From today, staff at the Government Offices of Sweden can no longer open ChatGPT or Claude in a browser on their work computers. They are being pointed to Microsoft Copilot instead, while a smaller group keeps access to both models through a Swedish-built platform.
By HahuPrime News Desk | October 26, 2026
STOCKHOLM — The internal auditors at Sweden’s Transport Administration found about ten different AI services in use across the agency earlier this year. “We have seen several examples of problematic use of open AI services,” they wrote, according to Sveriges Radio’s news service Ekot. “The examples concern both the sensitivity of the information that was entered and how the AI-generated answers were handled.”
That kind of finding is now shaping policy at the top of the Swedish state. On Monday, 26 October, the Government Offices, known in Swedish as Regeringskansliet, switched off browser access to OpenAI’s ChatGPT and Anthropic’s Claude on staff computers. The Government Offices supports the prime minister and the ministries, and the block applies to its own employees.
It is not a nationwide ban. Ordinary people, companies, schools and universities in Sweden can still use both chatbots. The measure covers work computers inside the Government Offices, and a handful of other agencies that have made their own decisions.
What the Government Offices changed
Ekot first reported the plan on 23 October, based on internal information it had seen. In an email to public broadcaster SVT, the Government Offices confirmed the step. “The browser interface for the AI services ChatGPT and Claude will no longer be reachable from the Government Offices’ work computers from 26 October,” it wrote.
The same email said all employees have access to Microsoft’s AI tool, Copilot. According to Ekot, the Government Offices reviewed Copilot and tightened its security settings before the switch. The press office declined to describe those changes, saying they concern security matters.
Mikael Wettercrantz, the director of digitalisation at the Government Offices, turned down an interview with Ekot. In a comment posted on the staff intranet, he said that using AI services without an agreement in place can have “consequences for our information security.”
The models themselves are not gone completely. The Government Offices told SVT that staff still have access to ChatGPT and Claude language models within Intric, a platform for building AI assistants. The Next Web reported that Intric is a Stockholm company that won a contract with the Government Offices in July and says its platform runs on servers in Sweden. Only part of the staff will have that access, according to Ekot.
Why agreements matter for government data
The concern is less about the chatbots’ answers and more about where the questions go. When an official pastes a draft memo or case details into a consumer AI service, the text leaves the organisation. Without a contract that sets out how the provider stores, protects and uses that data, the agency has little control over it.
Euronews reported that officials worried sensitive information entered into the models without such an agreement could be stored and used for training. TechRadar noted that missing enterprise contracts also make it harder to audit how staff use AI tools, and can raise data protection questions when information is processed in data centres outside Europe.
Security teams call this pattern “shadow IT”: tools that employees adopt on their own because they are useful, before the organisation has approved them. Copilot, by contrast, is covered by an agreement between Microsoft and the Swedish government and has passed a security review, Euronews reported.
The Government Offices had already limited the use of DeepSeek, the Chinese AI service, Ekot said. ChatGPT and Claude now join it on the list of tools staff cannot reach directly.
Other Swedish agencies are moving too
Ekot’s survey found the Government Offices is not alone. The Swedish Tax Agency (Skatteverket) and the National Board of Institutional Care (Statens institutionsstyrelse, SiS) have also introduced blocks on these AI services. The Swedish Enforcement Authority (Kronofogden) is preparing a position on the issue.
The decisions are being taken agency by agency rather than through one national law. That means rules can differ between offices, and some agencies may approve tools that others block. For staff, the practical message is simpler: use the services your employer has signed contracts for, and keep sensitive material out of everything else.
The move comes as governments everywhere try to balance two pressures. Civil servants want AI help with drafting, translation and summaries, and many already use it. At the same time, ministries handle personal data, security matters and policy drafts that cannot be allowed to leak. Sweden’s answer, for now, is to channel AI use through a small number of approved routes.
What it means beyond Sweden
For readers in Ethiopia and elsewhere in Africa, where government offices, banks and universities are adopting AI tools quickly, the Swedish case is a useful reference. The lesson is not that chatbots are dangerous in themselves. It is that institutions need written agreements, clear rules on what may be typed into an AI tool, and a record of who uses what.
Neither OpenAI nor Anthropic had publicly commented on the Swedish decision in the reports reviewed by HahuPrime. Both companies sell business and government versions of their products with contractual data protections, which are the kind of agreements the Government Offices says are missing from browser use.
Whether other Swedish authorities follow the Tax Agency and SiS, or opt for approved platforms like Intric, will become clearer as Kronofogden and others finish their reviews. For now, the Government Offices’ staff start the week with one fewer browser tab and a single approved assistant.
Related on HahuPrime: OpenCode Breach Leaves Developers With Surprise AI Bills, OpenAI Brings Its Models to Google Cloud and Anthropic Puts 5% of Its Shares Into a Public Benefit Fund.
Sources
- Sveriges Radio (Ekot): Regeringskansliet spärrar AI-tjänsterna Chat GPT och Claude
- SVT Nyheter: AI-tjänster spärras på Regeringskansliet
- Euronews: Swedish government blocks ChatGPT and Claude. Here’s why
- The Next Web: Sweden’s Government Offices block ChatGPT and Claude on work computers
- TechRadar: ChatGPT and Claude blocked by Swedish government over “problematic” security concerns
- Cybernews: Swedish government blocks ChatGPT and Claude, citing security risks
