The EU’s Digital Omnibus on AI: What the Provisional Agreement Really Changes
On 7 May 2026, the European Parliament and the Council of the EU reached a provisional agreement on the amendments to the European Commission’s proposed Digital Omnibus on AI.
How does the Digital Omnibus link to the EU AI Act?
The EU AI Act came into force on 1 August 2024, with its provisions starting to apply on a staggered basis. While some provisions are already in effect, most of the remaining provisions, including those relating to standalone high-risk AI systems, are set to apply from 2 August 2026. The obligations relating to high-risk AI systems embedded in regulated products are set to apply from 2 August 2027.
However, delays to guidance and standards on high-risk AI, coupled with concerns that the Act’s compliance framework is overly complex and could undermine EU competitiveness in the AI sector, prompted the European Commission to propose the Digital Omnibus. The Digital Omnibus contains a number of proposed amendments to the EU AI Act, aimed at simplifying the implementation of the EU AI Act across the EU and delaying the timetable for the application of high-risk AI system obligations.
What are the key proposed changes?
- Delayed application of high-risk AI obligations. Obligations for standalone high-risk AI systems would apply from 2 December 2027. Obligations for high-risk AI systems embedded in regulated products would apply from 2 August 2028.
- New prohibition on harmful AI uses. The list of prohibited AI practices would be expanded to ban AI systems (so-called “nudifier” apps) from generating non-consensual sexual or intimate content or child sexual abuse material (CSAM). This prohibition would apply from 2 December 2026.
- Grace period for transparency obligations for AI-generated content. Providers of generative AI systems that are already on the market by 2 August 2026 would have a grace period of four months until 2 December 2026 (under the original Digital Omnibus on AI proposal, this was six months (until 2 February 2027)) to comply with transparency/labelling requirements relating to artificially-generated audio, image, video or text content.
- Reinstatement of registration obligation for exempt systems. Where an AI system falls into one of the high-risk AI categories (for standalone AI) but the provider/developer concludes that it falls within an exemption (for example, because it is intended to perform a narrow procedural task or improve the result of a previously completed human activity), the provider is still required to register in the EU database of high-risk AI systems. The original Digital Omnibus on AI had proposed removing this requirement, but this has been reinstated.
- Reinstatement of the “strict necessity” standard for special category data. The current EU AI Act says that special category data can be processed where “strictly necessary” to ensure bias detection and correction in relation to high-risk AI systems (subject to various conditions). The original Digital Omnibus on AI had proposed watering down this requirement, but this has been reinstated.
- Support on the interplay between sectoral legislation and the EU AI Act. By no later than 1 August 2027, the European Commission would be required to publish guidelines on how to ensure consistency, avoid duplication, and minimise compliance burdens when complying with both sectoral legislation and the EU AI Act.
Next steps
The amendments agreed provisionally by the European Parliament and the Council of the EU must be endorsed by both institutions before the legislation is adopted formally and then published in the Official Journal of the EU. There will be pressure to complete these steps quickly enough so that the amendments take effect before 2 August 2026.
