EU postpones high‑risk AI obligations and bans deepfake nudifier apps

The European Parliament voted on 16 June to approve a package that delays the most demanding provisions of the EU AI Act. Obligations for stand‑alone high‑risk systems will now apply from 2 December 2027 and for AI embedded in regulated products from 2 August 2028. The watermarking rule requiring machine‑readable labeling of AI‑generated content moves to 2 December 2026, while a new ban on AI that creates child sexual abuse material or non‑consensual intimate imagery (often called “nudifier” apps) takes effect by the same date. The text still awaits formal adoption by the Council of the EU and publication before it becomes law, with the Council expected to act before 2 August 2026.

Critics say the deal weakens key safeguards: it reduces the information AI providers must upload to the public database, shifts AI in machinery to sector‑specific rules, and delays public‑interest controls, thereby setting a precedent for deregulation of digital rights laws. The final agreement also postpones high‑risk AI risk‑management and human‑oversight duties, allowing potentially harmful systems to remain on the market longer.

In parallel, the Digital Trust Summit in Brussels on 24 June 2026 will bring together policymakers, business leaders and technology experts to discuss how organisations can meet the upcoming transparency obligations that start on 2 August 2026 and turn EU AI principles into practical governance frameworks.