What it means
Human oversight is the active role people play in watching, reviewing, and intervening in what an AI agent does. It can mean approving an action before it happens, checking outputs while the agent works, or pausing and overriding the agent when something looks wrong. Good oversight is designed, not assumed: a named person, with the time, information, and authority to intervene, and a clear way to stop the agent.
Why it matters for governance
It matters for governance because it keeps a person accountable for consequential actions instead of letting the agent be its own final reviewer.
Example
An operator sees an agent about to send a message that does not match company policy, pauses the agent, reviews the case, and sends a corrected reply instead.
Related ActLattice products
General information only — not legal advice.