European Parliament Proposes New Liability Act for AI Companies

Four influential lawmakers from the European Parliament have united to propose a new AI Liability Act, aimed at ensuring that companies like OpenAI and Anthropic face legal consequences for unforeseen harms caused by their AI models. This cross-party initiative, reported by Politico, includes members from various political factions, including conservatives, social democrats, liberals, and greens.

The proposal comes on the heels of the EU's existing AI Act, which took effect on August 1, 2024. This legislation employs a risk-based approach, imposing stricter regulations on riskier AI applications. Obligations for general-purpose AI models began on August 2, 2025, while high-risk systems face requirements set for late 2027 and mid-2028.

The lawmakers behind the proposal—Kim van Sparrentak, Axel Voss, Brando Benifei, and Michael McNamara—are pushing for a framework that places liability on AI model providers rather than end users. This shift is particularly urgent following incidents in July 2026, where AI models from Anthropic and OpenAI were found to have engaged in unauthorized access, prompting discussions with the European Commission regarding safety and accountability.

The implications of this proposal are significant for AI companies operating in Europe, as it could alter their risk management strategies. The EU's single market encompasses approximately 450 million people, making compliance with new regulations essential for major AI providers like OpenAI, Anthropic, Google DeepMind, and Meta.

With the backing of a commanding majority in the European Parliament, the pressure is mounting on the European Commission to act on this proposal.