Sweden bans ChatGPT and Claude on government staff computers

Key takeaways

• Sweden's Government Offices (Regeringskansliet) will block ChatGPT and Claude on staff work computers from Monday, 26 October 2026, Sveriges Radio's news service Ekot reported on 23 October.
• Staff are being directed to Microsoft Copilot, which the Government Offices approved after a review that tightened the service's security, according to Ekot, SVT and Finwire.
• Ekot reports that digitalisation director Mikael Wettercrantz warned staff that AI services used without an agreement can have consequences for information security.
• This is not a ban on all AI. Ekot reports that some staff can still reach ChatGPT and Claude through an in-house AI assistant platform.
• Other Swedish agencies are doing the same. The Swedish Tax Agency (Skatteverket) and the National Board of Institutional Care (SiS) already block similar services, and the Enforcement Authority (Kronofogden) is reviewing the question.

Sweden bans ChatGPT and Claude on staff work computers at its Government Offices, headquartered at Rosenbad in Stockholm, from Monday, 26 October, citing information security. Staff will be directed to Microsoft Copilot instead, Ekot, SVT and Finwire report.

Sweden bans ChatGPT and Claude: what the Government Offices decided

Ekot first reported the block on 23 October 2026, and its report has since been carried by SVT Nyheter and the financial wire Finwire, published on Börsvärlden. All three accounts agree on the core facts. From Monday, people working at the Government Offices will no longer be able to open OpenAI's ChatGPT or Anthropic's Claude from their official computers.

The Government Offices (Regeringskansliet) is the central civil service behind the Swedish government. It includes the Prime Minister's Office, the ministries and the Office for Administrative Affairs. The rule applies to work devices across that organisation. It is not a national ban, and it does not apply to the public or to private companies in Sweden.

SVT says it has contacted the Government Offices for comment, but its report does not include a response. As of 23 October, Bullquill could not find a press release about the change on the government's website, regeringen.se. The details below come from Ekot's report, which was based on a notice posted on the internal intranet, and from the outlets that confirmed it.

When does the block start?

The reports say the block starts "on Monday". Ekot published on Friday, 23 October 2026, so that means Monday, 26 October 2026. None of the reports describe a transition period.

Why the Government Offices is blocking ChatGPT and Claude

The stated reason is information security. According to Ekot, as reported by Finwire, Mikael Wettercrantz, the Government Offices' director of digitalisation, wrote on the intranet that AI services used without an agreement can have consequences for information security.

The key phrase is "without an agreement". On consumer and individual plans, the provider's standard terms decide how prompts and uploaded files are stored, kept and, in some cases, used to improve models. An organisation-wide contract usually sets those conditions in advance. Government staff routinely handle unpublished policy drafts, negotiating positions and material that is classified or subject to secrecy rules. For staff with that kind of work, a tool without such a contract is hard to justify.

The reports do not say that either company did anything wrong or that any data has leaked. The decision is about procurement and control, not about a known incident. OpenAI and Anthropic had not commented publicly on the Swedish move at the time of writing.

Which AI tools Swedish government staff can still use

Staff are being steered to Microsoft's Copilot. According to SVT and Sejfer, the Government Offices reviewed the service and then raised its security level before approving it for use. The reports do not say which Copilot product or licence tier is involved.

The block also has an exception. Ekot reports, via Finwire, that some employees will still be able to use ChatGPT and Claude through an in-house platform for AI assistants. That suggests the underlying models are not ruled out, as long as access runs through a channel the Government Offices controls. Sejfer describes the change the same way: work is being moved to an approved alternative, and AI use as a whole is not being stopped.

The Government Offices had already restricted the Chinese chatbot DeepSeek before this decision, Finwire reports.

Other Swedish agencies with similar AI blocks

Ekot's survey found that the Government Offices is not alone. The table below sets out what has been reported about Swedish public bodies, alongside the two comparisons readers most often ask about: the UK civil service and the US federal government.

How Sweden compares with UK and US rules on AI in government

United Kingdom: guidance, not a ban

The UK has managed civil service use of chatbots through guidance rather than blocking specific brands. Its first generative AI guidance for civil servants in 2023 told officials not to enter sensitive information or personal data into public tools, UKAuthority reported at the time. That guidance was later replaced by the Generative AI Framework for HMG and then, in February 2025, by the AI Playbook for the UK Government. The Playbook reminds departments that AI systems can process personal data, tells them to seek data protection advice, and says staff should follow any organisation-specific policies, especially ones about security and data handling. Each department sets its own approved-tool list, which is broadly how the Swedish Government Offices is now working.

United States: a political fight centred on Claude

In the US, the dispute has been about one vendor rather than about data handling in general. On 27 February 2026, President Donald Trump directed federal agencies to stop using Anthropic's technology, and Defense Secretary Pete Hegseth moved to designate the company a supply-chain risk, according to a Congressional Research Service report. The dispute followed Anthropic's refusal to allow Claude to be used for mass domestic surveillance or fully autonomous weapons. Judge Rita Lin of the US District Court for the Northern District of California granted a preliminary injunction in March that blocked both the agency-wide ban and the Pentagon's steps to enforce the designation, FedScoop reported. The case was still being argued in the summer.

Sweden's step has a different logic. It covers both ChatGPT and Claude, it is presented as a security and contract question, and it leaves a route to the same models through an approved internal platform.

The prime minister who uses ChatGPT

The decision lands awkwardly for Prime Minister Ulf Kristersson. In August 2025 he told the business daily Dagens industri that he used AI tools, including ChatGPT and the French service Le Chat, quite often, "if for nothing else than for a second opinion", as quoted in English by TechSpot. Critics in the Swedish press questioned the practice. His spokesperson, Tom Samuelsson, said at the time that no security-sensitive information was shared with the tools.

The reports do not say whether the new block will change how the prime minister himself works, or whether ministers' political staff will be treated differently from civil servants.

What is still unknown

• Whether the Government Offices will publish its decision or the intranet notice.
• Which staff qualify for the in-house platform, and which models it runs.
• Whether the block also covers phones and personal devices used for work.
• Whether other agencies, including Kronofogden, will follow before the end of the year.

What it means

For anyone searching "Sweden bans ChatGPT", the short answer is that the Swedish government has not banned the chatbot for the public. Its own central offices have blocked direct access to ChatGPT and Claude on work computers and moved staff to a contracted tool. The pattern is spreading across Swedish agencies, from the Tax Agency to SiS. It matches how many European governments are trying to manage generative AI: they allow it through procured, controlled services and close off consumer accounts that sit outside a contract.

For OpenAI and Anthropic, the message is that public-sector access increasingly depends on enterprise agreements, data-handling terms and local controls, not on the popularity of a consumer app. For Microsoft, the decision shows the advantage of already being inside government IT contracts. Readers following Europe's wider rulebook can see how this sits next to the EU AI Act timeline. Recent Bullquill coverage of the companies involved includes Anthropic's Claude Duet agent app, OpenAI's $30 billion fundraising and DeepSeek's reported $15 billion round.

Sources

• Sveriges Radio Ekot: Regeringskansliet blocks ChatGPT and Claude (23 Oct 2026)
• SVT Nyheter: AI services blocked at the Government Offices (23 Oct 2026)
• Börsvärlden/Finwire: Government Offices blocks use of ChatGPT and Claude (23 Oct 2026)
• Sejfer: Government Offices blocks ChatGPT and Claude (23 Oct 2026)
• GOV.UK: AI Playbook for the UK Government (10 Feb 2025)
• UKAuthority: Civil servants told to use generative AI with caution (2023)
• Congressional Research Service: Federal Government and Anthropic
• FedScoop: Record "has gotten worse for the government" in Anthropic dispute, judge says
• TechSpot: Swedish prime minister admits using AI chatbots (Aug 2025)
• Wikipedia: Government Offices of Sweden