Sweden's Government Offices Will Block ChatGPT and Claude on Work Computers From Monday, Pointing Staff to Copilot Instead

Sweden's Government Offices, the Regeringskansliet that supports the prime minister and the cabinet, will stop employees from directly accessing ChatGPT and Claude on government computers beginning Monday, October 26, according to reporting by Swedish Radio's Ekot. Staff will instead be directed toward Microsoft Copilot, the only major assistant the organization has approved after a security assessment.

The move is an access-control decision, not a nationwide ban, and it does not claim either service caused an incident. Mikael Wettercrantz, the Government Offices' digitalization director, warned in a message on the internal network that using AI services without established agreements could create information security risks. He declined an interview with Ekot, and the press office refused to describe what security modifications were made to Copilot before its approval, beyond saying the service had been hardened following review.

There is a carve-out: some employees will continue reaching ChatGPT and Claude through an internally developed AI assistant platform, letting the organization keep using the underlying models through a system it controls. The arrangement captures the distinction Swedish authorities are drawing, between permitting AI use in principle and controlling which channels staff may use to get it.

The Government Offices is not acting alone. The Swedish Tax Agency and the National Board of Institutional Care already prohibit comparable AI services on their systems, and the Swedish Enforcement Authority is still deciding whether to follow. An audit earlier this year at the Swedish Transport Administration found employees were using around ten publicly available AI services, with documented problems involving sensitive information submitted to AI systems and the handling of AI-generated answers, though no security breach was identified.

Sweden's public sector has otherwise been moving in two directions at once. A February 2026 national strategy commits the country to broader AI adoption, the Swedish Post and Telecom Authority has run supervised trials of both Copilot and ChatGPT under strict information-handling rules, and Kammarkollegiet evaluated the two services before introducing digital assistants for its staff. A separate Justice Ministry inquiry, appointed in September 2026, is examining how the Swedish tradition of public access to official documents applies when authorities use AI systems and external cloud providers, with findings due by November 2027.

For Microsoft, the decision is a quiet enterprise win: a national administration steering thousands of civil servants toward Copilot not because it benchmarked best, but because it signed the agreements and passed the review. The episode also illustrates how governments are handling consumer-grade AI in the workplace, by treating model quality as almost irrelevant next to procurement, data handling and auditability. Expect more administrations to make the same trade, and expect employees to keep finding the loophole, which in this case is a government-built platform that quietly keeps the banned models available.