The European Parliament has voted to simplify parts of the Artificial Intelligence Act and push back key deadlines for compliance. The European Parliament writes in an official press release the vote passed by 569 votes in favour, 45 against, and 23 abstentions.
The most significant changes affect when companies must comply with rules governing high-risk AI systems. For systems listed directly in the regulation, including those used in biometrics, law enforcement, education, and border management, the new deadline is 2 December 2027. For AI systems covered by existing EU product safety laws, the deadline extends to 2 August 2028. Providers of AI-generated audio, image, video, or text content must add watermarks indicating the content’s origin by 2 November 2026.
Parliament also voted to ban so-called nudifier systems. These are AI tools that generate or manipulate images to appear sexually explicit while resembling a real, identifiable person without their consent. Systems with effective safety measures that prevent such outputs would be exempt from the ban.
Additional changes include allowing AI providers to use personal data to detect and correct bias, subject to strict safeguards. Companies that have grown beyond small and medium-sized enterprise status will gain access to support measures previously reserved for smaller businesses. Products already regulated under sector-specific EU laws, such as medical devices or toy safety rules, will face lighter AI Act obligations.
Negotiations with the EU Council on the final text can now begin.
