Glossary
Human Oversight
Arrangements that keep people able to understand, intervene in, or override an AI system’s operation where required.
Last reviewed 2026-08-22
In plain language
Human oversight appears throughout ethics guidelines and in the EU AI Act’s obligations for certain systems. Forms include human-in-the-loop, human-on-the-loop and human-in-command. Oversight fails when operators lack time, authority, information or incentives to disagree with a model. The presence of a human click is not the same as meaningful control.
Why it matters for AI governance
Policymakers often treat human oversight as a safeguard. Design quality, training, liability and workplace pressure determine whether it works in practice.
Authoritative sources
- Regulation (EU) 2024/1689 (AI Act) (opens in a new tab) — European Union