Recruitment AI compliance under the EU AI Act

Recruitment technology sits at the heart of the EU AI Act's high-risk provisions. Any AI system used to screen, rank, filter, or score candidates for employment is explicitly listed in Annex III Category 4. This covers CV parsing tools, automated interview analysis platforms, skills assessment engines, and any scoring model that influences which candidates a human recruiter sees. If you build or sell recruitment AI and it is used within the EU, you are building a high-risk AI system with obligations that come into full effect on December 2, 2027. The good news is that you have time to prepare, but the documentation and technical requirements are significant enough that starting now is not optional.

What the EU AI Act requires

High-risk recruitment AI systems must comply with Articles 9 through 17. Article 9 requires an ongoing risk management process that identifies potential harms to candidates, including discrimination, and puts controls in place. Article 10 is central to recruitment AI: your training and validation data must be scrutinised for bias across protected characteristics including gender, age, ethnicity, and disability status. If your model was trained on historical hiring data, you must assess whether that data encodes existing workplace discrimination and take corrective action. Article 11 requires comprehensive technical documentation. Article 13 requires that recruiters using your tool can understand why a candidate was scored or ranked as they were, with sufficient granularity to make a meaningful independent judgment. Article 14 requires that human oversight is not merely nominal: the Act explicitly states that humans must be able to understand the AI system's capabilities and limitations and to disregard, override, or reverse the output.

What this means for your business

A recruitment AI that presents a ranked shortlist without explanations does not meet Article 13. An AI video interview analysis tool that scores candidates on vocal or facial features needs extensive bias testing and documentation under Article 10, as well as genuine human review under Article 14. Enterprise customers in regulated industries will increasingly require supplier documentation as part of their own compliance programmes, making EU AI Act readiness a sales prerequisite rather than a regulatory afterthought. Non-compliant recruitment AI tools face both enforcement risk and commercial displacement by compliant competitors.

Steps to get compliant

1. Conduct a bias audit of your training data and model outputs. Document the methodology, the demographic groups tested, the error rates found, and any steps taken to mitigate disparate impact, as required by Article 10.
2. Build explainability into candidate scoring. Ensure every score or ranking presented to a recruiter includes a plain-language breakdown of the contributing factors, meeting Article 13 transparency requirements.
3. Design and document human oversight workflows. Define exactly how a recruiter overrides or disregards an AI recommendation, train users on this process, and include it in your instructions for use under Article 14.
4. Begin technical documentation. Start building the Article 11 technical file now, covering system architecture, training methodology, performance metrics, known limitations, and risk management measures.

Free EU AI Act risk assessment

Not sure where your AI system stands? The free ActComply risk screener classifies your system in under 5 minutes. No sign-up required.