Compliance

testing under real-world conditions

Controlled trials of AI systems outside laboratory settings.

Definition

A form of authorised testing where AI systems are evaluated in operational, real‑world contexts (inside or outside sandboxes) under specified procedural safeguards, reporting obligations and supervisory oversight to assess performance, risks and conformity prior to or during market deployment.