# Digital Omnibus adopted: what really changes in the EU AI Act > The Digital Omnibus amends rather than replaces the EU AI Act. It moves Annex III high-risk requirements to 2 December 2027 and Annex I requirements to 2 August 2028. General Article 50 transparency remains due on 2 August 2026. On 13 July 2026, the text was signed but not yet in force. - Tags: Digital Omnibus, AI Act, Article 50, AI literacy, high-risk AI - Main keyword: Digital Omnibus AI Act ## Introduction The Digital Omnibus changes part of the high-risk timetable without postponing the AI Act as a whole. Here are the dates, the Article 4 changes and the actions organisations should prioritise. ## Content The **Digital Omnibus on artificial intelligence** has completed the EU legislative process: the European Parliament approved the text on 16 June 2026 and the Council gave its final green light on 29 June. The signed wording is therefore settled. One legal distinction still matters, however. As of 13 July 2026, final publication in the Official Journal of the European Union had not yet been identified. The Omnibus will enter into force on the third day after that publication. This status point does not change the operational message. The Omnibus reschedules part of the high-risk framework, rewrites the AI literacy provision and adds prohibited practices. It does not postpone the AI Act as a whole. The general transparency duties under Article 50 are still due from **2 August 2026**. ## Adopted does not yet mean applicable The [European Parliament](https://www.europarl.europa.eu/news/en/press-room/20260611IPR45207/ai-act-ep-approves-simplification-measures-and-nudifier-app-ban) and the [Council of the European Union](https://www.consilium.europa.eu/en/press/press-releases/2026/06/29/artificial-intelligence-council-gives-final-green-light-to-simplify-and-streamline-rules/) have completed their examination. The [signed text PE-CONS 30/1/26 REV 1](https://data.consilium.europa.eu/doc/document/PE-30-2026-REV-1/en/pdf) records the agreed amendments. Three states should be kept separate: - **adopted and signed**: the legislative procedure is complete and the wording is settled; - **published and in force**: the text has appeared in the Official Journal and its entry-into-force period has elapsed; - **applicable**: a particular rule is producing effects for the actors concerned on the date assigned to it. Until Official Journal publication, [Regulation (EU) 2024/1689](https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng) remains applicable in its current form. The Omnibus does not replace that Regulation. It **amends** it when it enters into force, with several provisions following their own application dates. ## What actually changes ### Annex III high-risk systems move to 2 December 2027 The most visible change concerns stand-alone systems classified as high-risk under Article 6(2) and Annex III, including certain uses in recruitment, education, credit, essential services, justice and biometrics. The core requirements in Chapter III, Sections 1 to 3 β€” risk management, data governance, technical documentation, logging, transparency, human oversight and conformity assessment β€” are rescheduled to **2 December 2027**. For high-risk systems embedded in regulated products listed in Annex I, the new date is **2 August 2028**. The extra time does not make compliance a 2027 project. Organisations still need to inventory their systems, confirm their role, assemble evidence and set up governance. The change gives them a longer runway to do that work properly. ### Article 4 becomes more proportionate The Article 4 currently in force requires providers and deployers to take measures to ensure, to their best extent, a sufficient level of AI literacy among the relevant people. Point 5 of the Omnibus provides new wording focused on measures designed to **foster the development of AI literacy**. It also makes clear that an organisation is not required to guarantee a particular level for every individual. Training therefore remains relevant. Organisations must still select measures suited to technical knowledge, experience, education, the context of use and the people who may be affected. The practical emphasis moves towards showing that those measures are relevant and proportionate, rather than guaranteeing one uniform individual outcome. Until the Omnibus enters into force, the current Article 4 wording remains the applicable rule. ### New prohibited practices are added The Omnibus adds targeted prohibitions to Article 5 concerning certain systems that generate or manipulate non-consensual intimate or sexual representations and child sexual abuse material. The signed text defines their scope and schedules them for **2 December 2026**, subject to the amending Regulation first entering into force. Businesses developing image-generation or image-transformation features should already examine whether their products, datasets, safeguards and authorised uses could fall within this scope. ## What is not generally postponed ### Article 50 transparency remains due on 2 August 2026 The date relevant to a broad range of organisations does not move: the general Article 50 transparency obligations still apply from **2 August 2026**. They include, depending on the system and use case: - informing people when they interact directly with certain AI systems; - marking certain generated or manipulated content in a machine-readable format; - disclosures for emotion-recognition and biometric-categorisation systems; - disclosures for certain deepfakes and synthetic content. The Omnibus changes paragraph 7 on EU codes of practice and allows a transition until **2 December 2026** for certain generative systems placed on the market before 2 August 2026. That targeted transition is not a general postponement of Article 50. ### There is no single new date for the whole AI Act General-purpose AI obligations, governance rules and parts of the penalty framework have followed their own application timetable. The Omnibus changes several supervision and implementation mechanisms, but it does not move the entire Regulation to December 2027. The claim that β€œthe AI Act has been delayed until 2027” therefore combines different legal tracks. Only part of the high-risk timetable has shifted. ## Dates to keep on the compliance calendar | Topic | Date or trigger | Position on 13 July 2026 | | --- | --- | --- | | New Article 4 wording | Third day after OJ publication of the Omnibus | Signed, not yet in force | | General Article 50 transparency | 2 August 2026 | Date retained | | New Article 5 prohibitions | 2 December 2026 | Scheduled by the signed text | | Transition of certain legacy systems to Article 50(2) | 2 December 2026 | Scheduled by the signed text | | Annex III high-risk systems | 2 December 2027 | Rescheduled by the Omnibus | | Annex I high-risk systems | 2 August 2028 | Rescheduled by the Omnibus | Our [AI Act calendar](/calendrier-ai-act) keeps applicable rules separate from signed texts, proposals and dates that remain conditional. ## Four actions to take now 1. **Keep the Article 50 workstream moving.** Review conversational interfaces, synthetic content, marking mechanisms and user-facing notices before 2 August 2026. 2. **Replan high-risk compliance without pausing it.** Use the additional time to strengthen inventory, classification, documentation and ownership. 3. **Document proportionate AI literacy measures.** Define audiences, risks, learning objectives, awareness formats and evidence. Do not discard training programmes that are already under way. 4. **Screen for the new prohibitions.** Examine products and use cases involving intimate or sexual content generation and manipulation, including reasonably foreseeable misuse. The practical conclusion is straightforward: **the high-risk calendar changed, but the AI Act was not postponed as a whole**. The immediate task is to identify which timeline applies to each system. The [compaia assessment](/diagnostic) is a starting point for that classification. ## Frequently asked questions ### Is the Digital Omnibus already applicable? No. On 13 July 2026, the text was adopted and signed, but final publication in the Official Journal had not yet been identified. Article 4 of the Omnibus provides for entry into force on the third day after publication. ### Are all AI Act obligations postponed until December 2027? No. That date mainly concerns Chapter III, Sections 1 to 3 for Annex III high-risk systems. The general Article 50 transparency requirements remain scheduled for 2 August 2026. ### Do organisations still need AI literacy training? Yes. The future wording is more proportionate and does not require a uniform individual level to be guaranteed, but organisations still need suitable measures to foster AI literacy. The current wording remains applicable until the Omnibus enters into force. ### What should be prepared for Article 50? Identify systems that interact with people or generate and manipulate content, determine the applicable notice and marking duties, assign owners and retain evidence of the technical and editorial choices made. ## Official source - [Source](https://www.aiacto.eu/fr/blog/digital-omnibus-ai-act-report-2027)