Sweden's Government Offices Block ChatGPT and Claude, Keep Microsoft Copilot

Sweden's Government Offices sit inside the state machinery that handles official work, so a security review there sets a standard for what civil servants can use on government devices. The decision lands at a moment when public bodies are sorting through how generative AI fits with handling sensitive information and internal workflows. In this case, the approved tool is Microsoft Copilot, while ChatGPT and Claude are excluded from staff computers.

Civil servants who relied on general-purpose chatbots will have to move to the approved system if they want AI assistance on work devices. That narrows the range of models available inside a government environment and gives Microsoft Copilot a privileged position in official use. For vendors, public-sector approval now looks like a gate that can decide which products become normal in daily administration.

The split between blocking two tools and keeping one approved shows the trade-off between access and control in government AI use. Public agencies want the productivity gains of generative AI, but they also have to limit exposure of official data and workflows to systems that do not meet their requirements. Sweden is joining a pattern where governments test AI against security rules before letting it into routine state work.

Still unknown is what specific security tests ChatGPT and Claude failed, and whether either tool could return after further review. It is also unclear whether other Swedish agencies will adopt the same rule or build their own lists of approved AI systems. The next concrete signal would be follow-up guidance from Sweden's Government Offices on which services, if any, remain available for staff use.

The move was larger before the first headline than after. That is an observation, not an explanation: the market also moves on things nobody publishes.