EU’s Digital Omnibus Tweaks AI Act: Deepfake Ban Now, High‑Risk Deadlines Later

At the end of June 2026, the Council of the European Union gave final approval to the so‑called “Digital Omnibus VII,” a regulation that amends parts of the EU AI Act. The package pushes back several looming deadlines for high‑risk AI systems while fast‑tracking new rules against the most harmful uses of generative AI, especially sexual deepfakes and AI‑generated child sexual abuse material.

Under the updated timeline, standalone high‑risk AI systems are now slated to face full obligations from December 2, 2027, with embedded high‑risk systems in products following on August 2, 2028. By contrast, AI practices that generate non‑consensual intimate imagery—such as tools that remove clothing in photos or fabricate explicit pictures of real people—will be banned across the EU from December 2026. The grace period for content‑labelling duties is also cut from six months to three, meaning providers must implement AI‑generated content disclosures by December 2, 2026.

Brussels is attempting a familiar balancing act: easing pressure on companies building complex, safety‑critical AI while moving aggressively against already visible harms. For global platforms and model providers, the message is clear. They have more time to adapt risk‑management processes for high‑risk sectors, but almost no time left to keep operating gray‑zone generative services that can be abused for sexualised deepfakes or synthetic child abuse material within the EU.