EU finalises AI liability rules after late compromise
Negotiators agreed a scaled-back liability framework for general-purpose AI systems, dropping several provisions opposed by member states.
Key points
- EU negotiators agreed a scaled-back AI liability framework
- A compulsory insurance requirement was dropped
- Tiered disclosure for high-risk applications remains
- Formal vote is scheduled in the coming months
Why it mattersThe compromise sets the shape of AI accountability in Europe's largest market and will influence how providers design disclosure features.
European Union negotiators finalised a liability framework for artificial intelligence systems on Friday after a late compromise narrowed its scope.
The Verge reported that the agreed text drops a requirement for providers of general-purpose AI to carry compulsory insurance, a provision several member states had opposed. A tiered disclosure obligation for high-risk applications remains.
Officials said the framework would let people harmed by an AI system seek redress without having to prove exactly how the model reached its decision. The burden of disclosure would fall on the provider.
The compromise was criticised by some consumer groups, who argued the insurance removal weakened the rules. Industry groups welcomed the lighter touch, saying it preserved room for European developers to compete.
The text now goes to a formal vote in the coming months. It would apply 20 months after entry into force.
Reported by
- The Verge — EU agrees AI liability rules after insurance dropped · 2026-08-29T06:40:00Z
- Reuters — EU reaches compromise on AI liability framework · 2026-08-29T07:00:00Z