EU AI Act on Hold: Digital Omnibus Could Push High-Risk Duties to December 2027

Digital Omnibus 2026: The EU Debates Postponing the AI Act's High-Risk Obligations

The EU faces a politically sensitive decision. On November 19, 2025, the European Commission tabled the so-called Digital Omnibus on AI — a package designed to push the EU AI Act's core high-risk obligations from the original deadline of August 2, 2026 to December 2, 2027. The second trilogue between Parliament, Council, and Commission ended without agreement on April 28, 2026. The next attempt is scheduled for May 13, 2026.

What the Digital Omnibus Actually Changes

At the core are the duties on providers and deployers of high-risk AI systems used in areas such as recruiting, credit scoring, critical infrastructure, and law enforcement. The Commission justifies the delay with the slow designation of national supervisory authorities and the unfinished harmonized standards needed as a basis for conformity assessments. The obligation around staff AI literacy is also set to be softened from a hard requirement to a more open-ended expectation.

What Happens If the Omnibus Is Not Adopted in Time

If the Digital Omnibus is not formally adopted before August 2, 2026, the AI Act's obligations apply from that date as originally written. This also includes the standing requirement for each member state to set up at least one national AI regulatory sandbox by then. In parallel, the AI Office's enforcement powers against providers of large general-purpose AI models still kick in from August 2026, regardless of the Omnibus.

Pros and Cons — Why the Negotiations Are So Hard

Supporters, mainly from industry associations and parts of the EPP group, argue that without harmonized standards and designated authorities, businesses face legal uncertainty. Critics in Parliament, data protection authorities, and civil society warn that a delay of more than a year would lift oversight from exactly the high-risk systems the law was written for. That conflict line is what is currently blocking the trilogue.

Why This Matters

For companies running high-risk AI use cases — for example in recruiting or medical diagnostics — the uncertainty is expensive. Compliance teams must plan two scenarios in parallel: a full start of obligations on August 2, 2026 and a postponement to December 2027. Anyone who postpones conformity assessments, data quality reviews, or bias testing risks ending up with finished but unauditable systems on deadline day. The decision on May 13, 2026 is becoming a strategic anchor point for AI roadmaps across Europe.

Practical Example

A mid-sized HR tech company from Munich has built an AI system to automatically pre-screen job applications. Compliance plans two parallel releases: an "Annex III track" with documented training data, formal risk management, and notified bodies for a possible conformity assessment from August 2026, plus a leaner transition track with expanded monitoring in case the Digital Omnibus pushes the high-risk obligations out to December 2027. The product stays sellable regardless of the trilogue outcome — but with significantly more effort than was planned twelve months ago.