Aller au contenu

Regulation & compliance

AI Act: what was postponed, and what has applied since 2 August

The Digital Omnibus delayed high-risk obligations to December 2027, but Article 50 transparency has applied since 2 August 2026.

  • 11 OCTOBER 2026
  • 2 MIN READ
  • CITED SOURCES

The essentials

  • Regulation (EU) 2026/1744, in force since 27 July 2026, postpones high-risk AI system obligations to 2 December 2027 (Annex III) and 2 August 2028 (Annex I).
  • It does not postpone Article 50: since 2 August 2026, deployers must inform users of chatbots, label generated content, and disclose deepfakes.
  • Next milestone: 2 December 2026, with machine-readable marking for generative systems already on the market and two new prohibitions.

What the Omnibus actually delayed

Regulation (EU) 2026/1744, the “Digital Omnibus on AI”, published in the Official Journal of the European Union on 24 July 2026 and in force from 27 July, is the first amendment to the AI Act. It pushes back substantive requirements for high-risk systems: to 2 December 2027 for standalone systems in Annex III (recruitment, education, credit scoring, biometrics, essential services…) and to 2 August 2028 for those embedded in regulated products (Annex I).

It also introduces reliefs for SMEs and small mid-cap companies: simplified technical documentation, proportionate quality requirements, and reduced fine caps.

What already applies

The most common confusion is hearing “the AI Act is postponed”. That is inaccurate. Still in force: prohibited practices and the requirement for AI literacy among staff (since 2 February 2025), obligations for general-purpose AI models (since 2 August 2025), and, since 2 August 2026, the transparency obligations of Article 50:

  • inform people that they are interacting with an AI system, for example a chatbot;
  • machine-readably mark synthetic content: text, image, audio, video;
  • disclose deepfakes;
  • inform people exposed to emotion recognition or biometric categorisation.

One calendar flexibility exists on this point: machine-readable marking for generative systems placed on the market before 2 August 2026 is required only from 2 December 2026. The EU penalty regime is active: depending on severity, up to €35 million or 7% of worldwide turnover, or €15 million or 3% in other cases.

For an organisation that uses AI without building it

If you deploy an assistant on your site, draft content with generative tools, or integrate AI into your processes, you are very likely a “deployer” under the text. First steps are straightforward:

  1. 01Inventory AI uses, internal and customer-facing.
  2. 02Display a clear notice on assistants and generated content.
  3. 03Train teams: the AI literacy obligation was reframed as a duty of means, but it remains binding for every deployer.
  4. 04Document where data goes: a model run locally greatly simplifies the answer.

Sources

This note was drafted with AI tools from the cited sources, then reviewed and published under the responsibility of Jordan FOUASSIER, publication director. We summarise the facts and add our reading; source text and images are not reproduced. An error? Write to contact@aigyrosgroup.com.

Prendre contact

Does this affect your environment?

Describe your context. We reply with an analysis of what it means for you, not a catalogue.