The “Digital Omnibus” has pushed high-risk obligations back to 2027, but most of the regulation already applies, and December 2 is fast approaching.
“The AI Act has been postponed”: the phrase is accurate for only one part of the text. Regulation (EU) 2026/1744 entered into force on 27 July 2026.
What was postponed
- Stand-alone Annex III systems (recruitment, credit scoring, education): from 2 August 2026 to 2 December 2027.
- Systems embedded in regulated products (Annex I): from 2 August 2027 to 2 August 2028.
The dates have moved, but the scope has not: a system classified as high-risk before the omnibus remains high-risk.
What already applies
- Since 2 February 2025: prohibited practices and the AI literacy obligation.
- Since 2 August 2025: obligations for general-purpose AI models.
- Since 2 August 2026: the transparency rules of Article 50.
No published analysis reports any change to the penalty ceilings. They remain at €35 million or 7% of worldwide annual turnover for prohibited practices, and €15 million or 3% for other infringements.
December 2026: less than a month away
- Generative systems already on the market must carry machine-readable marking (Article 50(2)).
- Two new prohibitions apply: non-consensual intimate imagery and AI-generated child sexual abuse material.
Article 4 on AI literacy has also been revised: it was not delayed, but softened. Internal policies that cite the old wording should be updated.
Why waiting until 2027 is a miscalculation
Compliance means inventorying systems, classifying them, documenting data, and setting up human oversight. None of these steps depends on the application date. Those who start in 2026 will arrive in 2027 with an up-to-date register.
At Berexia, compliance is built on data
The AI Act is as much a data and governance issue as a legal one. This is where AI, data, cloud, and compliance intersect, and where Berexia supports its clients across France, Spain, the UK, Morocco, and Singapore. The delay buys time to do things properly, provided you start now.
