EU negotiators agree AI Act simplification and nudifier ban

Parliament and Council negotiators reached a provisional agreement early Thursday on changes to the EU AI Act under the digital omnibus package, aiming to make compliance easier while preserving the law’s risk-based structure. The deal postpones parts of the rules so standards and support measures can be in place before providers must comply.

Obligations for high-risk AI systems with designated use cases, including biometrics, critical infrastructure, education, employment, law enforcement and border management, would apply from 2 December 2027. AI systems used as safety components and covered by EU sectoral safety and market surveillance legislation would face obligations from 2 August 2028. Watermarking duties for AI-generated content would apply from 2 December 2026.

The agreement would ban AI systems designed to create child sexual abuse material or non-consensual depictions of an identifiable person’s intimate parts or sexually explicit activity, covering images, video and audio. Companies would have until 2 December 2026 to align their systems with the ban.

Negotiators also agreed to remove overlapping requirements for machinery products, narrow the definition of safety component, allow strictly necessary personal data processing to detect and correct biases with safeguards, extend SME exemptions to small mid-cap enterprises and streamline enforcement for certain general-purpose AI systems through the EU’s AI Office. The provisional text still needs formal adoption by Parliament and Council before entering into law.