← Concept IndexDEFINITION WHY IT MATTERS COMMONLY CONFUSED WITH SOURCES
Tool & permission scoping
Also called: least privilege, what the agent can touch
Deliberately limiting which tools an agent has and what each can access — read-only where possible, one mailbox not all of them, no ability to delete — so a mistake or injection has a small blast radius.
It's the most effective agent safeguard because it doesn't rely on the model behaving. Least privilege caps the worst case regardless of what the model is tricked into trying.
Trusting the model less in the prompt. Scoping is enforced outside the model, so it holds even when the prompt defence fails.
- Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1U.S. National Institute of Standards and Technology · 2023-01-26
First-pass citations, limited to primary sources; a reviewer will broaden and verify these before this entry leaves draft.