MEPs advance AI Act with two accountability routes

European Parliament committees backed an AI Act negotiating mandate on 11 May, voting 84–7 with 12 abstentions. It isn’t law: the committee text needs plenary endorsement, expected in June, before Council negotiations.

Photo: European flags. © Håkan Dahlström, Spain, 2010 (CC BY 2.0, via Wikimedia Commons). Resized and cropped. Historical EU illustration, not the announcement.

Foundation-model providers would assess and mitigate risks and register in an EU database. Generative foundation-model providers would publish summaries of copyrighted training data.

Two routes to accountability

MEPs also want people to receive explanations of decisions based on high-risk systems that significantly affect their rights, alongside a right to complain about AI systems.

These address different questions: provider documentation about a model, and explanations of consequential decisions. A training-data summary wouldn’t, by itself, explain that decision.