Why recruiting AI is classified as "high-risk"
Regulation (EU) 2024/1689 — the AI Act — ranks AI systems by risk level. Employment is in scope: Annex III, point 4 explicitly covers AI used to screen, filter applications and evaluate candidates. These uses are therefore 'high-risk'.
The reason is simple: a hiring decision has a direct impact on people. Lawmakers want to prevent an opaque AI from rejecting candidates on criteria that can be neither explained nor challenged.
The deadlines that matter
The AI Act entered into force in 2024, but its obligations apply in stages. Prohibited practices have applied since February 2025. Regulation (EU) 2026/1744 moved the application of Chapter III, Sections 1 to 3 for Annex III high-risk systems — including recruiting — to 2 December 2027.
The delay does not remove the high-risk classification or the value of acting now: documentation, human oversight, candidate information and testing take time.
Provider vs deployer: who carries what?
The AI Act draws two roles. The "provider" builds the AI system (here, the software vendor). The "deployer" uses it in their hiring (the recruiting company).
The provider carries the heaviest obligations: technical documentation, risk management, robustness and bias testing, transparency. The deployer must keep meaningful human oversight, inform candidates and never base a decision on the machine score alone.
As a recruiter, choose a provider that publishes both the controls already in place and the work still open, rather than relying on a blanket compliance claim.
Four habits to recruit with compliant AI
1. Keep a human in the loop: no application should be rejected automatically. The score is a reading aid, not a verdict.
2. Demand transparent criteria: you must be able to explain what the AI evaluates and with which weightings.
3. Inform candidates: applicants must know an AI takes part in analysing their file, and be able to request a review.
4. Check for bias: the provider must be able to show that equivalent profiles get equivalent scores, regardless of name, age or origin.
How aiKip approaches it
aiKip is built around these constraints, not against them. The AI score is explicitly "indicative" and no stage change is automatic: the recruiter stays in control. Criteria and weightings are public (skills 35%, experience 35%, education 15%, motivation 15%) and recomputed server-side.
An evaluation set regularly replays identical profiles with different names to surface any bias. And a public transparency page explains to candidates how the analysis works and what their rights are. The goal: AI you can audit, not take on faith.
Frequently asked questions
Does the AI Act ban AI in recruiting?
No. It does not ban it — it frames it. AI screening and evaluation of applications is "high-risk", which imposes obligations (human oversight, transparency, bias testing), but the use remains allowed if compliant.
Am I liable if I use an AI scoring tool?
As a "deployer" you have your own obligations: keep human oversight, inform candidates and never decide on the score alone. A transparent provider makes those duties easier to evidence.
What happens on 2 December 2027?
Chapter III, Sections 1 to 3 become applicable to Annex III high-risk AI systems, including recruiting, under the timetable amended by Regulation (EU) 2026/1744.