What happened
On 15–16 June 2026, the European Parliament plenary session voted to give final approval to amendments to the EU AI Act as part of the 'digital omnibus' simplification package. Key changes: (1) High-risk AI systems embedded as safety components covered by EU sectoral safety/market-surveillance legislation now apply from 2 August 2028 (deferred from 2 August 2026); (2) Watermarking obligations on AI-generated content are delayed until 2 December 2026; (3) A new explicit EU-wide ban on 'nudifier' tools (AI apps that generate synthetic nude images of real people without consent) and on AI-assisted creation of child sexual abuse material is added to the prohibited-practices list; (4) Overlapping requirements on AI for machinery products are removed. The Parliament press release was published 16 June 2026 (reference 20260611IPR45207).
Why it matters
This is a formal legislative amendment to the EU AI Act — not guidance — and changes compliance deadlines that thousands of companies had been working toward. The two-year deferral of high-risk obligations for safety-component AI gives industry more runway but also delays consumer protection. The nudifier ban is the first explicit prohibition of a specific AI application category added to the Act since its original adoption, establishing a precedent for content-specific bans. The watermarking delay to December 2026 reduces near-term burden for generative AI deployers in the EU but compresses the final timeline.
Action needed
Update EU AI Act compliance roadmaps: high-risk safety-component AI deadline is now 2 August 2028. AI-generated content watermarking must be in place by 2 December 2026. Immediately audit image-generation products for nudifier functionality — the ban is a prohibited practice with no grace period upon entry into force. Await Official Journal publication and confirm exact entry-into-force date.