Approval Gating Glossary
The control pattern that requires human consent before an agent performs sensitive, external, costly, or destructive actions.
Approval gating is the practice of placing explicit human consent in front of actions an AI agent should not take on its own: sending messages, spending money, deleting files, changing production systems, exposing private data, or making irreversible commitments.
In a chatbot, approval is often informal: the user asks, the model answers. In an agentic system, the model can act through tools. That makes the difference between suggestion and execution much more important. Approval gates turn I could do this into I will only do this after the human authorises it.
Approval gating records where the agent’s authority ends and the human’s begins. A system without that boundary can quietly turn a recommendation into an action.