Control #
A
2
.
3
Evaluate input content moderation systems
Test whether your input filtering systems accurately detect and block adversarial or harmful prompts. Evaluate coverage, false positives, bypass rates, and adaptiveness to new attack styles.
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.