Sweden's government offices block ChatGPT and Claude on work PCs

Sweden's Government Offices, which support the prime minister and the ministries, are blocking the browser versions of ChatGPT and Claude on staff work computers from Monday 26 October 2026. Staff are directed to Microsoft Copilot instead. According to Sveriges Radio's Ekot, the decision was first reported on 23 October 2026, and the government's press office confirmed it to SVT. The stated reason is information security: AI services used without a formal agreement can create risks. For businesses that automate work, the practical lesson is that the approval depends on the contract and the settings, not only on the AI model. Companies should know which AI tools their staff open on work machines and which of those have a signed data agreement.
In short
- Sweden's Government Offices block the browser versions of ChatGPT and Claude on staff work computers from 26 October 2026.
- Staff are pointed to Microsoft Copilot, which the offices reviewed and whose security settings they tightened.
- The stated reason is information security risk from AI services used without a formal agreement.
- Some staff keep access to ChatGPT and Claude models through the offices' internal Intric platform.
- For any business that automates work, the contract behind each AI tool matters as much as the tool itself.
What happened
Sveriges Radio's news programme Ekot first reported the decision on 23 October 2026, citing internal information. According to Ekot, from Monday 26 October 2026 the Government Offices will no longer let staff reach ChatGPT and Claude from their work computers. The Government Offices cover the prime minister's office and the ministries, so the change touches the civil service that supports the Swedish cabinet.
The press office later confirmed the block by email to SVT. In that email it said the browser interface of both chatbots will no longer be reachable from staff work computers from 26 October. It also said all staff have access to Microsoft's AI tool, Copilot, and that some AI services are switched off from Monday.
The reason given comes from Mikael Wettercrantz, the Government Offices' director of digitalisation. According to Sveriges Radio, he wrote on the staff intranet that using AI services without a formal contract could have consequences for information security. The offices also said they had reviewed Copilot and then tightened its security settings. The press office declined to say what those changes were, citing security reasons.
The block is narrower than the headlines suggest. According to SVT, staff can still use ChatGPT and Claude models through Intric, a platform the offices use internally. Decrypt also reported that the restriction applies to the browser route, and that the offices have not said how many people keep the internal access.
The current position at a glance
| Tool or route | Status for Government Offices staff from 26 October 2026 |
|---|---|
| ChatGPT in the browser | Blocked on staff work computers |
| Claude in the browser | Blocked on staff work computers |
| Microsoft Copilot | Approved after review, with security settings tightened |
| ChatGPT and Claude models via Intric | Still available to some staff, according to the press office |
Why it matters for businesses that automate work
Automation projects often connect AI to customer records, invoices and internal reports. The Swedish case draws a line between a tool that staff open for convenience and a tool that sits under a signed agreement. The government said the problem was missing agreements, not the models themselves. Decrypt reported that consumer AI tools send user queries to external servers, and that U.S. law can require those companies to hand over data on request.
Microsoft's own documentation for Microsoft 365 Copilot says that prompts, responses and data accessed through Microsoft Graph are not used to train the foundation models behind Copilot. That protection applies to organisations using Microsoft 365 commercial terms. It does not apply to a personal chatbot account opened in a browser. The difference is contractual, so a business has to check its own agreement rather than assume it.
Sweden is not the only public body moving this way. Ekot found similar blocks at the Swedish Tax Agency (Skatteverket) and the National Board of Institutional Care (Statens institutionsstyrelse, Sis). It also reported that the Swedish Enforcement Authority (Kronofogden) is preparing its own position.
What changes in practice, at any size
A five-person firm and a group with thousands of staff face the same first step: find out which AI tools people actually use on work machines. Staff often use personal accounts in a browser tab, and those tabs rarely appear in a software list. A short survey of staff, followed by a look at browser history on company devices, is usually enough to start.
Next, decide which tool may hold which kind of data. A reasonable split is to allow an approved business tool for internal drafts, keep customer identity data and payroll out of any tool without a signed agreement, and record the rule in writing. Smaller firms can do this in one page. Larger groups will need an administrator to set the rule on every device.
Automation can help here, but it cannot decide what is sensitive. A system can route tasks to an approved tool, log what was sent and flag unusual uploads. A person still has to classify the data, approve the tool and review the output before it reaches a client, a tax authority or a bank.
What to watch next
Three things are worth tracking. The first is whether Kronofogden adopts a similar block. The second is whether the Government Offices explains what it changed in Copilot's settings. The press office has so far declined to give details. The third is whether the Intric route stays open, and for how many staff, since the offices have not given a number.
Sources
How AiStaffo would automate this
AiStaffo would start by listing every AI tool and data route in use, including the browser accounts that staff open on work machines. Each routine task, such as drafting follow-ups or summarising reports, would be connected to the tool your business has agreed to use, and the system would log what goes into it. Routine runs would happen automatically, but the owner would still decide which tool may hold which data and would approve any new tool before it is connected. Results depend on your own figures and contracts, which the free audit measures. Book a free automation audit.
Questions people ask
Did Sweden ban ChatGPT and Claude?
Why did the Swedish government block the chatbots?
Why is Microsoft Copilot allowed when ChatGPT and Claude are not?
Does this mean my business should block ChatGPT and Claude?
Which other Swedish bodies have made similar decisions?
Book a free automation audit
Thirty minutes. We look at one process you run every week and tell you exactly what an AI worker would take off your desk, and what it would not.


































