Four senior MEPs from rival political blocs are pushing to make OpenAI and Anthropic legally accountable when their models cause unintended harm.
Four of the European Parliament’s most influential tech lawmakers have joined forces to demand a new AI Liability Act, a proposal that would force companies like OpenAI and Anthropic to face legal consequences when their most powerful models cause harm in ways nobody anticipated.
The cross-party push, first reported by Politico based on written exchanges between lawmakers, represents a rare alignment of the Parliament’s major political factions. Conservative, social democrat, liberal, and green MEPs all signed the September 16, 2026 note to the European Commission.
What the proposal would change #
The EU already has an AI Act, which took effect on August 1, 2024, and was celebrated as the world’s first comprehensive AI regulation. It uses a risk-based framework: the riskier the AI application, the stricter the rules. Obligations for general-purpose AI models kicked in on August 2, 2025, while high-risk system requirements were pushed back to late 2027 and mid-2028 through a July 2026 omnibus revision.
The four MEPs behind the push are Kim van Sparrentak from the Greens, Axel Voss from the center-right EPP, Brando Benifei from the social democrats, and Michael McNamara from the liberal Renew group. Voss and Benifei were co-rapporteurs on the original AI Act. The AI Office, which received full enforcement powers in August 2026, would likely play a central role in implementing any new liability framework.
The incidents driving urgency #
In late July 2026, AI models from both Anthropic and OpenAI were found engaging in unauthorized system access. The European Commission subsequently opened discussions with both companies about safeguards and responsibilities.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
The UK is taking similar notice. Its Parliament has summoned several AI companies for a hearing scheduled for October 13, 2026, focused specifically on safety and liability.
The proposed liability act would place responsibility squarely on the providers of the most advanced models, rather than on downstream deployers or end users who may have no insight into how the underlying system works.
Market and industry implications #
For AI companies operating in or selling to the European market, this proposal introduces a new variable into their risk calculus. The EU’s single market covers roughly 450 million people, making it impossible for major AI providers to simply ignore Brussels. OpenAI, Anthropic, Google DeepMind, and Meta’s AI division all serve European customers and would fall within the scope of any new liability regime. With the EPP, S&D, Renew, and Greens aligned, the coalition represents a commanding majority in the European Parliament. The European Commission, which holds the exclusive right to propose EU legislation, will face intense pressure to act.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our