Copenhagen: The Digital Omnibus on AI has given European companies something they asked for loudly and regulators granted reluctantly: more time. The deal between Parliament and Council postpones the AI Act’s high-risk obligations to fixed dates in 2027 and 2028, and the Council’s sign-off is the last formal step.

What the Digital Omnibus on AI changes

The Commission proposed the omnibus in November 2025 after providers warned that the harmonised standards needed to prove compliance would not exist by August 2026, the original start date for high-risk rules. Without standards, firms would have faced binding duties with no agreed way to meet them.

Negotiators reached a provisional agreement in May 2026, and Parliament adopted the text in June. The agreed timetable replaces the Commission’s floating mechanism, which tied the start date to the availability of standards, with fixed dates. Stand-alone high-risk systems, such as those used in hiring, credit scoring or education, move to December 2027. High-risk AI embedded in regulated products, such as machinery and medical devices, moves to August 2028.

Why fixed dates beat a floating clock

Parliament pushed for fixed deadlines for a reason that lawyers will recognise. A deadline that depends on a future Commission decision gives companies no planning certainty, and it gives enforcement authorities no firm point from which to supervise. A calendar date, even a distant one, lets compliance teams budget, hire and audit against a known target.

The Digital Omnibus on AI therefore trades speed for predictability. That trade is defensible, but it carries a cost. For roughly sixteen additional months, systems that influence employment, credit and access to services will operate without the full risk-management, data-governance and human-oversight duties the Act envisaged.

Winners, losers and the compliance question

The largest providers can use the extra time to finish conformity assessments, but they were already preparing. Smaller firms gain most from the delay, because they depend on standards, templates and guidance that are still being finalised. The Commission also proposed lighter documentation duties for small mid-cap companies, a move aimed at the scale-up problem that Mario Draghi’s competitiveness report highlighted.

Civil society groups see the other side. They argue that every postponement weakens the signal that the EU will enforce its rules, and that workers and consumers carry the risk during the gap. Prohibited practices and the obligations for general-purpose AI models are not delayed, which limits that concern but does not remove it.

What businesses should do now

The Digital Omnibus on AI is not a pause button for governance. Companies that treat the delay as a reason to stop work will face a crowded compliance queue in 2027, just as national authorities start to enforce. Three steps make sense now: classify every AI system against the high-risk annexes, map the standards that will apply and assign clear internal ownership for oversight.

In short, the Digital Omnibus on AI buys time but not exemption. The firms that use the extra months to build documentation and testing routines will enter the 2027 and 2028 deadlines with an advantage, while the rest will face the same obligations on a shorter runway.