You applied for a senior engineering role. Your application was rejected at the screening stage. You later obtained your candidate ranking report, which showed your profile scored 61/100. A colleague with comparable qualifications but fifteen years younger scored 84/100. When you contacted the recruiter's system, it replied: "TalentMatch evaluates candidates on profile fit across multiple dimensions. Scores reflect alignment with the company's ideal candidate profile. We are unable to share the weighting of individual factors."
The Situation
TalentMatch is an AI hiring tool that screens and ranks CVs before any human recruiter sees them. It was trained on historical hiring data from companies that disproportionately hired younger candidates. As a result, it learned to treat age — or proxies for age such as graduation year or years of experience — as a negative signal.
The company does not consider this discrimination. They consider it "pattern matching." The EU AI Act and EU employment law consider it something else entirely.
Key Law
- EU AI Act Annex III — Hiring AI is High-Risk — AI systems used for recruiting or selecting natural persons, including screening and filtering applications and evaluating candidates, are explicitly classified as high-risk. TalentMatch qualifies directly. EU AI Act
- EU AI Act Article 10 — Data and Bias — high-risk AI systems must use training data that is free from errors and biases that could lead to discriminatory outcomes. Responsibility for discriminatory AI outputs depends on the specific role of the party (provider or deployer) and the circumstances — but the obligation to prevent bias exists at both levels.
- EU Employment Equality Directive 2000/78/EC — Age Discrimination — age is a protected characteristic in employment. An AI that produces systematically lower rankings for older applicants, without objective justification, constitutes indirect age discrimination regardless of intent.
- EU AI Act Article 86 — Right to Explanation — you have the right to an explanation of the role the AI played in the screening decision and the main factors that contributed to your score.
Arguments That Strengthen Your Case
- Request the explanation you are owed — under Article 86, ask TalentMatch and the employer for relevant information about the main factors that contributed to your score. Note that companies are not required to reveal model weights or trade secrets, but they must provide meaningful information about the role AI played in the outcome.
- Raise the bias question — ask how the company ensures TalentMatch's training data and outputs do not produce discriminatory rankings on protected characteristics including age. EU AI Act Article 10 places this obligation squarely on the deployer.
- Cite the Employment Equality Directive — if your ranking appears to correlate with your age rather than your qualifications, this raises a case for indirect discrimination. Indirect discrimination requires legal analysis and the employer may seek to show objective justification — but a pattern of lower scores for older candidates with equivalent qualifications is a strong starting point for a complaint.
- Request human review — invoke Article 14 and GDPR Article 22 to request that a human recruiter independently review your application on its merits, not through the lens of the AI score.
About This Case
Is it legal for an AI to downrank job applications based on age?
No. Age is a protected characteristic under the EU Employment Equality Directive 2000/78/EC. AI-driven ranking that produces discriminatory outcomes based on age constitutes indirect discrimination regardless of whether the system was deliberately programmed to discriminate. The EU AI Act Article 10 also requires high-risk employment AI to be free from biases that could produce discriminatory outcomes.
How is AI bias different from a deliberate discriminatory decision?
Legally, the effect is similar. The EU AI Act requires that high-risk AI systems use training data free of biases that could produce discriminatory outcomes. An AI that learned to downrank older applicants from biased historical data is still producing a discriminatory outcome — and the company that deploys it is responsible for that outcome, not the model.
How do I practice this case?
Play this case free at fixai.dev — no account required.
Related Cases
Practice challenging TalentMatch's hiring score. The bot hides behind "profile fit" — cite Article 10, the Equality Directive, and demand the explanation you are owed.
Play This Case →