PolicyNews

Swedish government offices block ChatGPT and Claude: what the ban covers and why

Sweden's Government Offices block ChatGPT and Claude in the browser on staff work computers from Monday, October 26, 2026, as Sveriges Radio's Ekot reported and the offices confirmed to SVT; staff are pointed to Microsoft Copilot. What the ban covers, who can still use the models, and how it compares with other government restrictions.

A glass monitor with two dashed app tiles under a lit lock and one solid tile beside them, on red and black, titled From October 26

What’s new, in 20 seconds

  1. Sweden's Government Offices (Regeringskansliet) block ChatGPT and Claude on staff work computers from Monday, October 26, 2026, Sveriges Radio's Ekot reported on October 23. Regeringskansliet confirmed the block to SVT. Staff are directed to Microsoft Copilot, which Ekot reported was approved after a review and had its security tightened.
  2. Digitalization director Mikael Wettercrantz warned staff on the intranet that AI services used without an agreement can affect information security, according to Swedish outlets citing Ekot. The press office declined to give details because the matter concerns security.
  3. It isn't a total ban: staff still have access to the ChatGPT and Claude models through Intric, the offices' AI-assistant platform, Regeringskansliet told SVT. DeepSeek was already restricted, and the Swedish Tax Agency and the National Board of Institutional Care had introduced similar blocks, Ekot reported.
Contents

Sweden bans ChatGPT and Claude inside its central government offices: from Monday, October 26, 2026, staff at Regeringskansliet, the Government Offices that serve the Swedish government and its ministries, won't be able to reach OpenAI's ChatGPT or Anthropic's Claude in a browser on their work computers. Sveriges Radio's news program Ekot reported the change on Friday, October 23, and Regeringskansliet confirmed it in an email to SVT. Staff are being pointed to Microsoft's Copilot instead, and the stated concern is information security.

This story is based on Regeringskansliet's email to SVT, as quoted in SVT's report of October 23 (updated October 24), and on Ekot's report as relayed by Computer Sweden, Bulletin, Börsvärlden (Finwire), Breakit and Sejfer, checked October 25, 2026. Sveriges Radio's own page couldn't be opened, and Regeringskansliet hasn't published a statement on its website. Swedish quotes are 9to5AI's translations.

The short version

  • What's blocked: the browser interface of ChatGPT and Claude on Regeringskansliet staff work computers ("tjänstedatorer"), in Regeringskansliet's own words to SVT.
  • When: from October 26, 2026, the date Regeringskansliet gave SVT.
  • Why: AI services used "without an agreement" can affect information security, according to digitalization director Mikael Wettercrantz's intranet note, as reported by Ekot.
  • What staff use instead: Microsoft Copilot, which Regeringskansliet says all staff can use; Ekot reported it was approved after a review, with tightened security.
  • Exceptions: staff still have access to the ChatGPT and Claude models through Intric, Regeringskansliet's AI-assistant platform.

What does the ban cover?

The block applies to Regeringskansliet staff computers, and it isn't a ban on AI. In its email to SVT, Regeringskansliet wrote that "the browser interface for the AI services ChatGPT and Claude will no longer be reachable from Regeringskansliet's work computers from October 26" ("Webbläsargränssnittet för AI-tjänsterna ChatGPT och Claude kommer inte längre att nås från Regeringskansliets tjänstedatorer från och med den 26 oktober"). Sejfer, citing Ekot, described the same thing as a block on "direct access from the service computers" ("den direkta åtkomsten från tjänstedatorerna").

Employees keep access to the same models by another route. "Within Intric, staff still have access to the language models ChatGPT and Claude," Regeringskansliet told SVT ("Inom ramen för Intric har medarbetare fortfarande tillgång till språkmodellerna ChatGPT och Claude"). Ekot's version, as relayed by Börsvärlden, was narrower: "some staff will continue to be able to use ChatGPT and Claude through an in-house platform for AI assistants" ("Viss personal kommer fortsatt kunna använda ChatGPT och Claude genom en egenutvecklad plattform för AI-assistenter"). Neither says how many people work on the affected computers.

Regeringskansliet's wording covers the two services' browser interface on staff work computers. It doesn't mention phones, personal devices or the companies' APIs.

Why did Regeringskansliet block ChatGPT and Claude?

The reports give one reason: control over where government information goes. Wettercrantz declined an interview with Ekot, but in a comment on the internal intranet he warned that "AI services used without an agreement can affect information security," Bulletin reported ("AI-tjänster som används utan avtal kan påverka informationssäkerheten"). Computer Sweden quoted his phrase as "consequences for our information security" ("konsekvenser för vår informationssäkerhet").

The press office "does not want to go into details because it concerns security matters," according to Bulletin ("presstjänsten vill inte gå in på detaljer eftersom det rör säkerhetsfrågor").

The reports don't say whether the decision followed a specific incident, and they don't suggest one did.

What can staff use instead?

Microsoft Copilot. Regeringskansliet told SVT that all staff have access to Microsoft's AI tool. "Instead, staff are referred to Microsoft's AI tool Copilot, which has been approved after a review," Sejfer wrote ("Personalen hänvisas i stället till Microsofts AI-verktyg Copilot, som har godkänts efter en översyn"). According to Bulletin, the security around Copilot was also tightened after the review. Regeringskansliet hasn't said what changed.

The dividing line is whether a service is approved and covered by an agreement, not which model it runs: the two companies' models remain available, but only through Intric. For how Copilot's business plans handle data compared with ChatGPT's, see our ChatGPT vs Microsoft Copilot comparison.

Have other Swedish agencies done the same?

Yes, according to Ekot. Its survey found similar blocks at Skatteverket, the Swedish Tax Agency, and at Statens institutionsstyrelse (SiS), the National Board of Institutional Care. Kronofogden, the Swedish Enforcement Authority, was reportedly considering one, though Sejfer said its material didn't confirm a decision. Regeringskansliet had already restricted DeepSeek before this change.

How does this compare with other governments' AI restrictions?

Many government AI restrictions have targeted DeepSeek over its Chinese ownership; our DeepSeek ban explainer lists them. Sweden's move is different because it covers two US services while allowing a third. A comparable case is the US House of Representatives:

Government bodyRestrictionDateReason givenSource
Regeringskansliet (Sweden)ChatGPT and Claude browser interface blocked on staff work computers; Copilot available to all staff; the models stay reachable through IntricFrom Oct 26, 2026Information security for AI used without an agreementRegeringskansliet, via SVT; Ekot, via Computer Sweden and Bulletin
Skatteverket and SiS (Sweden)Similar AI blocksBefore Oct 2026 (date not reported)Not reportedEkot, via Sejfer
US House of RepresentativesFree ChatGPT banned; paid version allowed in limited waysJune 2023Not detailed in the Axios reportAxios
US House of RepresentativesMicrosoft Copilot banned and blocked on House Windows devicesMarch 2024"Threat of leaking House data to non-House approved cloud services"Axios

Table 1Like the House's 2024 Copilot ban, Sweden's block turns on whether data leaves approved services, not on the model. Source: Regeringskansliet via SVT and Ekot via Swedish outlets (October 23–24, 2026); Axios (March 29, 2024).

Last verified: October 25, 2026

For a sense of how ChatGPT and Claude differ as products, see our Claude vs ChatGPT vs Gemini guide.

What's confirmed, reported and uncertain

  • Confirmed (Regeringskansliet, in an email quoted by SVT): the browser interface of ChatGPT and Claude can't be reached from staff work computers from October 26, 2026; all staff have access to Microsoft's AI tool; staff still have access to the ChatGPT and Claude models through Intric.
  • Reported (Ekot, via five Swedish outlets that agree on the details): Wettercrantz's intranet warning, Copilot's approval after a review and its tightened security, the press office declining to give details, the earlier DeepSeek restriction and the blocks at Skatteverket and SiS.
  • Not available: a statement on Regeringskansliet's website, and Sveriges Radio's original article, which 9to5AI couldn't open.
  • Uncertain: how many staff are affected, whether every employee or only some can use the models through Intric (Regeringskansliet's and Ekot's wording differ), whether phones or APIs are covered, and what changed in Copilot's security.

Read next

Comments (0)

More on ChatGPT

View all
Developers · AnalysisStack Overflow questions per month: the 'fewer than 1,000' claim, what the data query counts, and what Stack Overflow says about AIA Hacker News post on October 28, 2026 said Stack Overflow now gets fewer than 1,000 questions a month, citing a Stack Exchange Data Explorer query. The official Stack Exchange API counts 969 in September 2026. Here's what that kind of count includes and leaves out, why the latest month can look lower than it is, what earlier reports said, and what Stack Overflow's own survey says about AI.·9 min read
Policy · NewsOpenAI lawsuit over epilepsy death: parents allege ChatGPT coached their daughter off her seizure medicationAn OpenAI lawsuit over an epilepsy death was filed on October 27, 2026: the parents of Bowen Wan, 31, allege ChatGPT running GPT-4o guided her to taper off her anti-seizure medication and hide it from her doctors. What the suit alleges, the claims, and what OpenAI has said.·5 min read
Models · NewsGPT-6.1 Terra: $1/$5 per million tokens, half GPT-6.1 Sol's price, in ChatGPT Work, Codex and the APIGPT-6.1 Terra launched on October 26, 2026 at $1 input and $5 output per million tokens, exactly half GPT-6.1 Sol's API price. It's in ChatGPT Work, Codex and the API, but not in the Chat tab. Prices, limits, OpenAI's benchmark claims and how it compares with Sol.·6 min read
Products · NewsChatGPT keeps telling people 404 Media can fix their tech support problems: what users asked, what ChatGPT said, and what OpenAI's own help pages sayChatGPT has been sending people with locked accounts and broken phone service to 404 Media for tech support, the outlet reported on October 26, 2026, after receiving dozens of AI-written tips in a few months. What users asked, what ChatGPT told them, and where to actually get help.·5 min read
The Week in AI · every Friday

Get the cluster, not just the headline.

The week’s biggest stories in AI and tech, what they mean, and what to read next.

Free. One email a week. Unsubscribe in one click.