VERIK / V048 / 02 JUL 2026
Five CategoriesGovernance

The Two Clocks That Were Written Into the Statute

On June 29, 2026, the Council of the European Union gave final approval to the simplification package known informally as the AI Omnibus. The press release used the language of streamlining. What the text actually did was inscribe two governance clocks into the statute and let them run at different speeds.

The transparency clock, anchored in Article 50 of the AI Act, was originally set to become enforceable on August 2, 2026. Under the amendments the Council has now adopted, the transparency and marking obligations for general purpose AI providers and for deployers of certain interactive systems are pushed back to December 2, 2026.

The substrate clock, which governs Annex III high risk systems used in biometrics, critical infrastructure, education, employment, essential services, migration, and the administration of justice, moves further still. Those obligations, originally scheduled for August 2, 2026, now become enforceable on December 2, 2027. Annex I embedded high risk systems, the ones integrated into regulated products such as lifts, toys, medical devices, and certain machinery, do not become enforceable until August 2, 2028. National regulatory sandboxes, the mechanism through which small and medium sized providers were supposed to be able to test high risk systems under supervision, are deferred to August 2, 2027.

Two new Article 5 prohibitions were also added: the generation of child sexual abuse material and the generation of non consensual intimate imagery. Those apply from December 2, 2026.

The Structural Fact

The Cyprus presidency framed the vote as a proportionality achievement. Marilena Raouna, the deputy minister for European affairs, said that "technological progress must always go hand in hand with the protection of our fundamental values." The Council statement presents the package as an alignment of the AI Act with the Digital Omnibus that governs the broader digital single market.

The structural fact is different. The Council has written into statute a governance regime in which the transparency layer, the layer that tells a user or a market surveillance authority that a system is generative or that content was produced by a model, is now formally decoupled from the substrate layer, the layer that determines whether a system falls under conformity assessment, human oversight, logging, and accuracy obligations because of the domain it operates in.

Under the original schedule those two layers moved on the same calendar. A market surveillance authority in Ireland or Italy could plan enforcement around a single August 2, 2026 anchor. A provider of a high risk employment screening system knew that both the disclosure obligations of Article 50 and the conformity obligations of Annex III would apply from the same date. Compliance planning and enforcement planning could share a clock.

The Omnibus removes that alignment. The transparency clock now runs sixteen months ahead of the Annex III clock and twenty months ahead of the Annex I clock. National sandboxes, the mechanism that was supposed to let providers rehearse the high risk regime before it applied to their products, arrive twelve months before Annex III enforcement and thirteen months after transparency enforcement.

What This Codifies

The Council's own case is that the deferral gives the AI Office, national authorities, and providers more time to develop harmonized standards and technical guidance. The European Commission draft guidelines on classification of high risk systems, published on May 19, 2026 with the consultation closing on July 23, 2026, are the mechanism through which some of that clarity is meant to arrive.

The counter reading is that the standards work has not caught up with the deployment tempo, and the statutory clocks have been adjusted to reflect where the work is rather than where the risk is. In sectors covered by Annex III, agentic systems are already being deployed. National labor ministries are procuring hiring screeners. Border management agencies are running biometric identification pilots. Financial services regulators are being asked to classify credit scoring tools that call out to third party foundation models. The substrate clock has been moved back to 2027, but the deployment clock has not.

The Osborne Clarke analysis of the draft guidelines observes that the classification test the Commission is proposing turns on whether the system materially influences a decision affecting individuals. That is a substantive test, not a formal one, and it is the kind of test that consumes market surveillance capacity. Under the Omnibus schedule the burden of that test begins in December 2027, not August 2026.

The DLA Piper reading of the same draft flags an operationally important point: the Article 6(3) filter that lets a system escape high risk classification if it performs a narrow procedural task or a preparatory task, or if it improves the result of a completed human activity, or if it detects decision patterns without replacing human assessment, cannot be applied to profiling systems under Annex III. Profiling is always high risk. That carve out sharpens the substrate clock: profiling systems, by law, will not be reachable through the escape valve and will land squarely under the December 2027 regime.

The Two Clocks in Practice

A market surveillance authority preparing an enforcement program in the second half of 2026 now has one enforceable regime and one that will not be enforceable for another sixteen months. Either the authority builds a transparency enforcement capability that will be idle relative to the substrate regime for over a year, or it builds a substrate capability that will not have statutory basis to act until December 2027.

A deployer procuring an agentic system for education, credit, or public benefits administration in late 2026 or early 2027 will do so under a transparency regime and a set of Article 5 prohibitions, but not under Annex III conformity assessment, logging, human oversight, or accuracy obligations. The Council has, by statutory choice, allowed a window in which agentic high risk deployment is legal, undisclosed as high risk under Annex III, and not subject to the conformity obligations that were meant to be the substrate check.

The Refrain

Publication in the Official Journal is pending. The ordinary legislative procedure 2025/0359 is closed. The clocks are set.

What remains on the table:

The policy instruments and the deployment tempo are not aligned.