The Digital Omnibus and AI hiring: what changed, what did not

The EU moved the deadline for high-risk AI hiring systems. It did not move the transparency deadline.

The Digital Omnibus on AI is adopted. Parliament passed it on June 16, 2026. The Council followed on June 29, 2026. For anyone running AI in hiring, the law now runs on two clocks, and most vendor decks only show one of them.

One clock covers the heavy obligations for high-risk systems. The other covers transparency toward the people who interact with your AI, and it starts this summer.

Last reviewed: July 2026.

What moved

The high-risk obligations for stand-alone Annex III systems now apply from December 2, 2027. Employment sits in Annex III. A tool that screens, ranks, or filters candidates for a job is inside the category the delay covers, so the core rulebook for AI hiring lands on that date.

The embedded-systems date moved further out. Annex I covers AI built into another product, and those obligations now apply from August 2, 2028. Most hiring AI is stand-alone software rather than a component inside something else, so December 2, 2027 is the date a talent leader should hold.

One more date in the omnibus has nothing to do with hiring. The new Article 5 prohibitions, covering non-consensual intimate imagery and CSAM generation, apply from December 2, 2026. It belongs on the full calendar even if it never touches your stack.

What did not move

Article 50 transparency duties still take effect on August 2, 2026. That is weeks after adoption and more than a year ahead of the high-risk rulebook. When a person interacts with an AI system, the person has to be told.

This is the part NODES already does. A candidate in a NODES AI interview is told that AI is assisting. The disclosure lives inside the interview itself, in front of the candidate, at the moment the interaction happens.

If your AI interviews, chatbots, or screening flows do not disclose themselves today, that gap comes due this summer. The omnibus did not buy you any time on it.

What the deferral is tied to

The deferral is tied to the registration mechanism. The new dates for high-risk systems run with the machinery that registers them, which is the part of the Act the omnibus reworks.

Read it that way and the delay has a shape. The extra time exists so the registration plumbing exists. It is not a reclassification of employment AI. The category did not move. The clock did.

What to do with the runway

That runway is a build window if you treat it as one.

• Meet the transparency duty first. Audit every candidate-facing AI touchpoint for disclosure before August 2, 2026. Interviews, chatbots, scheduling and screening flows. If a person interacts with AI, the person hears it from you.
• Inventory every AI system that touches candidates or employees, and classify each one against Annex III. Most teams find tools they forgot they bought, in departments that never went through procurement.
• Ask every vendor two questions. Who makes the reject-or-advance call, the tool or a person? And what record exists of it? Mobley v. Workday is what the unanswered version of those questions looks like in federal court.
• Use the time to connect your hiring data. The high-risk obligations will land on evidence: what the system did, on whose data, with what outcome. Teams that already link screening signals to results will answer from records. Everyone else will answer from memory. The published research shows what that connection surfaces.
• Build the governance layer once, not per vendor. The AI governance use case covers what an evidence layer across your whole AI stack looks like, and the security and compliance page covers the posture behind it.

The companies that spend this runway well will not be the ones that waited for December 2027. They will be the ones that built the evidence while the clock was generous.

Frequently asked questions

Did the Digital Omnibus delay the EU AI Act for hiring? For the high-risk obligations, yes. Stand-alone Annex III systems, including employment, now apply from December 2, 2027, and embedded Annex I systems from August 2, 2028. The Article 50 transparency duties were not delayed. They still take effect August 2, 2026.

When does the EU AI Act apply to AI hiring tools? Transparency duties apply from August 2, 2026. The high-risk obligations for employment systems apply from December 2, 2027. Both dates come from the Digital Omnibus on AI, adopted in June 2026.

Is employment still treated as high-risk under the omnibus? Yes. Employment remains in Annex III. What changed is the timing. Whether the obligations apply at all was never in question.

What is the registration mechanism the delay is tied to? The omnibus ties the new application dates for high-risk systems to the machinery that registers those systems. The deferral runs with that mechanism instead of arriving as a flat date change.

What should talent teams do before December 2027? Meet the transparency duty now, inventory and classify every AI system touching candidates, get straight answers from vendors on who makes each decision, and connect screening data to outcomes so the evidence exists before the obligations do.

Related reading

• Mobley and the agent theory
• AI governance and compliance use case
• Security and compliance
• The published hiring research

See how your AI stack classifies, and what a decision record looks like. Book a 30-minute walkthrough.