Both govern AI risk, but only one ends in a certificate. The NIST AI Risk Management Framework organizes the work — GOVERN, MAP, MEASURE, MANAGE — while ISO 42001 turns it into a certifiable management system. Here is the function-by-function crosswalk, and the honest answer on when each earns its keep.
The one-line difference: NIST AI RMF is a voluntary framework you use. ISO 42001 is a standard you certify against. They map cleanly onto each other — NIST itself publishes a crosswalk — so “both” is a strategy, not a contradiction.
Plain-English comparison · Function-to-clause crosswalk · Last reviewed July 2026
The NIST AI Risk Management Framework (AI RMF 1.0, January 2023) is a voluntary, non-certifiable framework organized around four functions — GOVERN, MAP, MEASURE, MANAGE. ISO/IEC 42001:2023 is a certifiable management-system standard with mandatory clauses and 38 Annex A controls. They come from different worlds — a US standards institute's guidance versus an international certification standard — but they describe the same underlying discipline, and NIST publishes an official crosswalk between the two. In practice the choice is rarely either/or: the RMF (and its Generative AI Profile, NIST-AI-600-1) is excellent working methodology for the risk work inside an AIMS, while ISO 42001 supplies what the RMF deliberately does not: an auditable structure, a Statement of Applicability, and a certificate that answers enterprise procurement in one line. US-facing vendors feel this most concretely — AI questionnaires increasingly name both.
Key Takeaways
- NIST AI RMF: voluntary framework, four functions (GOVERN, MAP, MEASURE, MANAGE), no certification — you attest to using it. ISO 42001: certifiable management system with clauses 4–10 and 38 Annex A controls.
- They crosswalk cleanly (NIST publishes one): GOVERN ≈ clauses 4/5/7 + A.2/A.3, MAP ≈ 4.3/6.1.4 + A.4/A.5, MEASURE ≈ 6.1.2/9 + A.6.2.4/A.6.2.6, MANAGE ≈ 6.1.3/8/10 + A.6.2.5/A.8/A.10.
- The GenAI-specific layer: NIST-AI-600-1 (July 2024) profiles the RMF for generative AI — useful methodology for the A.5 impact assessments and A.6 lifecycle controls of LLM systems.
- Decision rule: need a defensible proof for customers or regulators → ISO 42001. Need a working method your engineers can run tomorrow → start with the RMF. Selling AI to US enterprise → expect to reference both.
- Using the RMF does not certify you against ISO 42001, and holding ISO 42001 does not mean “NIST compliant” — keep the claims distinct on your trust page.
RMF Functions → ISO 42001 Clauses & Controls
Indicative mapping at the function level — the granular subcategory-to-control detail is in NIST's published crosswalk. This is the working view for planning an integrated program.
| RMF function | What it covers | ISO 42001 home |
|---|---|---|
| GOVERN | Culture, policies, roles, accountability, and risk tolerance for AI across the organization. | Clauses 4 (context), 5 (leadership & AI policy), 7 (support/competence) · Annex A.2 (AI policy), A.3 (roles & concerns reporting), A.10.2 (responsibility allocation). |
| MAP | Establish context per AI system: purpose, stakeholders, capabilities, and potential impacts. | Clause 4.3 (AIMS scope) and 6.1.4 (impact assessment) · Annex A.4 (resource documentation), A.5 (impact assessments per ISO/IEC 42005), A.8.5 (information for interested parties). |
| MEASURE | Analyze and track AI risks and trustworthiness characteristics with metrics and testing. | Clauses 6.1.2 (risk assessment) and 9 (monitoring, internal audit, management review) · Annex A.6.2.4 (verification & validation), A.6.2.6 (operation & monitoring), A.7.4 (data quality). |
| MANAGE | Prioritize, treat, and respond to mapped and measured risks. Allocate resources and respond to incidents. | Clauses 6.1.3 (risk treatment), 8 (operation), 10 (improvement) · Annex A.6.2.5 (deployment), A.6.2.8 (event logs), A.8.3/A.8.4 (external reporting & incident communication), A.10.3 (suppliers). |
When Each One Earns Its Keep
- Choose ISO 42001 when the output must be provable to someone else: enterprise customers, AI questionnaires (SIG/CAIQ AI modules), regulators, boards. Certification by an ISO/IEC 42006-accredited body is the strongest third-party signal available today.
- Choose NIST AI RMF when you need methodology now: it is free, immediately actionable, and its playbook + Generative AI Profile give engineering teams concrete practices without waiting for a management-system build.
- Use both when you are serious: run MAP/MEASURE/MANAGE as the working method inside your AIMS risk process (6.1.2–6.1.4), let GOVERN shape clauses 4–5, and let ISO 42001 wrap it in auditable structure. The crosswalk means near-zero duplicated work.
- US federal context: agencies and their vendors will meet NIST framing (including via OMB guidance) — an ISO 42001 program that speaks RMF vocabulary travels furthest.
ISO 42001 vs NIST AI RMF — Common Questions
Certifiability, crosswalks, and running both without duplication.
Can you get certified against the NIST AI RMF?
What are the four functions of the NIST AI RMF?
Is there an official crosswalk between the two?
What is the NIST Generative AI Profile (NIST-AI-600-1)?
We already follow the NIST AI RMF — how much extra is ISO 42001?
Which should an Indian or global SaaS company selling to US enterprise pick?
Related reading: the Learn hub, the NIST Cybersecurity Framework guide, AI risk vs impact assessment, ISO 42001 vs the EU AI Act, and the ISO 42001 hub. Terms in the compliance glossary.
Related Reading
ISO 42001 Knowledge Hub
AIMS controls, EU AI Act mapping, risk assessment and guides.
Read moreISO 42001 Annex A Controls
The 38 AI-specific controls grouped by domain, mapped to clauses.
Read moreAI Risk Assessment
Risk identification, impact analysis and treatment for AI systems.
Read moreEU AI Act Mapping
How ISO 42001 controls map onto EU AI Act obligations.
Read moreISO 42001 Certification Guide
From gap analysis to Stage 2 audit: the path to an AIMS certificate.
Read moreSOC 2 for AI Companies
Enterprise AI procurement, model/data security, and ISO 42001 pairing.
Read more