The EU just gave AI hiring vendors an extra sixteen months to get compliant. It did not give you extra time on the one requirement that actually touches your candidates, and almost nobody in talent acquisition seems to have noticed.
The Digital Omnibus on AI, the European Commission's package reworking parts of the AI Act, pushed back the deadline for the high-risk obligations covering hiring and worker-management systems. The original date was August 2, 2026. It's now December 2, 2027. The Commission tabled the proposal on November 19, 2025, Parliament and Council reached a provisional agreement in early May 2026, approved it formally in June, and it entered into force in late July. Recruitment tools sit in Annex III as a stand-alone high-risk use, which is why this lands on you directly and not just on the vendors selling you the tools.
What Actually Moved
- Stand-alone high-risk systems, including hiring and worker management: December 2, 2027
- High-risk AI embedded in already-regulated products: August 2, 2028
- Article 50 transparency obligations: August 2, 2026, unchanged
Read that last line again. The transparency requirement, the one with nothing to do with model documentation or risk assessments and everything to do with what you actually tell a candidate before they talk to a bot, kept its original date.
What Article 50 Actually Asks For
It's a small ask on paper. Tell candidates, clearly and at first interaction, that they're engaging with an AI system: what it does, and how to reach a human instead. That disclosure has to live inside the invitation itself. A link to a policy page buried in your footer doesn't count.
Why the Easy Part Is the Part Nobody Owns
The big compliance programme gets a budget line and a steering committee, because it's expensive and visible. The one-sentence disclosure gets neither, because it's cheap and invisible, and it falls into a gap between three teams who each assume it's someone else's job. Legal owns the larger AI Act programme. Vendors handle their own product claims. Recruiters send the invitations. The sentence that has to sit inside that invitation belongs to all three of them and none of them, and things that belong to nobody don't get done.
I've watched this exact pattern play out across hiring before. The expensive, visible part of a process gets attention. The cheap part that actually touches the candidate gets skipped.
What to Do Monday
Three things, all small enough to finish this week.
- Write the disclosure line and put it inside the invitation itself: what AI is used, what it does, and how to reach a human instead.
- Document every automated step in your hiring process, including the ones a vendor runs on your behalf. You can't disclose a step you haven't written down.
- Record the date you started disclosing. That's your evidence if anyone ever asks.
This is an afternoon of work. Possibly less.
What's at Stake
Article 50 breaches carry a maximum penalty of €15 million or 3% of global annual turnover, whichever is higher, with a reduced threshold for qualifying small and medium enterprises and proportionality built into how it's applied. That's the same maximum penalty as the rest of the AI Act, attached to the shortest, simplest obligation in the whole regulation.
The Part That Actually Matters
Sixteen extra months went to the vendors. None of it went to the candidate who is about to sit down, record themselves answering questions, and has no idea whether a person or a system is on the other end.
So: does your hiring process currently disclose that? And if it does, whose job was it to put it there?
Source: European Commission, Digital Omnibus on AI (2026), amending Regulation (EU) 2024/1689 (the AI Act).
Want a hiring process that's honest with candidates by default?
See our sample interview structure, learn more about embedded interviewers, or book a call to talk through what this means for your process.
