Rules and limits placed on an AI agent to keep its behavior inside acceptable bounds.
Guardrails set the boundary; oversight checks whether the agent actually stays inside it over time.
Related: Agent Policy & Guideline Enforcement, AI Agent Governance