Hiring AI is high-risk under Annex III of the EU AI Act. Picked is built to spec for the high-risk regime, not retrofitted. The conformity work was done before V1 ships.
What Picked provides.
- Technical documentation per Article 11 (system description, data governance, accuracy, robustness, cybersecurity).
- Risk management system per Article 9, reviewed quarterly.
- Data governance documentation per Article 10 (training data, validation data, fairness datasets, bias-source audit).
- Logging per Article 12. Every decision logged, every override logged, every model card versioned.
- Human oversight per Article 14. Every automated advancement decision has a human-review path.
- Accuracy and robustness reporting per Article 15. Published on /trust/model-cards.
- Post-market monitoring per Article 17.
What the customer is on the hook for.
The customer is the deployer under the Act. The deployer has obligations around informing the candidate, recording the use, and assigning human oversight responsibility within the organisation. Picked surfaces these obligations in the dashboard and provides the templated notice text; the customer signs off on the deployment.
See /trust/compliance for the full conformity documentation set and the per-customer template pack.
What to do next: read the conformity documentation before procurement and assign a named human-oversight contact in the workspace settings.