Governance, Law & Compliance
Human oversight
Meaningful human supervision of an AI system's decisions or actions.
Example
A qualified person can review, challenge, and reverse a high-impact automated decision.
Why it matters
It keeps consequential AI actions subject to accountable human judgment.
You'll hear it
“Can the reviewer actually stop this, or only watch it happen?”
Can you control it?
Yes
Yes. An organization can put human oversight in place through its policies, processes, records, training, or oversight structure. Individual users may participate without controlling the entire program.
Common questions
- Does oversight require checking every result?
- Not always. The appropriate review depends on the risk and the safeguards around the system.
- Can reviewers become too trusting over time?
- Yes. Routine agreement with AI can reduce attention, so training and checks remain useful.
- What if the system acts too quickly for a person to react?
- Use limits or approval before the action where after-the-fact intervention would be too late.