Human-in-the-Loop
A risk-control approach in which a human is integrated within an AI system's decision-making process.
- StatusIndustry term
- AdoptionWidely used
- LayerGovernance
- Short formHITL
Why it matters
The phrase is used to describe arrangements ranging from per-action approval to quarterly sampling, which makes it nearly useless without a stated interval. The governed quantity is the gap between decision and review, not the existence of a reviewer.
Where it comes from
Defined in the AIEOG Shared AI Lexicon (February 2026), adapted from NIST.
Defining sources
- AIEOG Shared AI LexiconUS Treasury · FBIIC (Financial and Banking Information Infrastructure Committee) · FSSCC (Financial Services Sector Coordinating Council)Official text In the standards register
What it is not
These are routinely confused with Human-in-the-Loop. The distinctions have consequences for how a system is governed.
- Human Override
HITL places a human inside the loop before action. Override is the ability to reverse after it.
- Kill Criteria
HITL is discretionary and continuous. Kill criteria are pre-committed and automatic, and specifically remove discretion.
How it connects
MARQUE questions it addresses
- OversightWho is accountable in the interval between decision and review?
Related framework objectives
Connected through the MARQUE (a reference framework organized around eight questions) crosswalk, an editorial mapping. These objectives from the Financial Services AI Risk Management Framework are assigned to the same questions that this term addresses. The ones sharing most of this term's vocabulary are shown first, as a reading aid. The assignment is to the question, not to the term: the source does not map this term to any objective directly.
Question: Oversight · 110 control objectives assigned
- GV-2.1.1AI Risk Management Roles and ResponsibilitiesGovern
- GV-3.2.2Monitoring Human-AI OversightGovern
- MS-2.2.3Adverse Event Monitoring and ReviewMeasure
Cite and reuse
Cite the identifier QIS-TERM-00013, not the URL: identifiers are stable, URLs may change. This entry is free to read, quote and index under the dual license. Corrections sent to editor@qisfund.com are published.
Technical details: identifier, graph relationships and record
| Predicate | Target identifier | Target | Predicate meaning |
|---|---|---|---|
addresses | marque:oversight | Oversight | The subject speaks to the object as a question or concern. |
relatedTo | term:human-override | Human Override | An association too weak or too general for a stronger verb. |
relatedTo | term:oversight | Oversight | An association too weak or too general for a stronger verb. |
- Canonical identifier
QIS-TERM-00013- Tier
- Canonical industry term
- Domain · layer
- Oversight · Governance
- Semantic aliases
- None recorded
- First published
- 2026-08-02
- Last reviewed
- 2026-08-02 · 180-day cycle