European Parliament Committees Approve Revised AI Act Ahead of Plenary Vote
European Parliament committees broadly support a revised AI Act that introduces new rules for foundation models and adjusts the definition of high-risk systems. The legislation now moves to a plenary vote in mid-June before entering final trilogue negotiations.
European Parliament committees adopt a provisional deal on the AI Act by an overwhelming majority, paving the way for a plenary vote scheduled for mid-June. This revised text diverges significantly from the original Commission proposal by broadening the definition of an AI system to align with OECD standards. Despite this wider definition, the overall scope of high-risk systems narrows because Annex III use cases now require proof of significant risk to qualify for strict regulation.
The updated legislation introduces new high-risk categories, explicitly bans certain AI practices, and establishes foundational ethical principles. Notably, the text imposes direct obligations on providers of foundation models, reflecting a major shift in how the EU approaches generative AI. The amendments also place a strong emphasis on AI-value chain governance, ensuring that responsibilities extend beyond just the initial developers to various actors involved in the deployment process.
Stakeholders show mixed reactions to the provisional deal, with some expressing concern that strict rules might hinder technological innovation. However, the clear progression of the Act allows companies to anticipate requirements and begin preparing for compliance. If Parliament approves the text in June, it serves as the negotiating basis for upcoming trilogue discussions under the Spanish presidency, with a final compromise text expected in early 2024.