AI Act Omnibus Moves Forward with Delayed High-Risk Obligations

On Thursday, the European Parliament adopted its position on the digital omnibus proposal amending the Artificial Intelligence Act, with strong cross-party support. The amendments focus on simplifying implementation and delaying the application of certain obligations for high-risk AI systems, allowing time for guidance and harmonized standards to be finalized. Parliament stressed the need for predictability and legal certainty through fixed application dates.

For high-risk AI systems explicitly listed in the AI Act, including those used in biometrics, critical infrastructure, education, employment, essential services, law enforcement, justice, and border management, MEPs propose an application date of 2 December 2027. High-risk AI systems already covered by EU sectoral safety and market surveillance legislation would instead become subject to the AI Act on 2 August 2028. In addition, providers would have until 2 November 2026 to comply with watermarking obligations for AI-generated audio, images, video, and text.

Parliament also proposes a new ban on AI-powered “nudifier” applications that generate or manipulate sexually explicit or intimate images resembling identifiable individuals without consent. Systems equipped with effective safeguards preventing such misuse would fall outside the scope of the ban. Further amendments would allow providers to process personal data to detect and mitigate bias in AI systems, subject to strict necessity and safeguards.

To support European innovation, MEPs backed extending certain SME support measures to small mid-cap enterprises as they scale. They also called for reduced overlap between the AI Act and existing sectoral product safety regimes, allowing lighter AI Act obligations for products already regulated under EU safety legislation. Interinstitutional negotiations with the Council will now determine the final shape of the revised AI Act.