EU member states approve final wording of AI Act

Members of the competent Council of Ministers' Permanent Representatives Committee (Coreper) voted to approve the AI Act at a meeting on Friday morning, according to the Belgian presidency of the Council.

In a statement on X, the Belgian presidency of the Council said Coreper had “confirmed the final compromise text found on the proposal on harmonised rules on artificial intelligence”. It described the AI Act as “a milestone, marking the first rules for AI in the EU, aiming to make it safe and in respect of EU fundamental rights”.

In December, the then Spanish presidency of the Council and the European Parliament announced that they had reached “a provisional agreement” on the proposed AI Act. The next stage in the legislative process was for the wording of a compromise text to be set and that text put before Coreper for its endorsement.

With Coreper’s approval now confirmed, the compromise text will be considered by MEPs who sit on the two committees within the Parliament that are jointly leading on scrutiny of the EU AI Act proposals on behalf of the Parliament – the committees on Internal Market and Consumer Protection and on Civil Liberties, Justice and Home Affairs.

Hear Nils Rauer discuss the AI Act on The Pinsent Masons podcast.

Assuming the two committees are happy with the legislative language, the proposals will then be put to a vote of all MEPs in a plenary session in the Parliament, for their adoption, likely in April, before the Council then formally votes itself on whether to adopt the text.

Under the EU AI Act, a new risk-based system of regulation applicable to AI will apply across EU member states.

Under the framework, some uses of AI will be prohibited entirely, while the strictest regulatory requirements are reserved for ‘high-risk’ AI systems and the providers and deployers of such systems. The technology would, for example, need to meet mandatory requirements around issues such as risk management, data quality, transparency, human oversight and accuracy, while the providers and deployers would face a range of duties, including around registration, quality management, monitoring, record-keeping, and incident reporting.

Regulatory requirements will also apply to general purpose AI models, under the AI Act, including the need to disseminate detailed summaries about the content used to train those AI models and – in the case of general purpose AI models that pose “systemic risk” – obligations around model evaluation and adversarial testing.