Is the AI Act delayed? What applies in 2026.

Last updated: July 2026

Abstract illustration of layered tiers

Updated 8 July 2026. We will revise this analysis when the Digital Omnibus is published in the Official Journal — or when the deadline passes.

Yes and no — and the nuance is exactly what most headlines leave out. The Digital Omnibus defers part of the AI Act, but today it is a provisional political agreement: it only takes effect if it is published in the EU Official Journal before 2 August 2026. And there are obligations that this date triggers no matter what.

The short answer

  • Article 50 transparency —disclosing when people interact with an AI and marking AI-generated content— applies on 2 August 2026 no matter what. The Omnibus does not touch this date.
  • High-risk would be deferred (Annex III → 2 December 2027; Annex I → 2 August 2028) only if the Omnibus reaches the Official Journal before 2 August 2026.
  • Everything already in force stays in force: prohibitions and AI literacy (February 2025), GPAI rules and penalties (August 2025).

What applies on 2 August 2026, no matter what

The Article 50 transparency obligations:

  1. AI interaction: if a chatbot or conversational system serves people, they must be told they are talking to an AI (unless it is obvious from context).
  2. Synthetic content: AI-generated or AI-manipulated content (text published to inform, images, audio, video, deepfakes) must be marked as such, in a machine-readable format where technically feasible. Generative AI systems already on the market before 2 August 2026 get a grace period until 2 December 2026.
  3. National sandbox: every Member State must have a regulatory sandbox operating. In Spain it is run by AESIA, with priority, free access for SMEs.

None of this depends on the Omnibus.

What would be deferred — and under which condition

The Omnibus pushes back the high-risk obligations:

Obligation Original calendar With the Omnibus published in time
Annex III high-risk (biometrics, employment, education, essential services…) 2 Aug 2026 2 Dec 2027
Annex I high-risk (AI embedded in regulated products) 2 Aug 2027 2 Aug 2028

The condition is literal: publication in the EU Official Journal before 2 August 2026. If the formal process does not make it in time, the deferral does not exist and the original calendar stands. That is not a footnote: it is the difference between planning with 17 months of margin or with zero.

What I would do this week

  1. Comply with transparency now — labelling AI-generated content and disclosing chatbots requires waiting for no one; it is cheap and it is mandatory in August.
  2. Inventory and classification up to date — list every AI tool in use (including shadow AI) and classify each use case against Annex III, documenting the reasoning.
  3. Don't bet your planning on the deferral — if a system is anywhere near high-risk, work with the 2026 date as the base scenario; the Omnibus, if it lands, becomes extra margin.
  4. Watch the Official Journal — one single check: has the Omnibus been published? That answer reorders the entire high-risk calendar.

Primary sources


Need the full picture —roles, obligations, penalties and the technical checklist—? It is in the guide: The AI Act for your SME: the technical side. And we do the technical side of compliance —inventory, classification, documentation and audit— from Zaragoza, for companies across Spain and Europe.

This is not legal advice: it is the engineering that complements your legal counsel.

Frequently asked questions.

  1. Q01
    Is the EU AI Act delayed?
    Only part of it, and conditionally. The Digital Omnibus would defer the high-risk obligations (Annex III to 2 December 2027 and Annex I to 2 August 2028), but it is a provisional political agreement: it only binds if published in the EU Official Journal before 2 August 2026. Article 50 transparency applies on 2 August 2026 no matter what.
  2. Q02
    What applies on 2 August 2026 no matter what?
    The Article 50 transparency obligations (disclosing when people interact with an AI and marking AI-generated or AI-manipulated content) and each Member State's duty to have a national regulatory sandbox up and running. Everything already in force stays in force: prohibitions and AI literacy (since February 2025), GPAI rules and penalties (since August 2025).
  3. Q03
    What is the Digital Omnibus?
    An EU digital-simplification legislative package that, among other things, adjusts the AI Act timeline. As of this analysis it is a provisional political agreement: until it is published in the EU Official Journal it has no legal effect.
  4. Q04
    When will the high-risk obligations apply?
    It depends on a condition with a deadline. If the Omnibus is published in the Official Journal before 2 August 2026, Annex III high-risk moves to 2 December 2027 and Annex I to 2 August 2028. If it is not published in time, the original calendar stands: Annex III on 2 August 2026 and Annex I on 2 August 2027.
  5. Q05
    Is content watermarking deferred too?
    No. Marking synthetic content is part of Article 50 and applies on 2 August 2026. Generative AI systems already placed on the market before that date get a grace period until 2 December 2026 to implement it.
  6. Q06
    What happens if the Omnibus is not published in time?
    Then there is no deferral: Annex III high-risk (CV screening, scoring, biometrics, essential services…) would be enforceable from 2 August 2026. That is why the technical recommendation is to have your inventory and classification ready and not to bet your planning on the deferral.