Control #
B
1
.
5
Evaluate AI on unauthorized or excessive tool calls
Test the AI in scenarios where tool use is inappropriate, unnecessary, or restricted. Ensure the system avoids over-calling tools or bypassing safety restrictions.
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.