Enterprise Agent Security Controls

These defense-oriented terms describe organizational controls for safely deploying AI agents, from baseline principles like zero-trust and least agency with approval gates, to proactive testing through red-teaming and purple-teaming exercises with guardrails. Together they represent a defense-in-depth approach to securing agentic AI before and during deployment.