GLOSSARY · LEADERSHIP & GRC
Human-in-the-loop
Human-in-the-loop is a design pattern in which a person reviews, approves, or can override an AI system's output before it takes effect.
It is the standard mitigation for hallucination and excessive agency, and several regulations expect human oversight for consequential decisions. It only works when the reviewer has the time, context, and authority to say no.