The 2 August deadline changed: here is what still applies and what doesn’t
On 29 June 2026 the Council of the EU gave its final approval to the Digital Omnibus. The text was signed on 8 July, published in the Official Journal of the EU on 24 July as Regulation (EU) 2026/1744 and entered into force on 27 July 2026. If your company had spent months racing towards 2 August, this moves the deadline, but not entirely, and the nuance matters.
What has changed, in one sentence
The high-risk obligations in Annex III of the AI Act (risk management, technical documentation, human oversight, quality management system) move from 2 August 2026 to 2 December 2027: 16 more months. The obligations for embedded high-risk AI in Annex I (AI built into products that are already regulated, such as medical devices or machinery) move from 2 August 2027 to 2 August 2028.
What has NOT changed
This is where most headlines fall short. Article 50, the transparency obligations (labelling AI-generated content and telling people when they are interacting with an AI system rather than a person), keeps its original timetable. It applies from 2 August 2026, unchanged. Only Article 50(2), on systems already placed on the market, is delayed to December 2026. Full guide: what the Article 50 transparency rules actually require.
The Article 5 prohibitions (in force since February 2025) and the obligations for general-purpose AI models (in force since August 2025) have not changed either. Article 4 on AI literacy has applied since February 2025, but the Omnibus reworded it: from “ensuring a sufficient level” to “taking measures to support its development”, without requiring a specific level for each person. The Omnibus also adds two new prohibitions, non-consensual intimate images and AI-generated child sexual abuse material, applicable from 2 December 2026.
Why this matters more than it seems
A 16-month delay sounds like relief, and partly it is. But there are two real risks in misreading it:
- Confusing “delayed” with “cancelled”. High-risk obligations are still coming, just in December 2027 instead of August 2026. An AI governance programme started 16 months ahead has room to be done properly; one started 3 months ahead because “there’s no rush now” does not.
- Forgetting that August 2026 is still a real date. If your company deploys chatbots, content generators or any system that interacts with people who don’t know they are talking to an AI, Article 50 applies to you now: the delay does not touch this.
What to do with this information
- If your project plan had the “big finish” for high-risk systems in August 2026, you can reschedule it with real breathing room, but don’t remove it from the roadmap.
- Check right now whether any of your systems fall under Article 50 (transparency): that one still arrives on the original date.
- Use the extra 16 months to do it properly: a complete inventory, a correct risk classification and documentation built calmly rather than against the clock.
Related: why it pays to implement Article 14 human oversight now, even though its deadline has also moved, and what the declaration of conformity that sums up all this work is.
Source
Regulation (EU) 2026/1744 (Official Journal of the EU, 24 July 2026), final approval by the Council of the EU on 29 June 2026, and Council and Pinsent Masons communications on the approved text of the Digital Omnibus.
What about the deadlines you’re already facing?
A free assessment of your risk level, or the checklist with every AI Act regulatory deadline in a single document.