Sweden has restricted the use of OpenAI’s ChatGPT and Anthropic’s Claude in government work due to concerns regarding information security and the risk of sensitive data leaks.
ChatGPT and Claude join the list of unapproved AI tools, alongside DeepSeek, which has been restricted previously.
A review at the Swedish Transport Agency uncovered problematic instances of AI usage in public service operations.
Mikael Wettercrantz, Chief Information and Digital Transformation Officer, warned that uncontrolled use of AI platforms could have consequences for national information security.
In the absence of enterprise contracts and appropriate governance mechanisms, data entered into chatbots risks being transmitted to developers, stored, or used for AI training.
Data processing at non-European data centers also poses compliance risks regarding GDPR data protection regulations.
Microsoft Copilot continues to be permitted for use, subject to stricter security control measures.
Certain employees still retain access to ChatGPT and Claude via specialized AI assistant platforms.
📌 Conclusion: Sweden is shifting from permissive AI usage to governance based on approved tool categories. The decision to restrict ChatGPT, Claude, and DeepSeek does not mean a complete ban on AI technology in the public sector.Microsoft Copilot continues to be used under strict control mechanisms. The policy focus is on protecting sensitive data,controlling information processing operations, and ensuring auditability and GDPR compliance.
