Definition
TEVV is the structured process of checking that an AI system actually works as intended, stays within acceptable risk limits, and is fit for its intended use — the AI equivalent of quality assurance and safety testing combined. NIST's new draft framework (AI 200-2) is the first detailed government methodology for doing this consistently.
Why it matters
Without rigorous TEVV, organizations have policies and risk registers but no real evidence their AI systems are safe to deploy — TEVV is what turns AI governance from paperwork into proof.