Oversight
The arrangement by which a human remains accountable for a system's behavior in the interval between a decision being taken and that decision being reviewed. The governed quantity is the interval, not the reviewer.
- StatusIndustry term
- AdoptionWidely used
- LayerGovernance
Why it matters
Applies to the named individual with authority to stop the system — not to a committee, which cannot be accountable because it cannot be fired. It bites continuously while the system is live. It fails when the review cycle is longer than the time in which loss accrues: an agent trading intraday under quarterly model review is unsupervised in every sense that matters, however complete the quarterly pack is. 'Human-in-the-loop' with no stated interval is not an oversight claim, it is a reassurance.
Where it comes from
MARQUE (a reference framework organized around eight questions) question 02. Carried in every published AI governance framework; the interval is almost never specified.
No published instrument defines this term yet; the note above is its origin.
What it is not
These are routinely confused with Oversight. The distinctions have consequences for how a system is governed.
- Monitoring
Monitoring observes. Oversight can intervene. A dashboard nobody is authorized to act on is monitoring.
- Governance
Governance is the whole apparatus. Oversight is the specific component operating between decision and review.
How it connects
MARQUE questions it addresses
- OversightWho is accountable in the interval between decision and review?
Related framework objectives
Connected through the MARQUE 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
- MS-2.9.2Interpretability and Decision ContextMeasure
- GV-1.3.4AI Use Risk Acceptance and Rejection FrameworkGovern
- GV-1.6.2AI Inventory and Data IntegrationGovern
Cite and reuse
Cite the identifier QIS-TERM-00026, 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-in-the-loop | Human-in-the-Loop | An association too weak or too general for a stronger verb. |
relatedTo | term:human-override | Human Override | An association too weak or too general for a stronger verb. |
relatedTo | term:duty-of-care | Duty of care | An association too weak or too general for a stronger verb. |
- Canonical identifier
QIS-TERM-00026- Tier
- Canonical industry term
- Domain · layer
- MARQUE questions · Governance
- Semantic aliases
- None recorded
- First published
- 2026-08-02
- Last reviewed
- 2026-08-02 · 180-day cycle