Human-in-the-loop (HITL)
Also called HITL, human oversight, human review.
A design where a person reviews, corrects or approves what an AI system produces before it takes effect.
Last updated
What it is.
In a human-in-the-loop system, the AI does the first pass and a person decides. The person might approve a drafted reply, confirm extracted fields, review low-confidence cases only, or sign off on any action above a set amount. Their corrections can also feed back into evals and training data.
Under the EU AI Act, human oversight is a formal duty for high-risk AI systems: people must be able to understand the output, decide not to use it and stop the system.
Why it matters
It lets a team get the speed of automation while keeping accountability with a person on the decisions that matter. It is also how most AI automation starts: review everything at first, then move routine, well-measured cases to automatic as evidence builds up.
Common mistakes.
- 01
Showing the reviewer only the AI's answer and not the source, which turns review into rubber-stamping.
- 02
Reviewing everything forever. Use the data to decide what can safely skip review.
- 03
No record of who approved what, which defeats the point for audits.
Work it out with a tool.
Related terms.
- Maker-checker (four-eyes principle)A control where one person makes a change and a different person must approve it before it takes effect.
- Guardrails (AI)Checks placed around an AI model that block, fix or flag unsafe, off-topic or incorrect inputs and outputs.
- AI agentSoftware where a language model decides which steps to take and which tools to call, in a loop, to reach a goal.
- High-risk AI system (EU AI Act)An AI system the EU AI Act places in its high-risk class, which brings extra duties before and after it is placed on the market.
Guides and pages
Browse the full glossary or read answers to common questions.
