What happened
Sweden’s Government Offices (Regeringskansliet), the body that supports the prime minister and the cabinet ministries, is blocking access to ChatGPT and Claude on staff work computers. Sveriges Radio’s news programme Ekot reported the decision first, on Friday 23 October 2026. SVT then reported that the Government Offices confirmed it in an email, saying the web interfaces of ChatGPT and Claude will be unreachable from government work computers from 26 October 2026.
We could not open the Sveriges Radio article (its site disallows automated access), so the account below rests on SVT’s report of the Government Offices’ own statement, plus secondary coverage. We have not seen a published decision document.
Key details
- What is blocked (SVT): the web versions of ChatGPT (OpenAI) and Claude (Anthropic) on work computers at the Government Offices.
- What stays available (SVT): in its email, the Government Offices said that staff still have access to the ChatGPT and Claude language models through Intric, a platform the Government Offices uses. All employees have access to Microsoft Copilot, which the Government Offices approved after a review that raised its security.
- Stated reason: SVT’s report gives no reason beyond the Copilot security review. Notebookcheck, citing Decrypt, reports that digitalisation director Mikael Wettercrantz wrote on the staff intranet that using AI services without a formal agreement can affect information security and that “we have seen several examples of problematic use of open AI services.” We have not verified that quote against the original.
- Other agencies (SVT): the Swedish Tax Agency (Skatteverket) and the National Board of Institutional Care (Sis) had already introduced similar blocks, and the Swedish Enforcement Authority (Kronofogden) is considering one.
- Company responses: none of the reports we read includes a comment from OpenAI, Anthropic or Microsoft.
Checked against SVT’s report and secondary coverage on 26 October 2026.
Why it matters
The change is narrow but concrete: civil servants at one of Europe’s governments move from the public chat websites to Copilot for everyday use, while a controlled internal platform keeps access to the same underlying models. The reports present it as a question of contracts and security controls over how data is handled, not a verdict on the quality of any model.
For organisations weighing AI policies, the pattern is worth noting: block the open consumer interface, approve one reviewed tool, and keep model access behind an internal platform. This is an early report based on a government email relayed by media, so details may change.
Disclosure: Claude, made by Anthropic, was one of the AI tools used to research and draft this article, and Claude is one of the services named in it. Anthropic is a potentially affected party; we have kept to what the reports say.
Sources
- AI-tjänster spärras på Regeringskansliet (AI services blocked at the Government Offices) Primary source , SVT (quotes the Government Offices' email statement)
- Chat GPT och Claude förbjuder Regeringskansliet Sveriges Radio (Ekot)
- Swedish Government Offices block ChatGPT and Claude Daily Northern
- Sweden's government blocks ChatGPT and Claude, keeps Copilot Notebookcheck