EU simplifies its AI Act
- EU governments and European Parliament negotiators reached a provisional agreement on May 7 to simplify parts of the AI Act and delay key high-risk rules. - The deal pushes stand-alone high-risk AI obligations to Dec. 2, 2027, while product-embedded systems would shift to Aug. 2, 2028. - The European Commission’s draft high-risk guidance, published May 19, is open for stakeholder consultation until June 23, 2026.
The European Union has moved from writing the AI Act to rewriting parts of how it will work in practice. On May 7, the Council presidency and European Parliament negotiators reached a provisional agreement to streamline parts of the law under the bloc’s wider “Omnibus VII” simplification package. The changes keep the AI Act’s risk-based structure but delay key high-risk obligations and narrow some compliance burdens, according to the Council of the EU. The European Commission added a second step on May 19, publishing draft guidelines on how to decide whether an AI system counts as “high-risk” under Article 6 of the law. Those guidelines are now open for stakeholder feedback and are meant to help companies, deployers and national authorities apply the rules more consistently across the bloc. (consilium.europa.eu) ### What did EU lawmakers actually change? The May 7 agreement delays the application of high-risk AI rules that had been due to start on Aug. 2, 2026. Under the provisional deal, stand-alone high-risk AI systems would instead face the new rules from Dec. 2, 2027, while high-risk AI embedded in products would move to Aug. 2, 2028. (digital-strategy.ec.europa.eu) The Council of the EU said the package also extends some regulatory exemptions previously aimed at small and medium-sized enterprises to small mid-caps, reduces requirements in a limited number of cases, broadens the possibility of processing sensitive personal data for bias detection and mitigation, and seeks to reduce governance fragmentation. The co-legislators also added a new prohibition on AI practices involving non-consensual sexual and intimate content or child sexual abuse material. (consilium.europa.eu) ### Why is Brussels changing a law that only took effect in 2024? The AI Act entered into force on Aug. 1, 2024, as Regulation (EU) 2024/1689, but the Commission, member states and lawmakers have spent the following months working through how the law would be enforced in sectors with overlapping product, safety and digital rules. The Council said the simplification package is intended to reduce recurring administrative costs, improve legal certainty and smooth implementation across the union. (consilium.europa.eu) Marilena Raouna, Cyprus’s deputy minister for European affairs, said in the Council statement that the agreement would support companies by cutting recurring administrative costs while keeping protections in place. Euronews reported that supporters cast the changes as a way to cut red tape, while critics argued they favored large technology companies. (consilium.europa.eu) ### What counts as “high-risk” AI under the draft guidance? The Commission’s draft guidance says an AI system is high-risk in two main cases. One is when the system is itself a regulated product, or a safety component of one, under the EU legislation listed in Annex I and subject to third-party conformity assessment. The other is when it falls within one of the use cases listed in Annex III of the AI Act. (consilium.europa.eu) The 19 May draft says it is meant to support providers, deployers and market surveillance authorities in assessing classification and to provide practical examples of systems that should or should not be treated as high-risk. The Commission split the material into sections covering general principles, Annex I products and Annex III use cases. (digital-strategy.ec.europa.eu) ### What happens next for companies and regulators? The Commission’s stakeholder consultation on the draft high-risk guidelines runs until June 23, 2026, according to Commission-linked materials and legal summaries published after the release. Feedback is meant to inform the final guidance on classification and related obligations. (digital-strategy.ec.europa.eu) The next formal step on the simplification package is adoption by the EU institutions after the provisional May 7 deal. If that text is approved, the revised timetable would leave companies preparing for Dec. 2, 2027, for stand-alone high-risk systems and Aug. 2, 2028, for product-embedded systems, while the Commission finalizes the guidance on Article 6 classification. (consilium.europa.eu) (hunton.com)