Home Blog Mapping AI Controls to NIST AI RMF and ISO 42001

Mapping AI Controls to NIST AI RMF and ISO 42001

Q: Does Vaikora provide a NIST AI RMF mapping?

Yes. Vaikora provides a crosswalk that maps NIST AI RMF subcategories across Govern, Map, Measure, and Manage to specific platform capabilities. Each mapping includes a coverage level and a defined evidence artifact that GRC teams can use directly in a NIST profile.

Q: Can I use this crosswalk for both frameworks at once?

Yes. The crosswalk includes a combined mapping that shows how individual Vaikora capabilities satisfy both NIST AI RMF subcategories and ISO 42001 Annex A controls. This allows organizations to support dual-framework programs without duplicating mapping work.

Q: What is the difference between Full, Partial, and Reference coverage?

Full coverage means Vaikora is the primary control owner and generates the required evidence directly. Partial coverage means Vaikora provides a technical component while other elements are handled by broader organizational controls. Reference coverage means Vaikora contributes supporting evidence for controls owned elsewhere, such as governance or training processes.

Q: Is this a substitute for an audit?

No. The crosswalk is an evidence-mapping tool to support audits, not a certification. Formal certification, such as ISO 42001, requires assessment by an accredited auditing body against the full organizational program.

Q: How often is the crosswalk updated?

The crosswalk is updated when frameworks are revised or when platform capabilities change in ways that affect evidence mapping. Each version is tracked so GRC teams can reference a specific revision in their compliance documentation.

Q: How does this connect to the rest of the AI security program?

The crosswalk links directly to operational telemetry. Each evidence entry maps to audit log fields, policy engine outputs, or SIEM events. This means the same data used to operate security controls also provides the documentation needed for compliance.

AI Runtime Control, Real-time AI Security, Threat Intelligence, Vaikora

April 25, 2026

This is a working crosswalk that maps NIST AI RMF 1.0 functions (Govern, Map, Measure, Manage) and ISO/IEC 42001:2023 controls to concrete Vaikora capabilities — policy engine, 7-factor risk scoring, SHA-256 hash-chained audit, content-free logging, and automated reporting. CISOs are increasingly asked to demonstrate AI control coverage against both frameworks; most teams find no off-the-shelf mapping. The full crosswalk is published as a downloadable XLSX so a GRC team can drop it into evidence packages and prospect deliverables. The blog version below highlights the highest-cited rows from each function and explains the design choices behind the mapping.

Why Both NIST AI RMF and ISO/IEC 42001?

NIST AI RMF 1.0 (released January 2023) is a voluntary framework structured around four functions — Govern, Map, Measure, Manage — with 72 sub-categories underneath. It is the framework most U.S. enterprise programs cite, and the framework U.S. regulators reference when articulating AI expectations. ISO/IEC 42001:2023 is the first certifiable management-system standard for artificial intelligence, with a Plan-Do-Check-Act structure and Annex A controls modeled on the ISO 27001 / 42001 lineage. It is the framework most international and certifiable programs cite.

Most enterprise AI security programs end up needing to satisfy both. The Govern function in NIST AI RMF and Clauses 4–10 in ISO 42001 cover the same governance terrain in different vocabularies; mapping the two at the control level is the highest-leverage piece of work a GRC team can do.

How the Crosswalk Was Built

Anchor on framework function / control IDs. NIST AI RMF subcategories use IDs like GOVERN-1.1; ISO 42001 uses Annex A control IDs like A.6.2.5. The crosswalk preserves these exact IDs so an auditor can match them against the source documents.
Map to capability, not feature. Each row in the crosswalk maps a framework requirement to a Vaikora capability — “deterministic policy enforcement,” “7-factor probabilistic risk scoring,” “SHA-256 hash-chained audit” — rather than to product menu items, so the mapping is durable across product versions.
Distinguish full coverage from partial. Coverage is marked Full where Vaikora is the primary control owner, Partial where Vaikora supplies one part of a multi-control answer, and Reference where Vaikora produces the evidence but the control is owned elsewhere.
Include the evidence artifact. Every row names the evidence artifact (audit log entries, redaction summaries, policy export, SIEM event stream) so a GRC team can lift the evidence directly into a NIST AI RMF profile or ISO 42001 Statement of Applicability.

NIST AI RMF 1.0 — Highlighted Crosswalk Rows

The full crosswalk in the XLSX covers all four functions across every relevant subcategory with coverage level and evidence artifact. The bullets below highlight the rows most often requested in evidence packages, grouped by function.

Govern

GOVERN-1.1 (legal / regulatory requirements identified) — Full. Six compliance presets (standard / strict / permissive / hipaa / pci-dss / gdpr) selectable per workspace. Evidence: workspace policy export.
GOVERN-1.4 (accountability structures) — Full. SAML SSO + SCIM-provisioned roles; the AuthN/Z layer of the reference architecture. Evidence: identity provider mapping; role assignments.
GOVERN-1.5 (ongoing monitoring) — Full. SHA-256 hash-chained audit; SIEM connectors. Evidence: audit log export; SIEM dashboards.
GOVERN-2.1 (roles and responsibilities documented) — Partial. Workspace-bound policy ownership; per-route compliance preset. Vaikora provides the technical control surface; documentation is owned by the program.
GOVERN-4.1 (workforce receives training) — Reference. Owned by the program; training records are owned by the program.

Map

MAP-1.1 (intended use defined) — Partial. Per-route compliance preset and capability scoping. Vaikora binds intended use to a workspace; the declaration of intended use is owned by the program.
MAP-2.3 (data sources documented) — Partial. Audit log records action_type and serving provider; routing policy enumerates upstream providers. Provenance of the data itself is owned upstream.
MAP-5.1 (risks evaluated) — Full. 7-factor probabilistic risk score on every request; risk_score field in the audit log. Evidence: risk score distribution by route.

Measure

MEASURE-2.5 (data integrity) — Full. Reversible PII redaction (synthetic / mask / hash); 4-layer detection on tool and RAG outputs. Evidence: redaction summaries; detection event stream.
MEASURE-2.7 (security and resilience) — Full. 12+ detection vectors / 4 layers; deterministic policy + probabilistic risk score; provider fallback. Evidence: detection event stream; fallback decisions.
MEASURE-2.8 (privacy) — Full. content: false metadata-only logging; reversible redaction with format preservation. Evidence: audit entries showing content: false.
MEASURE-2.11 (incident response readiness) — Full. SHA-256 hash-chained audit replayable for IR; SIEM integration. Evidence: replayed incident timeline.

Manage

MANAGE-1.1 (risks prioritized) — Full. Risk-score distribution per workspace; route-level risk metrics. Evidence: risk dashboards per workspace.
MANAGE-2.3 (rollback plans) — Partial. Provider fallback policy; deterministic block decisions. Vaikora handles upstream rollback; model rollback is owned by the model lifecycle.
MANAGE-4.1 (post-deployment monitoring) — Full. Continuous detection event stream; behavioral baselines per session and agent. Evidence: behavioral detection events.

ISO/IEC 42001:2023 — Highlighted Annex A Controls

ISO 42001 Annex A enumerates controls organized by clause. The bullets below are the controls most often referenced in Statements of Applicability and certification evidence.

5.2 (AI policy) — Full. Per-workspace compliance preset bound to a deterministic policy engine. Evidence: policy export per workspace.
6.2.5 (AI system impact assessment) — Partial. 7-factor risk score; risk distribution per route. Vaikora supplies measurement input; the impact assessment is authored by the program. Evidence: risk-score histogram per route.
7.2 (data quality for AI systems) — Full. 4-layer detection on inputs and tool outputs; reversible PII redaction. Evidence: detection event stream; redaction summaries.
7.4 (privacy) — Full. content: false logging; reversible redaction with format preservation. Evidence: audit entries; redaction summaries.
8.2 (operational planning and control) — Full. Per-route compliance preset; provider fallback policy. Evidence: routing policy export.
8.3 (information security for AI) — Full. Inline AI gateway with the 5-layer reference architecture; SAML SSO + SCIM. Evidence: architecture diagram; identity provider mapping.
8.4 (resources for AI systems) — Full. Budget caps and rate limits per workspace and per key. Evidence: budget configuration export.
9.2 (internal audit) — Full. SHA-256 hash-chained audit; replayable evidence; SIEM export. Evidence: hash-chain integrity report.
9.3 (management review) — Partial. Periodic risk-score and detection-event reports for management. Vaikora produces the data; the management review cadence is owned by the program.
10.2 (continual improvement) — Partial. Detection event stream feeds policy tuning; quarterly preset review. Vaikora supplies the feedback signal; the improvement loop is owned by the program.

NIST AI RMF ↔ ISO 42001 Rosetta (Highest-Cited Pairings)

Programs that have to satisfy both frameworks find this rosetta useful — the same Vaikora capability satisfies a NIST subcategory on one side and an ISO 42001 Annex A control on the other. The full XLSX includes the complete rosetta; the highlights below are the pairings most often called out in dual-framework evidence packages.

Vaikora capability	NIST AI RMF subcategory	ISO/IEC 42001 Annex A control
Six compliance presets bound to deterministic policy	GOVERN-1.1	A.5.2
Reversible PII redaction (synthetic / mask / hash)	MEASURE-2.5, MEASURE-2.8	A.7.2, A.7.4
12+ detection vectors / 4 layers	MEASURE-2.7	A.8.3
7-factor probabilistic risk score	MAP-5.1, MANAGE-1.1	A.6.2.5
SHA-256 hash-chained, content-free audit	GOVERN-1.5, MEASURE-2.11	A.9.2
SAML SSO + SCIM provisioning	GOVERN-1.4	A.8.3
Per-workspace budget caps and rate limits	MANAGE-2.3	A.8.4
Provider fallback across 12 LLM providers	MANAGE-2.3, MANAGE-4.1	A.8.2

Downloadable Crosswalk (XLSX)

The full crosswalk is published as an exportable XLSX so a GRC team can drop it into a Statement of Applicability or a NIST AI RMF profile without retyping. Hand the file to your auditor; the column headers are stable across releases. The XLSX contains four sheets:

NIST AI RMF. All four functions across the relevant subcategories with Vaikora capability, coverage level, and evidence artifact.
ISO 42001 Annex A. Annex A controls grouped by clause with the same four columns.
NIST subcategory ↔ ISO Annex A control pairings sharing the same Vaikora capability.
Evidence index. Each evidence artifact named once with the audit-log fields used and where to find it.

What This Mapping Does Not Claim

It does not claim certification. Mapping a capability to a control is not the same as a control being audited and certified. The XLSX is a crosswalk, not a certification.
It does not replace program documentation. NIST AI RMF and ISO 42001 expect program-level artifacts (AI inventory, impact assessments, training records, management review minutes) that Vaikora supplies inputs to but does not own.
It does not freeze in place. Both frameworks evolve. The crosswalk is a working document; the XLSX is versioned and updated when a framework revision changes the underlying IDs or expectations.

Next Steps

Download the XLSX, scope it to the workspaces your AI program covers, and walk through the coverage column with your GRC team. The natural next read is “AI Compliance in 2026: What CISOs Must Prove” for the eight evergreen evidence patterns and “Running a DPIA for AI Workflows” for the GDPR Art. 35 view.

A NIST AI RMF / ISO 42001 mapping for AI security tooling is a crosswalk that ties each framework’s functions and Annex A controls (e.g., GOVERN-1.5, MEASURE-2.7, A.7.4, A.9.2) to concrete platform capabilities — for Vaikora: deterministic policy engine with probabilistic risk scoring, reversible PII redaction, content-free SHA-256 hash-chained audit, SAML / SCIM identity, and SIEM-native reporting.

Your AI Agents Need a Control Layer

See how Vaikora intercepts, evaluates, and enforces policy on every AI agent action — in real time, before execution.

Frequently Asked Questions

Does Vaikora provide a NIST AI RMF mapping?

Yes. The crosswalk maps every relevant NIST AI RMF subcategory (across Govern, Map, Measure, Manage) to a concrete Vaikora capability with coverage level (Full / Partial / Reference) and the evidence artifact a GRC team can lift into a NIST profile.

Does Vaikora provide an ISO/IEC 42001 mapping?

Yes. The crosswalk includes ISO/IEC 42001:2023 Annex A controls grouped by clause, with the same coverage and evidence columns. This drops cleanly into a Statement of Applicability.

Can I use this crosswalk for both frameworks at once?

Yes — that is the design intent. The third sheet of the XLSX is a rosetta that shows which NIST subcategory and which ISO Annex A control are satisfied by the same Vaikora capability, so a dual-framework program does not have to map twice.

What is the difference between Full, Partial, and Reference coverage?

Full means Vaikora is the primary control owner and produces the evidence directly. Partial means Vaikora supplies one part of a multi-control answer (typically the technical surface, while program-level documentation is owned by the GRC team). Reference means Vaikora produces evidence that supports the control even though the control itself is owned elsewhere (training records, governance committees, etc.).

Is this a substitute for an audit?

No. The crosswalk is an evidence-mapping tool that an auditor can use, not a certification. Certification (e.g. ISO 42001) is a separate process that a certifying body conducts against the program as a whole.

How often is the crosswalk updated?

On framework revision (e.g. when NIST releases an updated AI RMF profile or ISO updates 42001) and on Vaikora capability changes that affect the evidence column. The XLSX is versioned so a GRC team can pin to a specific revision in their evidence package.

How does this connect to the rest of the AI security program?

It connects through the audit log. Each row’s evidence column points back to a field in the audit log, an export from the policy engine, or a SIEM event. That means the same operational telemetry that runs the security program also produces the framework evidence — no separate evidence pipeline.