Did the EU delay the AI Act rules for hiring?

Yes — the application date for high-risk AI obligations, which cover AI used in recruitment and employment, moved from 2 August 2026 to 2 December 2027 under the "Digital Omnibus on AI." The European Parliament endorsed the change on 16 June 2026 and the Council of the EU gave final approval on 29 June 2026. The obligations themselves were not repealed; only the deadline moved.

Is AI used in recruitment high-risk under the EU AI Act?

Yes. The Act classifies AI used for recruitment, candidate selection, performance evaluation, task allocation, worker monitoring, and promotion or termination decisions as high-risk (Annex III). High-risk does not mean banned — it means the system must meet obligations such as bias mitigation, human oversight, transparency and candidate disclosure, and registration in the EU database.

What are the penalties under the EU AI Act?

Non-compliance with high-risk operator obligations can draw administrative fines of up to EUR 15 million or 3% of total worldwide annual turnover, whichever is higher (Article 99). Using an outright-prohibited AI practice carries a higher tier of up to EUR 35 million or 7%. The 2026 delay deferred the compliance timeline, not the penalty amounts.

What did the Digital Omnibus not change?

The substance of the high-risk regime. Bias mitigation, human oversight, transparency, and registration all remain; only the date they apply moved to December 2027. The Act’s outright prohibitions have been in force since 2 February 2025 and are unaffected, and the fine structure is unchanged. The EU rescheduled the hiring rules, it did not reverse them.

Does the EU AI Act apply to companies outside the EU?

It can. The obligations attach to AI systems used in the EU and to employment decisions affecting people in the EU, so a non-EU employer using an AI hiring tool on candidates or workers located in a member state can be in scope. As with other EU employment rules, what matters is where the affected people are, not only where the company is based.

What should employers do during the delay?

Use the time, do not bank it. Inventory every AI tool that screens, ranks, scores, or monitors candidates and workers; press vendors on bias testing and documentation now; design genuine human oversight into automated decisions; prepare candidate-disclosure notices; and track the final text in the Official Journal. Retrofitting these into a live system later is far harder than building them in now.

EU AI Act Hiring Rules: 2027 | RemoteStaffingWiki

A Reprieve on the Clock, Not on the Rules

Companies that use AI to screen, rank, or manage candidates just got more time — but not a way out. The EU AI Act’s obligations for "high-risk" AI systems, the category that covers AI used in recruitment and employment decisions, were due to apply from 2 August 2026. Under the "Digital Omnibus on AI" simplification package, that deadline has been pushed to 2 December 2027. The European Parliament endorsed the change on 16 June 2026 and the Council of the EU gave its final green light on 29 June 2026 (Council of the EU).

The temptation is to read a delay as a reprieve and move on. That would be a mistake. The obligations themselves — bias testing, human oversight, candidate disclosure, registration — were not softened at their core; only the date they bite has moved. This piece explains what the AI Act actually requires of AI hiring tools, what the delay does and does not change, and why the sensible response is to keep building toward compliance rather than to shelve it.

It pairs naturally with the EU pay transparency explainer as part of the EU compliance calendar that anyone hiring into Europe now has to track, and it sits alongside the broader Remote Hiring Compliance 2026 overview.

Why Hiring AI Is "High-Risk" Under the Act

Annex III covers specified employment-related AI uses, including recruitment and worker-management decisions. Classification depends on the intended use and the Act’s conditions and exceptions; not every résumé-processing tool is automatically high-risk merely because it is used by HR.

High-risk does not mean banned — it means regulated. For systems in this tier, the Act imposes a set of obligations designed to make automated hiring decisions auditable and contestable rather than a black box.

Risk and conformity assessment, including bias mitigation — the provider must test the system and document that it meets the Act’s requirements before it goes to market.
Human oversight — a person must be able to understand, monitor, and override the system’s output, so hiring decisions are not fully delegated to the model.
Transparency and disclosure — deployers must be able to explain the system and, in employment contexts, inform the workers and candidates subject to it that a high-risk AI system is being used.
Registration duties depend on the actor and system under Article 49. Check which duties attach to the provider and which attach to a particular deployer rather than assuming every private employer has the same registration obligation.

The distinction between provider (who builds the system) and deployer (the employer who uses it) matters: buying an off-the-shelf hiring tool does not offload every obligation, because deployers carry their own duties around oversight, disclosure, and using the system as intended.

What the Delay Changes

The Digital Omnibus moves the application date for high-risk obligations from 2 August 2026 to 2 December 2027 — roughly sixteen extra months. For an employer mid-way through vetting an AI hiring vendor or building an internal tool, that is real breathing room to get conformity assessments, human-oversight processes, and disclosure practices in place without a hard summer-2026 wall.

The practical value of the extra time is mostly in the plumbing: running bias tests, wiring in a genuine human-in-the-loop rather than a rubber stamp, updating candidate-facing notices, and getting vendor documentation in order. None of that is quick, and the delay is best understood as a chance to do it properly rather than a signal that it no longer matters.

What the Delay Does NOT Change

Here is the part that gets lost in the "deadline delayed" headlines. The substance of the high-risk regime is intact.

The obligations themselves stand. Bias mitigation, human oversight, transparency, and registration are unchanged — only the date they apply has moved.

The penalties are unchanged. Non-compliance with high-risk operator obligations can draw administrative fines of up to EUR 15 million or 3% of total worldwide annual turnover, whichever is higher (AI Act, Article 99). The Omnibus deferred the timeline, not the fine structure.

The Act’s outright prohibitions are already live. The banned-practices provisions took effect on 2 February 2025 and are not part of this delay — for example, AI that infers emotions in the workplace is restricted regardless of the high-risk timeline.
The direction of travel is fixed. The EU has not reversed course on regulating hiring AI; it has rescheduled it. Any tool bought or built now will have to meet the same bar, just by December 2027.

What This Means If You Use AI to Hire

For companies hiring in or into the EU — including non-EU employers whose candidates or workers are in a member state — the reasonable posture is to use the extra time, not to bank it.

Inventory your AI. List every tool that screens, ranks, scores, or monitors candidates and workers, including features bundled into your applicant-tracking or HR platform. You cannot manage scope you have not mapped.
Press vendors on conformity now. Ask how a system was tested for bias, what documentation they will provide, and how they will support your deployer obligations. Build it into procurement while you still have leverage.
Design real human oversight. Ensure a person can review, understand, and override automated hiring decisions — and that this is a genuine checkpoint, not a formality.
Prepare candidate disclosure. Draft the notices that tell applicants and workers when a high-risk AI system is used in decisions about them.
Track the final amending text, applicable dates, and regulator guidance for the system’s actual classification. Keep a record of the source used for the compliance timetable.

The Bottom Line

The EU AI Act’s hiring obligations are now scheduled to apply from 2 December 2027 rather than 2 August 2026, after the Parliament and Council signed off on the Digital Omnibus in June 2026. But the requirements — bias testing, human oversight, candidate disclosure, registration — and the fines behind them (up to EUR 15 million or 3% of global turnover) are unchanged, and the Act’s outright prohibitions have applied since February 2025. The honest read is that the market got more time to comply, not permission to skip it. Employers who use the runway to build compliant, auditable hiring processes will be ready in 2027; those who treat the delay as a pass will simply face the same requirements later, with less room to maneuver.

RemoteStaffingWiki is an educational resource operated by LegelpTech Outsourcing Pvt Ltd, and is editorially independent. This article is general information, not legal advice. If you need a managed shortlist of remote staff, you can compare Zedtreeo as one provider option.

Council of the EU: AI Act legislative timeline

The EU AI Act’s Hiring Rules Just Slipped to 2027 — What Changed, and What Didn’t