The European Council released its general approach on a package of EU AI Act amendments on March 13, 2026. Two elements define it. The first is the one your compliance team has likely already tracked: a proposed extension pushing enforcement of high-risk AI system obligations from August 2026 to approximately December 2027, up to 16 months from the original schedule. The European Commission proposed that delay in November 2025. The second element is new.
The Council’s package adds a provision prohibiting AI tools that generate non-consensual sexual or intimate imagery. The Record reports the Council’s position explicitly targets AI practices used to produce such content. The amendment reportedly also addresses child sexual abuse material, per coverage of the Council package, though that specific provision was not confirmed in the full text available at publication. Treat the CSAM element as reported, not confirmed, until the official Council press release is located.
Neither element is enacted law. The Council’s position must now be reconciled with the European Parliament before the amendments become binding, a process that follows the March 18 committee vote. Compliance teams should not treat December 2027 as a confirmed operative date.
One critical nuance: the deadline extension applies to specific high-risk system provisions. Some August 2026 obligations may remain in effect. Reporting confirms the Council framed this as a targeted extension, not a blanket delay. Do not assume all August 2026 requirements have moved.
For generative AI providers whose models can produce intimate imagery, the nudification ban is the element requiring product assessment now, regardless of where the broader deadline extension lands after Parliament.