Digital Omnibus on AI

The Digital Omnibus on AI started as Commission proposal COM(2025)836 of 19 November 2025. On 7 May 2026, the Council and European Parliament reached a provisional political agreement on the AI Act amendments. That agreement is not yet the same as binding law: legal-linguistic revision, formal approval and publication in the Official Journal still need to follow. Until then, the existing AI Act remains the legal baseline. For organisations, the deal is still concrete enough to update planning now.

The biggest practical change in the political agreement is the deferral of high-risk AI obligations. For most high-risk systems listed in Annex III, the core obligations move to 2 December 2027. For high-risk AI embedded in products covered by existing EU sectoral safety legislation, such as medical devices or machinery, 2 August 2028 becomes the relevant date. This gives organisations more implementation time, but it does not remove the need to classify systems now, set supplier expectations and build evidence.

The original Commission proposal sought to shift Article 4 from a direct organisational obligation to an encouragement role for the Commission and Member States. The EDPB and EDPS criticised that move and advised keeping the direct obligation. The official communication on the 7 May 2026 political agreement does not spell out Article 4 as a settled amendment. The correct message is therefore: the existing AI Act remains in force until a formal amendment applies, and AI literacy remains commercially and legally prudent to organise with evidence.

The Commission originally proposed removing the registration requirement for certain non-high-risk claims. The EDPB and EDPS warned this would weaken accountability and public scrutiny. The political agreement changes that course: providers placing an AI system on the market in a high-risk area while claiming an exception means it is not high-risk must register that assessment in the EU database. That makes classification evidence more important, not less important.

The political agreement sets the transparency deadline for AI-generated content at 2 December 2026. That is later than the original AI Act date, but earlier than the Commission proposal of 2 February 2027. The deal also introduces an explicit ban on AI systems that generate or manipulate child sexual abuse material or non-consensual intimate images; according to the agreement communication, providers must comply with these new requirements by 2 December 2026. For GPAI-based systems, centralised AI Office supervision remains an important simplification element.

The Digital Omnibus continues to push proportionate application for SMEs, startups and small mid-cap companies. This includes simplified quality management systems, better access to regulatory sandboxes and implementation support. That lowers execution burden, but it does not remove the core questions: which AI system are you using, does it fall under Annex III, what is your role as provider or deployer, and what documentation can you show to regulators or buyers?

On 21 January 2026, the EDPB and EDPS published their Joint Opinion on the Digital Omnibus on AI. Their core message remains relevant after the political agreement: simplification must not lower the level of fundamental rights protection. Some of their criticism appears to be reflected in the political outcome, especially around registration and stronger safeguards for processing special-category personal data in bias detection. Article 4 remains the key open point to communicate carefully until the final text is available.