Stockholm has tightened its grip on artificial intelligence in the public sector. As of Monday, government employees in Sweden are barred from using browser-based versions of OpenAI's ChatGPT and Anthropic's Claude on their work computers. The move follows a security review that flagged risks tied to how these tools handle sensitive information.
The decision stems from an internal assessment at the Swedish Transport Administration earlier this year, which uncovered what officials described as “problematic” uses of AI platforms in government work. The core worry: entering confidential data into these models without a formal agreement on data protection could result in that information being absorbed into the models and used for training.
This practice, often called shadow IT, poses a direct threat to national information security, according to Mikael Wettercrantz, director of digitalisation at the Government Offices of Sweden. “Consequences for our information security” are a real risk, he said, underscoring the need for stricter oversight.
Copilot as the approved alternative
In place of the banned tools, the Swedish government is steering its workforce toward Microsoft's Copilot. That platform has passed a security screening and is covered by a data-processing agreement between Microsoft and the Swedish state. Officials argue it offers a safer way to harness AI without compromising sensitive data.
Government employees will still be able to use AI for their tasks, but under tighter controls. Only a limited number of staff will be allowed to continue using ChatGPT and Claude, and only through a dedicated platform for AI assistants developed by Swedish authorities. This ensures that any use of these models is monitored and aligned with security protocols.
The restrictions extend beyond the central government. According to Sveriges Radio, Sweden's public broadcaster, the Swedish Tax Agency and the National Board of Institutional Care have also imposed limits on ChatGPT and Claude. The Swedish Enforcement Authority is currently reviewing its own position.
These AI tools now join DeepSeek on Sweden's list of unapproved AI services. Last week, the United States announced a ban on all products and services from the Chinese AI company, citing similar data security concerns. Sweden's move reflects a broader European trend of scrutinizing AI tools that lack robust data protection guarantees.
For European observers, this is a reminder that the continent is navigating a delicate balance between technological innovation and data sovereignty. As AI becomes more embedded in public administration, governments across the EU are watching closely. The Swedish approach—combining a ban on risky tools with a vetted alternative—could serve as a model for other member states.
Meanwhile, the debate over AI regulation continues to evolve. The European Union's AI Act, which aims to set a global standard for AI governance, is expected to add another layer of oversight. For now, Sweden's decision highlights the practical challenges of integrating AI into government while safeguarding sensitive information.
