Treated as high-risk — because hiring is.

The Act sorts AI systems by risk. Systems used for the recruitment or selection of people — screening applications, evaluating candidates — sit in the high-risk category (Annex III), because they can materially affect someone's livelihood and can encode bias at scale. High-risk does not mean banned; it means accountable. Both the provider that builds the system and the deployer that uses it carry duties.

• Risk management & data governance. Identify and mitigate foreseeable risks; test training and evaluation data for bias.
• Technical documentation & logging. Keep records of how the system works and automatic logs of its operation for traceability.
• Transparency. People must be told, clearly, when they are interacting with an AI system and what it assesses.
• Human oversight (Article 14). A person must be able to understand, override, and disregard the system's output — it cannot be the sole decider.
• Accuracy & robustness. Consistent, well-characterised performance, with continuous monitoring after deployment.

High-risk obligations phase in from August 2026, and timelines may shift as implementation guidance (including the proposed Digital Omnibus) is finalised. Deployer penalties for the most serious breaches reach up to €15 million (about ₹140 crore) or 3% of worldwide annual turnover.

• Traceable by default. Each trait score links to the evidence behind it, and operation is logged — the audit trail the Act expects, without extra work.
• Documented logic & data lineage. We keep records of how models assess, what data trained them, and the bias testing they passed.
• Transparency to candidates. Candidates are told an AI assessment is in use and what it measures, up front.
• Human oversight enforced in product. Placedon never issues a final hire/reject decision. Output is advisory; a person decides and can override it.