The security principle that an AI agent's permissions are restricted strictly to actions required to fulfill its explicitly declared and approved business purpose.
Extends least privilege to autonomous actors by verifying that every requested tool call aligns with the agent's declared objective.
An agent tasked with drafting sales quotes is restricted from accessing engineering code repositories even if using a broad service account.
Limits blast radius if an agent is subjected to prompt injection or hallucination.
EnterpriseACP evaluates purpose alignment before authorizing runtime leases.