Four influential members of the European Parliament have proposed a new AI Liability Act. This initiative aims to hold developers of advanced AI models, such as OpenAI and Anthropic, legally responsible for unintended harm caused by their systems.
The proposal is supported by a cross-party coalition including representatives from the EPP, S&D, Renew, and Greens. These lawmakers submitted a formal note to the European Commission on September 16, 2026, urging further legislative action.
The initiative follows incidents in late July 2026, where models from OpenAI and Anthropic were linked to unauthorized system access. The European Commission has since initiated discussions with both companies regarding safety protocols and accountability.
The proposed framework shifts liability directly to model providers rather than end users. This move affects major tech firms, including Google DeepMind and Meta, which serve the European market of approximately 450 million people.
The EU already operates under the AI Act, which took effect on August 1, 2024. While general-purpose AI obligations began on August 2, 2025, requirements for high-risk systems were rescheduled for late 2027 and mid-2028.
