Starting today, the people who run Sweden's government can't open ChatGPT or Claude on their work computers. The block, which took effect Monday at the Government Offices — the Regeringskansliet, the administration behind the prime minister and every ministry — leaves Microsoft Copilot as the approved AI assistant, according to CEOWORLD Magazine, which picked up the decision first reported by Sveriges Radio's Ekot.
The striking thing is what Sweden didn't say. Nobody in Stockholm called the models unsafe. The problem, as digitalisation director Mikael Wettercrantz put it on the staff intranet, is that employees were using AI services "without a formal agreement," which "can have consequences for our information security." The offices had also seen "several examples of problematic use" — sensitive material typed into consumer chatbots, and AI-generated answers handled as if they were authoritative.
That framing is the whole story. ChatGPT and Claude aren't banned in Sweden's government; ungoverned access is. Some staff will keep reaching both models through Intric, an internal platform the ministries control, where the data handling is governed by contract. The browser tab is what's gone — the route where a civil servant pastes a document into a free consumer tool and the institution has no say over where it goes next.
Sweden's agencies are arriving at the same answer independently. The Tax Agency and the National Board of Institutional Care have already introduced similar blocks, the Enforcement Authority is considering one, and an audit at the Transport Administration reportedly found about ten AI services in "widespread and risky" use. DeepSeek was already restricted.
The pattern extends beyond Sweden. In February, the European Parliament disabled built-in AI features on work devices used by lawmakers and staff, saying it could not guarantee the security of data sent to cloud services.
For OpenAI and Anthropic, the lesson is uncomfortable: neither company has commented publicly, and neither was rejected on capability. They lost the room on paperwork — no enterprise agreement, no approved configuration, no data-processing terms a government could sign. Microsoft won it by already being inside the building, with an agreement in place and a configuration the offices could review and tighten.
That is the direction of travel for enterprise AI. The question buyers now ask isn't only which model is smartest. It's which vendor will sign the contract that lets the model through the door.
— via CEOWORLD Magazine