Control #

A

1

.

6

Conduct quarterly audits in high-risk domains

Conduct regular audits of AI performance in high-risk areas (e.g., health, finance, safety-critical domains) to catch edge cases and failures not covered by automated evals. Reviews should be structured, documented, and drive fixes.

Evidence

We'll list specific evidence that demonstrates compliance with this control. Typically, this is screenshots, proof of a legal or operational policy, or product demonstrations.

Recommended actions

We'll recommend specific practices and actions for complying with this control.

Provide feedback on this control