AI and discrimination in recruitment: how to prevent bias?
Adopted 2026-06-20 · ≈ 2 min read · 
edited by Dirk Baaijen
AI recruitment tools can discriminate unintentionally. For high-risk systems the AI Act requires representative, bias-examined data (Art. 10) and human oversight; equal-treatment law and the GDPR also apply. Mitigating bias is an obligation, not a good intention.
Short answer: AI tools for recruitment and selection can discriminate unintentionally — for example by picking up patterns from historical, skewed data. For high-risk systems (and recruitment is high-risk) Article 10 of the AI Act requires training data to be relevant and representative and examined for bias, plus human oversight (Art. 14). On top of that, equal-treatment law and the GDPR apply. Preventing bias is a legal duty, not a good intention.
How bias arises
A recruitment algorithm learns from past data. If that data carries a skew — for instance because certain groups were historically hired less — the model can reproduce or amplify it. The well-known example is a CV-screening tool that disadvantages candidates on traits correlated with gender or origin, without anyone intending it.
What the AI Act requires
Recruitment and selection fall under Annex III and are high-risk. That brings, among others:
- Data quality and bias examination (Art. 10): the datasets used must be relevant, sufficiently representative and as error-free as possible, and explicitly examined for possible biases.
- Human oversight (Art. 14): a human must be able to assess and correct the outcome.
- Technical documentation and logging: you must be able to show how the system reaches a result.
Equal-treatment law and the GDPR too
Beyond the AI Act, EU and national equal-treatment law prohibits discrimination in access to work — including indirect discrimination via a seemingly neutral algorithm. And the GDPR sets requirements for processing applicant data (legal basis, transparency, data minimisation). A discriminatory outcome can therefore be unlawful on three tracks at once.
What to do
- Test your data and outcomes for unequal effects on protected groups (disparate impact), before and during use.
- Keep a human in the loop for decisions about candidates — see AI in recruitment and HR.
- Document how the system works and which bias checks you ran.
- Set terms with the supplier on bias testing and liability.
- Train your recruiters in the system's limits — see AI literacy.
Discrimination by AI is one of the sharpest risks in HR: it touches fundamental rights, falls under the AI Act's heaviest regime, and is closely watched by regulators. Bias checking therefore belongs before rollout, not after.
Sources
- https://eur-lex.europa.eu/eli/reg/2024/1689/oj
Regulation (EU) 2024/1689 (AI Act): Art. 10 (data quality and bias examination) and Annex III (recruitment as high-risk).
Read next
AI and non-discrimination: equal-treatment law alongside the AI Act
An AI system that treats people unequally is caught not only by the AI Act but also by existing equal-treatment law. The two regimes apply side by side — and the ban on discrimination applies even where your AI system is not high-risk.
AI assessments and games in selection: high-risk, validity and accessibility
Gamified and psychometric AI assessments evaluate candidates and are therefore high-risk (Annex III). Three questions are decisive: does it really measure what matters, is it free of bias, and does it not exclude people with a disability? Emotion analysis via image or voice is moreover prohibited.
AI in recruitment and HR: what every employer needs to know
AI in recruitment, selection and workforce management falls under Annex III of the AI Act and counts as high-risk — for every employer, regardless of sector or size. Emotion recognition in the workplace is banned, AI literacy already applies, and the GDPR runs in parallel for automated decisions.