The Digital Omnibus created two EU AI Act deadlines. Most operators have mapped one.
What the Digital Omnibus actually changed
The political agreement reached on 7 May 2026 between the European Parliament and Council on the Digital Omnibus — a legislative package consolidating several digital regulation simplification measures — included a significant adjustment to the EU AI Act's implementation timeline. The Annex III high-risk AI system obligations, which were scheduled to become enforceable in August 2026, are now postponed to 2 December 2027. That is a sixteen-month extension for a category of obligations that requires companies to conduct conformity assessments, maintain technical documentation, and implement quality management systems for AI systems classified as high-risk under the Act's specific use-case taxonomy.
The postponement is real, it is material, and it is being interpreted in many organisations as a signal to defer all EU AI Act preparation work until the revised deadline approaches. That interpretation will prove costly for the companies that adopt it.
The obligation the Omnibus did not touch
Article 50 of the EU AI Act — the deployer transparency obligations — was not included in the Omnibus adjustment. The Article 50 requirements remain on schedule to become enforceable on 2 August 2026.
Article 50 covers a specific set of transparency obligations that apply whenever an AI system interacts with people in ways that are not immediately obvious. The clearest instance is the obligation to inform people when they are communicating with an AI system rather than a human, in contexts where a reasonable person might not expect this: customer service chatbots, AI-generated responses in support workflows, automated decision communications. Deployers must ensure that people are informed before or at the point of that interaction, and must maintain the technical and operational means to do so.
What this requires operationally is not a documentation exercise. It requires a live inventory of every AI system in production that interacts with people, a review of where those systems produce outputs without disclosure, the implementation of disclosure mechanisms where they are absent, and the internal governance processes to keep this state current as the stack evolves. For a company with twenty AI systems across customer service, HR tools, marketing automation, and financial workflows, this is a multi-week project.
Why the confusion is widespread
The Omnibus communication was framed, understandably, in terms of compliance burden reduction. The industries most affected by Annex III — manufacturers, financial services, healthcare — had been lobbying heavily for timeline relief on the conformity assessment requirements, and the agreement was presented as delivering that relief. The coverage focused on the postponement of Annex III obligations as the headline, with Article 50's continued August deadline mentioned in passing or omitted entirely.
The result is a widespread assumption among operators that the August 2026 deadline is now either gone or substantially reduced in scope. It is neither. Companies entering the second half of June 2026 without a current inventory of their AI systems, without disclosure mechanisms reviewed against Article 50's requirements, and without a designated person responsible for AI literacy under Article 4 — which has been in force since August 2025 — are behind the compliance state the Act requires of them by 2 August.
Mapping the two exercises as separate projects
The practical approach is to treat Article 50 and Annex III as two distinct projects with two distinct owners, timelines, and scopes.
The Article 50 exercise begins with an inventory. Every AI system the organisation deploys or uses in a professional context needs to be identified, described, and assessed against the Act's transparency requirements. That inventory becomes the foundation for the disclosure review: where does an AI system interact with a person, what does that person know about the nature of that interaction, and what is required to bring disclosure posture into compliance by 2 August. The AI literacy obligation under Article 4 is embedded in the same exercise.
The Annex III exercise is genuinely deferred to December 2027. But the inventory produced for Article 50 is the prerequisite for the conformity assessment that December 2027 will require for any high-risk system. Companies that complete Article 50 preparation thoroughly will have a material head start on the subsequent exercise.
The sixteen-month extension is real relief for a genuinely demanding set of obligations. The companies that benefit from it are the ones using the extension for Annex III preparation — not the ones treating it as grounds to defer work that was due in August regardless.