In a significant move, the European Parliament has voted to delay core enforcement deadlines of the EU's landmark Artificial Intelligence Act. The decision, passed with broad support, grants developers of high-risk AI systems—those considered a serious threat to health, safety, or fundamental rights—until the end of 2027 to comply. For AI integrated into products already governed by strict sectoral rules, such as medical devices, the proposed deadline extends to August 2028.

Additionally, new rules mandating that providers watermark AI-generated audio, video, and images will not take effect until November 2026. All these obligations were initially scheduled for August of this year.

The vote also endorsed incorporating a ban on so-called 'nudify' applications into the revised law. While specifics are absent, the proposal suggests an exemption for systems with robust technical safeguards that prevent the generation of non-consensual intimate imagery. This follows public and political anger across Europe earlier this year after a surge of sexually explicit deepfakes, attributed to the Grok AI model, circulated on the platform X.

The postponement prolongs a regulatory limbo for the tech industry, which has already contended with shifting timelines and incomplete guidance from Brussels. The proposed changes are not yet final; Parliament must now negotiate the final text with the European Council, comprising ministers from the 27 member states, raising questions about whether any agreement can be reached before the original August date.

Source: The Verge