Sweden bans browser use of ChatGPT and Claude for government staff over data security

On 26 October, Sweden implemented a ban that prevents government employees from using browser-based versions of OpenAI’s ChatGPT and Anthropic’s Claude on official devices, citing fears that sensitive data could be captured and used to train the models. The move stems from an internal audit at the Swedish Transport Administration that identified “problematic” applications of these services in public work. Mikael Wettercrantz, director of digitalisation at the Government Offices of Sweden, said that without a formal data-protection agreement, the platforms could jeopardise information security.

As a replacement, Microsoft’s Copilot, which has passed a security screening and is covered by a government agreement, is being recommended. Only a small group of staff will retain access to ChatGPT and Claude through a specially developed AI-assistant platform, and similar restrictions have been applied at the Swedish Tax Agency, the National Board of Institutional Care, and are under review by the Swedish Enforcement Authority. The ban adds DeepSeek to the list of AI tools barred by the Swedish government for security reasons.