Regulation & Ethics
The Digital Omnibus on AI: What Changed in the AI Act and What Did Not
The Digital Omnibus AI Act amendment, Regulation (EU) 2026/1744, deferred high-risk duties to December 2027 but left transparency, literacy and the prohibitions in force.
On this page
Regulation (EU) 2026/1744, the Digital Omnibus on AI, has been in force since 27 July 2026. It deferred the AI Act's high-risk obligations, Annex III to 2 December 2027 and Annex I to 2 August 2028, reworded the Article 4 literacy duty and added two prohibitions. It did not delay the Article 50 transparency rules, which applied on 2 August 2026, or anything already in force. "Delayed" is the wrong word for most of the Act.
Eight months from proposal to Official Journal
The speed of the file is itself worth explaining to clients who assume EU amendments take years.
- 19 November 2025. The Commission publishes the Digital Omnibus package, including the proposal to defer the high-risk deadlines that would otherwise fall on 2 August 2026. DLA Piper's tracker followed it from the first draft.
- 18 March 2026. Parliament committees adopt their position.
- 6 and 7 May 2026. Political agreement between Parliament and Council in trilogue. Pinsent Masons' Nils Rauer called the deal "a pragmatic step towards greater legal certainty" (Out-Law).
- 13 May 2026. Council confirms the agreement.
- 24 July 2026. Publication in the Official Journal as Regulation (EU) 2026/1744.
- 27 July 2026. Entry into force, the third day after publication, as White & Case's alert notes.
- 2 August 2026. Six days later, the AI Act's next scheduled tranche applies anyway: Article 50 transparency, AI Office enforcement powers over GPAI models, and national supervision of Article 4.
The compressed timetable had one purpose: to get the deferral into law before 2 August 2026, the date the high-risk obligations would otherwise have started. It achieved that with six days to spare.
What changed
| Provision | Before the Omnibus | After the Omnibus |
|---|---|---|
| Annex III stand-alone high-risk systems (employment, education, essential services, law enforcement, migration, administration of justice, biometrics, critical infrastructure) | Obligations from 2 Aug 2026 | Obligations from 2 Dec 2027 |
| Annex I high-risk AI embedded in regulated products | Obligations from 2 Aug 2027 | Obligations from 2 Aug 2028 |
| Article 4 AI literacy | "take measures to ensure, to their best extent, a sufficient level of AI literacy" | "take measures to support the development of AI literacy"; new paragraphs on Commission and Member State support and AI Board recommendations |
| Article 5 prohibitions | Original list | Two additions: AI generating non-consensual intimate imagery and child sexual abuse material, with a transition to 2 Dec 2026 |
| Article 50 watermarking of synthetic content | From 2 Aug 2026 | From 2 Aug 2026, with a grace period to 2 Dec 2026 for providers of existing generative systems |
| Requirements for certain categories of business | Standard | Eased, according to White & Case's summary; the details depend on category and are outside the scope of this page |
Gibson Dunn's analysis is the most complete public account of the agreed text; the Commission's own AI Act policy page now shows the new dates in its application timeline.
What did not change
This is the list clients need and rarely get.
- The prohibitions. Article 5 has applied since 2 February 2025. Emotion recognition in the workplace, social scoring and the rest were not touched, except to add two more.
- AI literacy. Article 4 applies from the same date. Softer words, same deadline, and now with national supervision. The Article 4 page shows the old and new text side by side.
- GPAI obligations. Model providers have been subject to Chapter V since 2 August 2025. The Commission's AI Office gained its enforcement powers on 2 August 2026 as scheduled: documentation requests, evaluations, corrective measures, withdrawal from the market and fines up to 15 million euro or 3 percent of turnover. The GPAI glossary entry explains what a general-purpose model is.
- Transparency. Chatbots must identify themselves as AI and deepfakes must be labelled since 2 August 2026. The only concession is the watermarking grace period for existing systems.
- The risk categories and roles. The Omnibus moved dates. It did not remove any Annex III category, redefine the deployer, or lower a fine tier. A system that was going to be high-risk still will be.
- Fine tiers. 35 million euro or 7 percent for prohibited practices, 15 million euro or 3 percent for most other obligations, 7.5 million euro or 1 percent for incorrect information.
The Cloud Security Alliance's research note titled its analysis "Deferred, Not Cancelled", which is the sentence to put at the top of any client memo.
Two new prohibitions
The Omnibus is usually described as deregulatory. In one respect it tightened the Act. Two practices were added to Article 5: AI systems that generate non-consensual intimate imagery, and AI systems that generate child sexual abuse material. Both carry the top fine tier and have a transition period to 2 December 2026. For law firms they are irrelevant in practice; for firms advising platforms, image generators, hosting providers and app stores they are a live compliance item with a deadline this year. Criticism came from the other direction as well: the European Disability Forum argued the package rolls back fundamental-rights safeguards, a reminder that "simplification" is contested and that further amendment is possible.
What clients keep getting wrong
| What the client says | What is true | What to tell them |
|---|---|---|
| "The AI Act is delayed" | Only the high-risk chapter is deferred | Three of four chapters are already in force; name them |
| "We have until 2028" | 2 Aug 2028 applies only to Annex I product-embedded AI | Annex III, which covers HR tools and most service uses, is 2 Dec 2027 |
| "Our chatbot can wait" | Article 50 has applied since 2 Aug 2026 | Label it now |
| "Nobody is enforcing it yet" | AI Office powers over GPAI and national supervision of Article 4 started 2 Aug 2026 | Enforcement is early, not absent |
| "Literacy training got scrapped" | Article 4 was reworded, not repealed | Keep the record; the AI literacy guide has a syllabus |
| "High-risk was cancelled" | Deferred by sixteen months for Annex III | Use the extra time; conformity assessment is a project, not a form |
| "The Omnibus only relaxed things" | Two prohibitions were added | Platform clients have a 2 Dec 2026 deadline |
The HR point deserves a second look. Recruitment screening, task allocation, performance monitoring and termination decisions supported by AI are Annex III high-risk uses. Every client with more than a handful of employees and an HR software vendor that has added "AI features" has a December 2027 project; the employment page covers what that project involves.
The client memo
A memo on the Omnibus that survives a general counsel's reading has a fixed shape. The skeleton below is written for a corporate client; adapt the third section to the client's systems.
MEMORANDUM: Digital Omnibus on AI (Regulation (EU) 2026/1744) — what changes for [Client]
1. Bottom line
The Omnibus, in force since 27 July 2026, defers the AI Act's high-risk obligations
(Annex III to 2 December 2027; Annex I to 2 August 2028). It does not defer the
prohibitions, the AI-literacy duty or the transparency rules, all of which already
apply to [Client]. Two prohibitions were added, effective 2 December 2026.
2. What moved and what did not (table)
[Before / after table, six rows, as above]
3. Application to [Client]'s systems
3.1 Systems in scope of Article 50 today: [customer chatbot; marketing image generation]
3.2 Article 4 status: [training delivered on …; record held by …]
3.3 Candidate Annex III systems: [recruitment screening tool; workforce scheduling]
-> compliance plan against 2 December 2027, owner, budget
3.4 Provider-side exposure: [none / product X placed on market under own name]
3.5 New Article 5 prohibitions: [not applicable / applicable to platform Y]
4. Recommended actions, dated
- Label chatbot (now)
- Update AI-literacy record (Q4 2026)
- Annex III gap analysis (Q1 2027), conformity plan (Q2 2027)
- Monitor Commission guidelines and delegated acts (ongoing)
5. Caveats
Secondary legislation and guidelines are still being issued; national penalties for
Article 4 vary; further amendment of the Act is politically possible.
Firms that produce this memo for several clients keep the first two sections as a template and regenerate the third from the client's tool inventory; the memo method that survives partner review applies. Module 6 of the AI Lab for Lawyers course covers risks and regulation for practitioners, the ground this kind of memo stands on.
Where this leaves you
Treat the Omnibus as a calendar change and update three dates in every AI Act document you hold: 2 December 2027 for Annex III, 2 August 2028 for Annex I, 2 December 2026 for the new prohibitions and the watermarking grace period. Then check the things that did not move, because those are where a client is exposed today: an unlabelled chatbot, a missing training record, a vendor that never signed the GPAI Code. Set up regulatory monitoring for the guidelines and delegated acts still to come; the hub page will carry the next set of dates when they land.
Frequently asked questions
What is the Digital Omnibus on AI?
It is Regulation (EU) 2026/1744, an amending regulation proposed by the Commission on 19 November 2025, politically agreed on 6 and 7 May 2026, confirmed by the Council on 13 May, published in the Official Journal on 24 July and in force since 27 July 2026. It defers the AI Act's high-risk obligations, softens the wording of Article 4, adds two prohibitions and eases some requirements for certain categories of business.
Is the EU AI Act delayed?
Partly. The high-risk chapter is deferred: Annex III stand-alone systems to 2 December 2027 and Annex I product-embedded systems to 2 August 2028. Everything else runs to the original timetable. The prohibitions and AI-literacy duty have applied since 2 February 2025, GPAI obligations since 2 August 2025, and transparency duties since 2 August 2026.
When do high-risk AI obligations now apply?
Stand-alone high-risk systems listed in Annex III, which include employment, education, essential services, law enforcement, migration and the administration of justice, must comply from 2 December 2027. High-risk AI embedded in products regulated under Annex I, such as medical devices and machinery, must comply from 2 August 2028. Both dates were confirmed by the Commission on its AI Act policy page.
Did the Omnibus change Article 4 AI literacy?
It changed the wording, not the date. The duty to ensure a sufficient level of AI literacy became a duty to take measures to support its development, and new paragraphs were added on Commission and Member State support and AI Board recommendations. The obligation has applied since 2 February 2025 and national authorities have supervised it since 2 August 2026.
Put this into practice
AI Lab for Lawyers: a live, hands-on course for practising lawyers
4 live sessions (8 hours), recordings included, taught by Dr. Niklas Schmidt of Wolf Theiss. Rated 4.7/5 from 381 ratings; alumni from Norton Rose Fulbright, Paul Hastings, Pillsbury and Taylor Wessing.
Related reading
Regulation & Ethics
The EU AI Act for Lawyers and Law Firms: What Applies in 2026
The EU AI Act applies to law firms as deployers. Article 4 literacy and Article 50 transparency are live; high-risk duties are deferred to December 2027.
9 min read
Regulation & Ethics
EU AI Act Article 4: The AI Literacy Duty for Law Firms, Explained
Article 4 of the AI Act requires every firm whose staff use AI to support their AI literacy. In force since 2 February 2025, supervised nationally since August 2026.
7 min read
Questions
Does the EU AI Act Apply to Law Firms? A Direct Answer
Yes. An EU law firm whose staff use ChatGPT, Claude or Copilot is a deployer under the EU AI Act. Article 4 literacy applies now, Article 50 since August 2026.
6 min read
Use Cases
AI Regulatory Monitoring: Agents That Track Law Changes for You
AI regulatory monitoring works as a scheduled fetch, diff, classify and route workflow on official sources. How to build a weekly digest and cut false alarms.
10 min read
Practice Areas
AI for Employment Lawyers: Advice, Investigations and the AI Act's HR Rules
AI for employment lawyers: advising on AI in hiring and monitoring after the AI Act's Annex III deferral to December 2027, plus investigations and policies.
11 min read
Use Cases
Writing Legal Memos with AI: A Method That Survives Partner Review
An AI legal memo survives partner review when the model drafts only from sources you supply, every authority is checked three ways and the checks are logged.
10 min read