EU Backs Provisional AI Act Simplification With Deepfake Sex Ban and Later Deadlines
The proposal pairs a targeted deepfake ban with extra time for companies to comply.
Overview
- EU Parliament negotiators and the Council presidency reached a provisional deal to streamline parts of the AI Act that still needs formal sign-off.
- The text forbids selling in the EU any AI system that creates non-consensual sexual or intimate images, video, or audio, responding to a surge in “nudification” apps.
- Providers must label AI-generated content by December 2, 2026, with companies given until the same date to adapt their systems.
- Obligations for high-risk uses shift to August 2, 2027, extending timelines for tools used in hiring, facial recognition, and critical infrastructure.
- Sectors already covered by EU safety rules and many small firms would gain broader exemptions, easing duplicate compliance burdens.