Discover our new Policy Prototyping report on the EU AI Act's value chain requirement  

Read and download here
A middle-aged Caucasian woman juggles deadlines while sipping coffee at her standing desk, managing emails, video meetings, and detailed Gantt charts for a high-stakes corporate rollout
15.07.2026

brAInfood - Digital Omnibus on AI: (new) deadlines AI Act

Following the Council’s final approval of the Digital Omnibus on AI, formally known as the Regulation on the Simplification of the Implementation of Harmonised Rules on Artificial Intelligence, several provisions and key deadlines of the AI Act have been amended. In essence, the AI Act establishes different regulatory regimes for specific categories of AI systems, each accompanied by corresponding obligations for providers and deployers.

To help navigate the updated implementation timeline, the Knowledge Centre Data & Society has prepared a timeline highlighting important changes to the implementation dates of the AI Act and explaining what organisations need to be aware of in the coming months and years. 

Download the timeline

Download the AI Act timeline via the button below. 

brAInfood EU AI Act timeline Knowledge Centre Data & Society 2026 (pdf, 259KB)

Download

Key amendments to the AI Act

Next to the revised timeline, a number of other adjustments were made. Those changes include: 

  • The AI literacy obligation is softened. Providers and deployers of AI systems shall take measures to support the development of AI literacy of their staff and other persons on their behalf instead of ensuring, to their best extent, a sufficient level of AI literacy of those persons. There will be no obligation anymore to guarantee any specific level of AI literacy of any individual. The Commission and the Member States have a best-efforts obligation to support and promote AI literacy among the personnel of providers and deployers. 
  • Broader scope of processing special categories of personal data. Providers and deployers of all AI systems and models, not only those classified as high-risk, will be permitted to process special categories of personal data for the purposes of detecting and mitigating bias, provided that appropriate safeguards are in place. Such processing is allowed only where it is strictly necessary, namely in exceptional circumstances where bias detection and mitigation cannot be achieved through other means, including the use of synthetic or anonymised data. 
  • Introduction of new prohibited AI practices. As of 2 December 2026, a new prohibition will apply to AI systems that can generate or manipulate non-consensual intimate content (NCIC) and child sexual abuse material (CSAM). For providers, the prohibition covers not only AI systems intended to generate or manipulate such material , but also systems where such generation or manipulation is a reasonably foreseeable and reproducible outcome and where there are no reasonable and adequate technical safety measures and other safeguards to reliably prevent the generation or manipulation of such material or performance. For deployers, the prohibition applies where an AI system is used for the purpose of generating or manipulating NCIC or CSAM. 
  • Postponement of high-risk rules. The AI Act will apply from 2 December 2027 to AI systems referred to in Article 6(2) and Annex III. AI systems covered by Article 6(1) and Annex I will be subject to the AI Act starting on 2 August 2028. 
  • An override rule in favor of AI systems of Annex I. For high-risk AI systems that are covered by the Union harmonisation legislation listed in Section A of Annex I and simultaneously fall within a category identified in Annex III, the provider is required to comply with the conformity assessment procedure established under the relevant Union harmonisation legislation referred to in Section A of Annex I. 
  • The Machinery Regulation is exempted from the requirements for high-risk AI systems of the AI Act. AI machinery products will only have to comply with sector-specific safety rules, rather than the rules set out in the AI Act, since these products have been moved from Annex I, Section A to Section B. 
  • broader SME exemption for small mid-caps enterprises (SMCs). The SME exemption has been expanded on several points to SMCs, allowing them to benefit from lighter regulatory requirements. The reasoning beyond this amendment is that, although SMCs grow faster and have a higher level of innovation, they face similar challenges as SMEs in terms of administrative burden, while being subject to the same rules as large companies.  
  • The possibility of limited requirements for Annex I systems. For high-risk AI systems listed in Annex I, the application of certain AI Act requirements may be limited where the relevant Union harmonisation legislation already provides an equivalent or higher level of protection for health, safety, or fundamental rights. To clarify the scope and conditions of such limitations, the Commission is required to adopt delegated acts by 2 August 2027. 
  • The AI Office's remit will be expanded. It will have exclusive responsibility for supervising and enforcing the obligations of AI systems based on general-purpose AI models, provided that the model and system are developed by the same provider, or the providers are part of the same group (with some exceptions). Additionally, the AI Office will have exclusive responsibility for AI systems that constitute, or are integrated into, a very large online platform or very large online search engine, as defined in the Digital Services Act. 

  

Next steps

At present, the Digital Omnibus on AI has not yet been published in the Official Journal of the European Union, although publication is expected prior to the upcoming deadline of 2 August 2026. Once published, the legislative act will enter into force on the third day following its publication. Furthermore, the Commission intends to provide greater clarity as quickly as possible through guidelines, codes of practice, and delegated acts, helping organisations to navigate compliance obligations and address implementation challenges more effectively. 

About

The brAInfood is available under a CC BY-NC-ND 4.0 licence. This means that this publication may be shared under the following conditions:

  • Attribution: You must correctly attribute the licensor, the Knowledge Centre Data & Society, include a link to the licence and indicate whether any changes have been made. This may be done in any reasonable manner, as long as it does not give the impression that the licensor endorses you or your use.
  • Non-commercial: You may not use the material for commercial purposes.
  • No Derivatives: If you remix, transform, or build upon the material, you may not distribute the modified material.

     

COPYRIGHT IMAGEs and illustrations

Image by wolfhound911 via Adobe Stock. 

The illustrations in this brAInfood were created by Toast Confituur Studio

Authors

Sabra Ibnouthen

Sabra Ibnouthen

email hidden; JavaScript is required
Sultan Erdogan

Sultan Erdogan

email hidden; JavaScript is required
Shannen Verlee

Shannen Verlee

email hidden; JavaScript is required