Sweden's Government Offices will block ChatGPT and Claude on staff work computers starting Monday, October 26, 2026, directing civil servants to Microsoft's Copilot instead. Sveriges Radio's news program Ekot reported the decision on Friday, October 23, citing internal documents and a post on the staff intranet by Mikael Wettercrantz, the Government Offices' director of digitalisation, who warned that using AI services without a contract could have "consequences for our information security." The Government Offices, known as the Regeringskansliet, runs the prime minister's office and every ministry. The block covers the computers staff use for work, and it is not a ban on AI itself. Copilot is the approved service, and its security setup was tightened after an internal review, though the press office would not detail the changes, saying the matter concerns security. Some employees will keep reaching ChatGPT and Claude through an AI-assistant platform the Government Offices built itself. Use of China's DeepSeek on official devices was already restricted.
What the block actually covers
The restriction is narrower than a headline ban and more serious than a policy memo. It is a technical block on staff service computers, aimed at the consumer web versions of ChatGPT and Claude that employees were opening with no agreement in place. The line being drawn is contractual: Copilot cleared a security review, the other two did not. Wettercrantz declined an interview with Ekot, but the intranet post makes the reasoning plain. Uncontracted AI services mean government text flowing to servers under terms the government never negotiated.
Two details show this was planned, not reactive. First, the Government Offices upgraded Copilot's security controls before the block takes effect, so staff have a reviewed alternative on day one. Second, the in-house AI-assistant platform gives selected employees a sanctioned route to the very models being blocked, which means the objection is to the channel, not the capability. Neither OpenAI nor Anthropic has commented publicly.
The government AI ban tracker
Sweden is not the first government to draw this line. We compiled every government-level restriction on general AI assistants we could verify, and the pattern is consistent: the bans target uncontracted consumer tools, and each one ends with an approved alternative.
| Government | Date | What was restricted | Status |
|---|---|---|---|
| Italy (Garante, data protection authority) | March 31, 2023 | ChatGPT blocked nationwide over GDPR concerns | Lifted April 28, 2023, after OpenAI added opt-outs and age checks |
| United States (DHS) | October 2026 | DeepSeek removed from federal civilian IT systems (order 2026-21447) | In force |
| Sweden (Tax Agency, National Board of Institutional Care) | 2026 | Unapproved AI services blocked, per Ekot reporting | In force; Enforcement Authority considering similar steps |
| Sweden (Government Offices) | October 26, 2026 | ChatGPT and Claude blocked on staff work computers; Copilot approved | Takes effect Monday |
Gra
phic: verified government-level restrictions on general AI assistants, March 2023 to October 2026. Sources: Garante, DHS order 2026-21447, Sveriges Radio Ekot. Chart: ByteHerald.
The approval model, and why it matters beyond Sweden
What makes the Swedish move worth attention outside Scandinavia is the completeness of the loop. Most organizations issue a policy memo and hope. The Regeringskansliet implemented a network-level block and paired it with a sanctioned alternative that cleared a security review. That is the full procurement loop most enterprises still have half of: ban the unsanctioned tools, then actually provide one that passed review, so employees have somewhere legitimate to go.
The contrast with the American approach is instructive. The DHS order removing DeepSeek from federal civilian systems targets a foreign model on supply-chain grounds, while Sweden's block applies to the two most widely adopted Western assistants on data-residency grounds. The criteria differ, but the destination is the same: government work consolidates onto whichever vendor can sign the security paperwork. Right now, that is Microsoft, whose Copilot benefits from exactly this kind of review everywhere it runs. We mapped the laws driving these decisions in our worldwide AI regulation comparison; Sweden's move is what enforcement looks like when the law is already on the books.
What to watch next
Whether this stays a Swedish story comes down to what happens next. If the Enforcement Authority follows the Tax Agency and the institutional-care board, the block becomes the default across Swedish public administration. The in-house platform is the other test: how many employees get sanctioned access to ChatGPT and Claude, and under what logging, will show whether this is a genuine security architecture or a two-tier workaround. And then the copycats: every government IT department watching this now has a template, which is a block plus an approved alternative, not a memo. Expect the tracker above to grow.
By ByteHerald Staff. Sweden's Government Offices blocking ChatGPT and Claude on staff work computers from October 26, 2026, reported by Sveriges Radio's Ekot on October 23, 2026 (Mikael Wettercrantz intranet warning on information security; Copilot approved after a security review; in-house AI-assistant platform retains access for some staff; DeepSeek already restricted). Swedish Tax Agency and National Board of Institutional Care blocks, and the Enforcement Authority's pending decision, via Ekot and SVT reporting on October 23, 2026. Italy's Garante blocking ChatGPT on March 31, 2023 and reinstating it on April 28, 2023 after OpenAI added privacy measures, via the regulator's announcements. The US DHS order 2026-21447 removing DeepSeek from federal civilian IT systems in October 2026, via ByteHerald's own reporting.
Comments
Post a Comment