The European Commission tabled the Digital Omnibus on AI on 19 November 2025, aiming to simplify and delay parts of the AI Act's high-risk regime before it began to bite. The European Parliament voted to adopt the provisional agreement reached with the Council on 16 June 2026, and the Council gave its final, definitive approval on 29 June 2026. The result is now part of Regulation (EU) 2024/1689 itself: the deadlines that firms advising on AI compliance had been working toward all year just moved.
What actually moved
Two deadlines changed, and they are not the same deadline. Standalone high-risk systems under Annex III, the category covering most AI used in recruitment, credit scoring, insurance pricing and similar decisions about people, now have until 2 December 2027 to comply, instead of 2 August 2026. AI embedded in products already regulated under harmonised EU product-safety law, such as the Machinery Regulation (EU) 2023/1230, now has until 2 August 2028, instead of 2 August 2027. A firm that built a client's compliance roadmap around the original August 2026 date needs to redraw it, and needs to redraw it differently depending on which of the two categories the client's system falls into.
What did not move
The Omnibus is a postponement, not a general amnesty. The AI Act's prohibited practices have applied since February 2025 and are untouched. Obligations on providers of general-purpose AI models have applied since August 2025 and are untouched. The duty to label AI-generated content and disclose synthetic media, which several sources tie to Article 50, is delayed only for systems already on the market before 2 August 2026, and only until 2 December 2026, a shorter grace period than the six months the Commission had originally proposed. The new ban on AI-generated child sexual abuse material and non-consensual sexualised deepfakes takes effect on that same date, 2 December 2026. National regulatory sandboxes, meant to let companies test high-risk systems under supervision, are due by 2 August 2027.
Why this matters for a firm advising clients in Spain or Italy
Spain and Italy were each building their own national layer on top of the EU Regulation, and neither layer moved with the Omnibus; in Spain's case, it has since stopped altogether. The Council of Ministers approved the Proyecto de Ley Orgánica for the good use and governance of AI on 26 May 2026 and sent it to Congress, where the amendment period was extended six times, to 7 October 2026. It never got that far: Royal Decree 806/2026 dissolved the Cortes on 6 October 2026 and called a general election for 29 November, and the bill lapsed with the legislature. Whichever government emerges will have to table it again from the start, so Spain has, for now, no national implementing law on any timetable, and the Regulation's own postponed dates are the only ones a firm can plan around there. In Italy, Legislative Decree 179/2026, in force since 23 October 2026, keeps its own deadlines: bar associations and other professional bodies must build AI literacy into mandatory training by 23 April 2027, and the fee parameters tying a lawyer's use of AI to the equo compenso must be set by 23 October 2027 (see our coverage of that decree). A client operating in both countries is therefore tracking two live clocks, the EU's new high-risk dates and Italy's own decree deadlines, plus a stopped one in Spain that will restart only when a new Congress takes the bill up again. None of them resets the others.
What to check now
Any gap analysis or compliance roadmap a firm delivered to a client before July 2026 was almost certainly built on the original 2 August 2026 date and needs a written update, not a quiet assumption that "it moved." The two postponed dates reward precise classification: a firm should confirm, system by system, whether a client's AI falls under Annex III as a standalone high-risk use or under Annex I as a component embedded in an already-regulated product, since the 2027 and 2028 dates apply to different categories. The European Commission's guidelines on how to classify high-risk systems were still in public consultation through 23 June 2026 and remain pending in final form; a firm advising on classification should watch for their publication rather than relying on the draft. Finally, the obligations that did not move, the prohibited practices, the general-purpose AI duties, and the shortened grace period for transparency labelling, which ends on 2 December 2026, deserve their own line in the client memo, so that "the AI Act was delayed" does not get repeated as a blanket statement it was never meant to be.
At Lexpower, tracking which obligation moved, which did not, and which runs on a separate national clock is exactly the kind of granular, document-anchored work a specialist AI can prepare and a licensed professional in the relevant jurisdiction then reviews, verifies and signs. See how our specialists work.
Sources
- El Consejo de la UE aprueba el Ómnibus Digital de IA — Cuatrecasas
- EU AI Act Omnibus Agreement — Postponed High-Risk Deadlines and Other Key Changes — Gibson Dunn
- EU Lawmakers Reach Provisional Agreement to Delay Key EU AI Act Obligations — Sidley
- Commission seeks feedback on the draft guidelines for the classification of high-risk artificial intelligence systems — European Commission
- Real Decreto 806/2026, de 5 de octubre, de disolución del Congreso de los Diputados y del Senado y de convocatoria de elecciones — BOE
- Proyecto de Ley Orgánica para el buen uso y la gobernanza de la inteligencia artificial, expediente 121/000096 (caducado) — Congreso de los Diputados