Sweden Blocks ChatGPT, Claude in Ministries
Sweden's Government Offices, which run the prime minister's office and all ministries, will block ChatGPT and Claude on staff work computers from Oct. 26 and steer staff to Microsoft Copilot, citing information security, Sveriges Radio's Ekot reported Oct. 23.
Sweden's Government Offices, the body that runs the prime minister's office and every ministry, will block ChatGPT and Claude on staff work computers starting Monday, Oct. 26, and steer employees to Microsoft Copilot instead, public broadcaster Sveriges Radio's Ekot reported on Friday, Oct. 23. The Government Offices cited information security risks.
Key takeaways
- From Oct. 26, staff at the Government Offices (Regeringskansliet) can no longer reach ChatGPT or Claude directly from work computers.
- Microsoft Copilot stays approved after a security review, though the press service would not say what was changed.
- The Swedish Tax Agency and the National Board of Institutional Care already block AI services, and the Enforcement Authority is weighing a similar step.

Why Sweden's Government Offices blocked ChatGPT and Claude
Mikael Wettercrantz, the Government Offices' director of digitalisation, announced the change in a post on the staff intranet, according to Ekot's reporting as relayed by several outlets. He warned that using AI services without an agreement in place could affect the organization's information security. He declined an interview with Ekot.
The Government Offices said the decision does not mean all use of generative AI is banned. Some employees will still be able to use ChatGPT and Claude models through an AI assistant platform the Government Offices built itself, the press service confirmed to Ekot. It did not say which staff have that access or how the platform handles data.
The move follows an earlier restriction on China's DeepSeek inside the Government Offices. OpenAI and Anthropic have not commented publicly.
Copilot stays, but its Claude option is a separate switch
Microsoft Copilot remains an approved AI service within the Government Offices, which said it reviewed the tool and added measures to protect information.
The choice has a wrinkle. Microsoft 365 Copilot can itself run Anthropic's Claude models in features such as Researcher and Copilot Studio. Microsoft's own documentation says those Anthropic models are "currently excluded from the EU Data Boundary," and that customers in the EU, EFTA and the U.K. have them disabled by default. Administrators must opt in through a setting called "AI providers operating as Microsoft subprocessors." It has not been reported whether the Government Offices turned that setting on.
In practice, the ban targets the consumer and direct web services, not the underlying models. Copilot itself has changed shape this year, after Microsoft merged its two Copilot apps in August.
Other Swedish agencies restricting AI tools
Ekot's survey found that the Swedish Tax Agency (Skatteverket) and the National Board of Institutional Care (SiS) have already introduced blocks on AI services. The Swedish Enforcement Authority (Kronofogden) is working on the issue but has not confirmed a decision.
An internal audit at the Swedish Transport Administration (Trafikverket) this year also found extensive use of about 10 AI services, Swedish tech site HurBra reported.
Context: Sweden wants more public-sector AI, under rules
The block does not reverse national policy. In January 2025 the government received national guidelines for generative AI from the Agency for Digital Government (Digg) and the Authority for Privacy Protection (IMY). According to the government's announcement dated Jan. 21, 2025, they set out 18 guidelines across seven areas, including information security, data protection, copyright and procurement, with the stated goal of increasing AI use in the public sector.
The Government Offices' approach fits that framework: keep AI, but only through tools covered by contracts and controls it manages. It is an internal IT policy, not a regulatory order against a provider.
What's next
The block takes effect Monday. Open questions include how many staff get access through the in-house platform, what Copilot settings were tightened, and whether Kronofogden and other agencies follow.
FAQ
Is Sweden banning ChatGPT nationwide? No. The block applies to work computers at the Government Offices, which serve the prime minister's office and the ministries. Some other agencies have set their own blocks.
Can Swedish government staff still use AI? Yes. Microsoft Copilot remains approved, and some employees can reach ChatGPT and Claude models through an in-house AI platform, according to the press service.
References
- Anthropic models in Microsoft Online Services — Microsoft Learn (updated Sept. 18, 2026)
- Regeringen har tagit emot nationella riktlinjer för generativ AI — Government Offices of Sweden (Jan. 21, 2025)
- Regeringskansliet spärrar ChatGPT och Claude — HurBra (Oct. 23, 2026)
- Regeringskansliet spärrar ChatGPT och Claude — Sejfer (Oct. 23, 2026)
- Sweden's Government Offices Block ChatGPT and Claude, Approve Copilot — CEOWORLD Magazine (Oct. 24, 2026)
Editor responsible for AI Newsway. Writes about AI, developer tools, cloud and productivity technology.
Comments 0
Comments are currently disabled.