EU member states back compromise text revising parts of AI Act implementation
EU governments have agreed on a compromise proposal that would adjust timelines, obligations, and enforcement mechanisms under the Artificial Intelligence Act while introducing new prohibited AI practices.
The Council of the European Union has approved a compromise text proposing targeted revisions to implementation of the Artificial Intelligence Act.
The agreement was confirmed by the Permanent Representatives Committee on 13 May following negotiations between EU institutions earlier in the month. The proposal forms part of the Digital Omnibus package, which seeks to simplify parts of the EU’s digital regulatory framework.
The changes focus mainly on implementation rather than rewriting the core structure of the AI Act. According to the text, delays in technical standards, conformity assessment systems, and national governance arrangements created heavier compliance burdens than initially anticipated.
One major element concerns timelines for high-risk AI systems. The proposal would postpone some obligations until late 2027 and 2028, depending on the category of high-risk system involved.
The compromise also revises several operational requirements, including AI literacy obligations, regulatory sandboxes, conformity assessment procedures, and registration requirements.
A notable change affects AI literacy rules. Instead of requiring providers and deployers to ensure a sufficient level of AI literacy among personnel, the revised wording would require them only to take measures supporting AI literacy. The adjustment reduces the legal burden linked to proving individual competency levels.
The text also introduces a new prohibited AI practice. It would ban AI systems used to generate realistic non-consensual intimate content or child sexual abuse material involving identifiable individuals, subject to limited legal exceptions.
Another important aspect is institutional coordination. The proposal clarifies certain responsibilities of the EU AI Office and attempts to reduce overlaps between the AI Act and sector-specific legislation, including rules affecting aviation and product safety.
The compromise reflects growing concern within EU institutions about implementation capacity. Rather than weakening the AI Act’s overall approach, the revisions are aimed at slowing and restructuring parts of enforcement so that technical standards, supervisory systems, and compliance infrastructure can catch up with the regulation’s timeline.
