NIST AI RMF 1.0, MP-2.2

Knowledge limits and oversight information

Unit OH-45 in the Impact theme. The held citation is NIST AI RMF 1.0, MP-2.2, and the row asks for 3 pieces of evidence.

At a glance

Ref: OH-45
Theme: Impact
Held source: NIST AI RMF 1.0
Citation: NIST AI RMF 1.0, MP-2.2
Evidence: 3 items
Gaps: 2 common

Evidence to keep

Knowledge limits and oversight informationDocumented knowledge limits
Knowledge limits and oversight informationGuidance on permitted use and interpretation
Knowledge limits and oversight informationThe human oversight expected over the output, and by whom

Questions before you sign

What was the system not built for?
Does the operator have that in writing?
What should they do when the input is outside it?

The requirement, in our words

Document the system's knowledge limits and how its output may be used and overseen by humans, in enough detail for a downstream actor to make an informed decision.

What it means for human oversight

The limit is the most important thing the overseer can know. Document the conditions the system was not built for, and deliver that with the system rather than leaving it with the build team.

How to evidence it

Attach the artefacts above to the Knowledge limits and oversight information row with the system name, the owner and the date. The first one is the anchor: Documented knowledge limits. The OH-45 file is complete when the named person can produce that artefact and explain the decision it records, without asking someone else where it is.

The reviewer will sample one live decision and ask for Guidance on permitted use and interpretation. If that artefact for OH-45 cannot be produced, the row is a gap, whatever the policy says.

OH-45 carries 3 evidence items and 2 common gaps. For knowledge limits and oversight information, link the row to the assessment page that names the harm, because an overseer cannot watch for a consequence the file never describes. The OH-45 statement above is ours, written from NIST AI RMF 1.0, MP-2.2, not a reproduction of the licensed text.

Common gaps

Limits known to the build team and absent from what the deployer receives
Output guidance silent on out-of-distribution input

Where the FDIS fits. The public abstract says the draft applies to all types of organizations across the AI system life cycle, so knowledge limits and oversight information is one of the places the guidance is expected to operate. For OH-45, the FDIS clause numbering and wording are not public, so this page cites NIST AI RMF 1.0, MP-2.2 above and does not quote the draft.

Related units

Source for this page: NIST AI RMF 1.0, MP-2.2. Open NIST AI RMF 1.0, MP-2.2 and compare it with the OH-45 statement above. Retrieved 2026-10-21.