Sweden Blocks ChatGPT and Claude on Government Work Computers
From Monday 26 October 2026, staff in Sweden's Government Offices can no longer open the ChatGPT or Claude websites from their work computers. The block, first reported by Sveriges Radio's Ekot on 23 October and confirmed by the Government Offices to broadcaster SVT by email, is one of the clearest examples yet of a government deciding that how staff reach an AI model matters as much as which model they use.
The body in question is the Regeringskansliet, usually translated as the Government Offices. It is not a single ministry but the administrative machine that supports the prime minister and the cabinet, bundling the Prime Minister's Office with the ministries under one roof. When it changes a rule about work computers, it changes it for the people drafting legislation, handling policy and writing official documents.
The block is easy to misread, and much of the early coverage did. It is not a ban on ChatGPT or Claude in Sweden. No Swedish law changed. Nothing stops a Swedish citizen from using either product on a personal phone. It is a workplace restriction, on managed devices, aimed at one specific route into those models: the public web interfaces.
Credit: Rosenbad, Swedish Government Offices, via Arto
What is blocked, and what is not
From 26 October 2026, the browser interfaces for ChatGPT and Claude are no longer reachable from the Government Offices' work computers. That is the confirmed scope, in the offices' own email to SVT: "The browser interface for the AI services ChatGPT and Claude will no longer be reachable from the Government Offices' work computers from October 26."
Microsoft Copilot, by contrast, remains approved for all staff. According to Ekot and SVT, the Government Offices reviewed Copilot and raised its security settings after that review before allowing it. That distinction is the whole point of the exercise. Copilot is not approved because it is a different model in some abstract sense. It is approved because the Government Offices have a reviewed service, with security settings they control, for official work.
The models are not fully gone, either. Staff still have access to the ChatGPT and Claude language models through Intric, an internal AI assistant platform. Ekot reported that part of the staff will keep that access, and Swedish AI company Intric won the contract to supply the platform in the summer of 2026. The block therefore targets the public door, not the model behind it.
Why the Government Offices did it
The stated reason is information security, and specifically contracts. Mikael Wettercrantz, director of digitalisation at the Government Offices, wrote on the offices' intranet that using AI services without an agreement can have consequences for information security. He declined an interview with Ekot, and the press office declined to give details beyond that intranet line, citing security.
According to reports, the Government Offices had seen examples of problematic use, including concerns about the sensitivity of information entered into AI tools and how AI-generated responses were handled. That is a practical concern, not an abstract one. Government staff handle policy drafts, official documents and sensitive information every day. If that material is pasted into a public chatbot under consumer terms, the government cannot control where it goes, how long it is stored, or whether it is used for training.
A contract changes that. A business or government agreement can set how long data is stored, where it is processed, who can access it and whether it can be used for training. A consumer account, by contrast, leaves those questions to the provider's standard terms and to a privacy setting the user may or may not have changed. The Swedish decision is, at its core, a decision to require a contract before official work goes into an AI system.
Credit: ChatGPT logo icon, via Vecteezy
Why Copilot is different, and why that matters
Copilot's approval shows that the Government Offices are not opposed to AI at work. They are opposed to uncontrolled AI at work. After reviewing Microsoft's service and raising its security settings, they allowed it for all staff. That is a procurement decision, not a technology judgement about which chatbot is smartest.
There is a wrinkle worth noting honestly. Reporting has noted that the Government Offices have not said whether Claude is switched on inside their Copilot setup, and Copilot can, in some configurations, run Claude models. That does not make the block meaningless, because the route, the contract and the security settings are still different. But it does show how hard it is to draw a clean line around a model when models are increasingly embedded inside other products.
Other Swedish administrations are moving in the same direction. Reporting noted that the Tax Agency and the National Board of Institutional Care already have similar blocks, and the Enforcement Authority was deciding. The Government Offices are therefore not an outlier inside Sweden. They are the most prominent example of a wider pattern.
What this means for ordinary users
The Swedish block is about government work computers, but its reasoning applies to private users too. With a regular consumer account, prompts are processed on the provider's servers under standard terms. A business contract can set storage, processing location and training use. On consumer ChatGPT and Claude accounts, a privacy setting decides the training question, and many users never change it.
For anyone who handles sensitive work, the practical lesson is simple: check whether you are using a consumer account or a contracted service, and check what that service does with your data. The model name on the screen is less important than the contract behind it. Sweden's Government Offices blocked two famous names and kept a less famous one, precisely because the contract, not the brand, was the deciding factor.
For other governments, Sweden's move is a template that is easier to copy than a law. An internal IT decision can be made quickly, without legislation, and it can be reversed just as quickly if a contract is signed. Expect more governments to use this tool, especially as AI tools become standard office equipment rather than novelties.
What is not proven
No public statement says a leak happened, and none of the sources says an incident prompted the block. The stated logic is the absence of a contract, not a documented breach. That matters because it would be wrong to read the block as proof that ChatGPT or Claude leaked Swedish government data. The Government Offices have not said that, and the reporting does not support it.
It would also be wrong to read the block as a judgement that Copilot is inherently safer than ChatGPT or Claude in all settings. Copilot was approved after a review and with raised security settings, in a specific government deployment. A different deployment, with different settings, would be a different decision.
Credit: Microsoft Copilot logo, via Ars Technica
Key Takeaways
- From Monday 26 October 2026, Sweden's Government Offices block the ChatGPT and Claude websites on staff work computers, as first reported by Ekot on 23 October and confirmed to SVT by email.
- Microsoft Copilot remains approved for all staff, after the Government Offices reviewed it and raised its security settings.
- The block covers public web interfaces on managed devices only; some staff still reach ChatGPT and Claude models through the internal Intric platform.
- The stated reason is information security and the absence of a formal agreement, not a documented leak or a new law.
- Digitalisation director Mikael Wettercrantz warned on the intranet that using AI services without an agreement can affect information security.
- The decision is a workplace IT rule, not a ban on either product in Sweden, and no Swedish citizen is stopped from using them on a personal device.
Sweden's block is a small decision with a large lesson. Governments are starting to treat AI like any other office tool: useful, but only through a reviewed service with a contract behind it. The model on the screen matters less than the door you use to reach it, and Sweden has just closed two very public doors.
Sources
- Gobi Daily — ChatGPT and Claude Blocked on Sweden's Government PCs From October 26
- Times Now News — Why Sweden Is Blocking ChatGPT And Claude On Govt Computers While Allowing Copilot
- ExplainX — Sweden Blocks ChatGPT and Claude From Oct 26 (Ekot)
- ZFR Build — Swedish government offices to ban ChatGPT and Claude



Comments
Post a Comment