sensitivity of information employees entered into AI tools and how they subsequently handled AI-generated responses. These 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. Sweden's Agency for Digital Government, Digg, has already issued guidance recommending that public authorities keep generative AI use under organisational supervision. Its recommendations emphasise protecting confidential information, controlling access to institutional data and establishing clear rules about which information employees may process using AI services. A Wider Question for European Public Institutions Sweden's decision comes as public administrations across Europe are exploring how generative AI can support administrative work without exposing sensitive information or weakening accountability. The Swedish approach demonstrates one possible response: restricting uncontrolled access while maintaining AI capabilities through approved services. For 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. There is also a practical distinction between the security of an AI model and the security of the environment in which that model operates. An organisation may use the same underlying language model through different services, each with its own contractual arrangements, permissions and information-handling controls. This is particularly relevant to public institutions considering AI adoption wi…