Sweden's government blocks ChatGPT and Claude Monday

Sweden's civil servants lose ChatGPT and Claude at their desks on Monday. Copilot stays.

Sweden's Government Offices, the body that supports the prime minister and the cabinet, will block both chatbots on staff computers from Oct. 26 and direct employees to Microsoft Copilot as the approved AI tool, Decrypt reported on Oct. 24, citing Sveriges Radio.

The reason is a missing contract

Mikael Wettercrantz, the Government Offices' director of digitalization, wrote on the staff intranet that using AI services without a formal contract can have "consequences for our information security," according to Decrypt.

He also wrote: "We have seen several examples of problematic use of open AI services."

Copilot passed because the Government Offices reviewed it and raised its security settings afterward, Decrypt reports. ChatGPT and Claude, used as consumer tools, did not get that review.

Some staff will still reach ChatGPT and Claude through an internal AI platform the Government Offices built. The press office has not said how many people that covers.

Sweden's Government Offices is not the first Swedish agency to do this. Decrypt reports the Tax Agency and the National Board of Institutional Care have similar bans, the Enforcement Authority is preparing its own decision, and an internal audit at the Transport Administration found roughly 10 AI services in use in a "widespread and risky way."

The model was never the problem

Here's the thing. Claude may not even be gone.

Microsoft added Claude models to Copilot earlier this year, Decrypt notes, and in the EU they are off by default until an administrator turns them on. The Government Offices has not said whether its Copilot has Claude enabled.

So the line Sweden drew is not between good models and bad ones. It is between a tool covered by a security agreement and a tool someone signed up for with a work email.

That is the line most large organizations will draw. The European Parliament disabled AI features on its work devices in February, ChatGPT, Claude and Copilot included, saying it could not guarantee the security of the data, according to Decrypt.

What it means if you sell, or use, AI at work

If you sell AI to public bodies or regulated companies, lead with the contract. Data handling, retention, where it is processed, who can compel access. Decrypt notes US law lets authorities compel US companies to hand over data, and that worry is exactly what this block answers. Terms like zero data retention are what get a tool onto the approved list.

If you use a chatbot at work on a personal account, assume this is coming to you. Ask your IT team which tools are approved and on what terms, before they ask you.

And if you build on top of these models, your customer's procurement team will ask which sub-processors touch their data. Have that answer written down.

Watch the other Swedish agencies

The Enforcement Authority's decision is next. Watch whether other EU governments copy the approved-tool model, and whether OpenAI and Anthropic answer with government contracts in Sweden.

Approved beats better. Get approved.