The practice of collecting evidence that AI systems meet certain levels of performance, quality, safety, security and governance.