NIST AI Risk Management Framework — execution-evidence crosswalk

Source version: NIST AI 100-1, AI RMF 1.0 (January 2023)
Crosswalk version: nist-ai-rmf@1.0.0
Source authority: National Institute of Standards and Technology (NIST)
Last reviewed: 2026-08-13

What NIST AI RMF is

A voluntary framework organised around four functions — Govern, Map, Measure, Manage — for managing risks of AI systems.

What AIEF is

AIEF is an independent, implementation-agnostic framework for AI execution integrity. It defines the evidence properties an AI or agent execution must have to be reconstructed, attributed and independently verified after the fact — execution artifacts, integrity protection, version pinning, retention and independent validation.

Where they overlap

Outcomes concerning documentation, traceability, accountability, third-party dependency awareness and post-deployment monitoring depend on records of what a system actually did.

Where they do not

AIEF does not address fairness, bias, model quality, explainability, safety, security posture generally, environmental impact or broader societal harm — all of which the AI RMF treats as first-class concerns.

Control mapping — 8 AIEF controls, 9 mapped references

Relationship strength is stated explicitly and deliberately conservative. Where a relationship is uncertain it is downgraded rather than overstated. Provisions with no credible execution-evidence relationship are left unmapped.

AIEF control to NIST AI Risk Management Framework execution-evidence crosswalk
AIEF control	External reference	Relationship	Rationale	Limitation
AIEF-01Execution Artifact Completeness	MEASURE 2.8Risks associated with transparency and accountability are examined and documented	Supporting relevance	Examining transparency and accountability risks depends on being able to say what a system actually did. Structured execution records make that determinable rather than inferred.	AIEF supplies evidence of execution. It does not examine or document transparency and accountability risk, which remains an organisational activity.
AIEF-01Execution Artifact Completeness	MANAGE 4.1Post-deployment monitoring plans are implemented	Supporting relevance	Post-deployment monitoring consumes operational records. Automatic, structured execution records are a reliable source for that monitoring.	AIEF neither creates a monitoring plan nor evaluates whether monitoring detects the risks that matter.
AIEF-04Version Preservation and Context Pinning	MEASURE 2.8Transparency and accountability — traceability of system behaviour over time	Supporting relevance	Accountability degrades when historical behaviour cannot be interpreted against the versions and configuration in force at the time. AIEF-04 preserves that context in the artifact.	This is narrow technical traceability, not model transparency, interpretability or explainability.
AIEF-06External Dependency Evidence / Tool Calls	MAP 4.1Approaches for mapping risks of third-party technology and data are in place	Supporting relevance	Mapping third-party risk requires knowing which external dependencies actually influenced an outcome. AIEF-06 records those calls in the evidence.	AIEF-06 identifies dependencies that ran; it does not assess or rate the risk those dependencies carry.
AIEF-08Retention, Portability and Offline Verification	GOVERN 1.5Ongoing monitoring and periodic review of the risk management process	Supporting relevance	Periodic review across time requires records that outlive the running system. Retained, portable evidence makes retrospective review possible.	AIEF does not define the review cadence, scope, governance structure or the risk management process being reviewed.
AIEF-05Independent Validation Capability	GOVERN 4.2Organisational teams document risks and impacts of the technology they design and deploy	Related consideration	Documented claims about system behaviour become materially more credible when the underlying evidence can be validated by someone other than the team that produced it.	Independent verifiability of evidence is not the same as independent review of risk documentation.
AIEF-07Multi-Step Chain Integrity	MEASURE 1.3Internal experts and independent assessors are able to evaluate system performance	Related consideration	Evaluating multi-step or agentic behaviour requires the sequence of actions to be preserved and its integrity checkable, not reassembled from unrelated events.	AIEF assesses chain integrity, not the correctness, quality or safety of the behaviour the chain records.
AIEF-02Tamper-Evidence	MANAGE 2.4Mechanisms are in place and applied to supersede, disengage or deactivate systems that demonstrate performance inconsistent with intended use	Contextual only	Acting on inconsistent performance presumes the records evidencing it are trustworthy and have not been altered after the fact.	AIEF-02 concerns detectability of evidence tampering only, and says nothing about intervention or deactivation mechanisms.
AIEF-09Privacy, Minimization and Redaction Controls	MAP 5.1Likelihood and magnitude of impacts, including privacy considerations, are understood	Contextual only	Capturing execution evidence can itself create privacy exposure. AIEF-09 requires minimisation and redaction that does not silently break integrity guarantees.	AIEF-09 does not assess privacy impact, lawful basis or data subject rights.

Strong supporting relevance: The AIEF capability produces evidence of the kind the external provision is concerned with. It does not satisfy the provision.
Supporting relevance: The AIEF capability may help provide relevant evidence, but does not itself satisfy the external requirement.
Related consideration: A material conceptual relationship. Satisfying one does not imply satisfying the other.
Contextual only: Useful context only. No evidentiary claim is made.

Limitations and source

The AI RMF can evolve. This crosswalk is pinned to AI RMF 1.0 and must be re-reviewed against any successor edition rather than assumed equivalent.

Authored against National Institute of Standards and Technology (NIST) — https://doi.org/10.6028/NIST.AI.100-1 (NIST AI 100-1, AI RMF 1.0 (January 2023)). Crosswalk nist-ai-rmf@1.0.0, reviewed 2026-08-13.

Run an AIEF assessment