EU AI Act for HR & Recruitment: Hiring AI Compliance

Which HR AI Is High-Risk

Annex III point 4 covers employment and worker management in two parts:

• Point 4(a) — recruitment and selection. AI used to place targeted job advertisements, filter or screen applications, and evaluate candidates in the course of interviews or tests is high-risk.
• Point 4(b) — the employment relationship. AI used to make or materially inform decisions on promotion and termination, to allocate tasks based on behaviour or personal traits, and to monitor and evaluate performance and behaviour is high-risk.
• Breadth. This is one of the most clearly enumerated categories in Annex III — CV parsers, interview-analysis tools, gig-work task allocators, and performance-scoring systems are all in scope.

The classification does not depend on whether a human makes the final call — a tool that materially informs a hiring or firing decision is still high-risk.

Interplay: Worker Consultation, NYC LL144 & the EEOC

Hiring AI sits at the intersection of the AI Act, EU labour law, and — for US operations — bias-audit rules.

• Worker information and consultation. Before deploying a high-risk system at work, deployers that are employers must inform affected workers and their representatives (Article 26(7)), and national labour law and works-council consultation duties apply on top.
• NYC Local Law 144. US firms using automated employment decision tools for candidates or employees in New York City must commission an independent annual bias audit and publish a summary — a distinct obligation that runs in parallel to the AI Act.
• EEOC and US anti-discrimination law. Title VII, the ADA, and EEOC guidance on algorithmic tools mean a screening system can be lawful under one regime yet expose the employer under another — a single bias-aware audit trail helps you answer to both.

Run one governance program that satisfies the strictest applicable rule rather than a separate scramble per jurisdiction.

Core Obligations for Hiring AI

For a high-risk employment system, the deployer-facing duties matter most:

• Article 14 — human oversight. A person must be able to review, override, and disregard the AI's ranking or recommendation; the hiring decision cannot be effectively automated away.
• Article 13 — transparency to deployers. The provider must give HR teams enough information to interpret the tool's output, understand its limitations, and use it correctly.
• Article 10 — data governance. Training and evaluation data must be examined for bias so the tool does not encode discrimination against protected groups.
• Article 12 — logging. Automatic records let you reconstruct why a candidate was screened out — the evidence a bias audit or a discrimination claim will demand.