ISO/IEC 42001:2023 is a certifiable international standard for an AI management system, published on 18 December 2023 (ISO/IEC 42001:2023 on iso.org). The NIST AI Risk Management Framework (AI RMF 1.0, NIST AI 100-1) is free, voluntary US guidance released on 26 January 2023, with no certification.
The two are not rivals. ISO/IEC 42001 tells you what an audited AI management system must contain. The AI RMF gives you a detailed vocabulary of AI risk outcomes. ISO/IEC 42001 even points to the AI RMF in its own clause 4.1, for the roles organizations play across the AI life cycle.
ISO 42001 and the NIST AI RMF compared, attribute by attribute
ISO/IEC 42001 is a requirements standard you can be certified against; the NIST AI RMF is a voluntary framework you adopt in part or in full. Neither is required by US law.
| Attribute | ISO/IEC 42001:2023 | NIST AI RMF 1.0 |
| Type | International management system standard with requirements (“shall”) | Voluntary framework, “rights-preserving, non-sector-specific, and use-case agnostic” |
| Issuer | ISO and IEC, joint committee JTC 1/SC 42 | National Institute of Standards and Technology, US Department of Commerce |
| Published | 18 December 2023, edition 1 | 26 January 2023, version 1.0 (NIST AI 100-1) |
| Certifiable | Yes, by accredited certification bodies; ISO itself does not certify | No NIST certification program found |
| Structure | Clauses 4 to 10; Annex A reference controls and Annex B implementation guidance (both normative); Annexes C and D (informative) | Four functions (Govern, Map, Measure, Manage), 19 categories, 72 subcategories |
| Audience | Any organization that provides or uses AI systems, any size | Organizations and AI actors across the AI life cycle, any sector |
| Cost | CHF 225 for the standard, plus audit fees if you certify | Free to download |
| Companion material | ISO/IEC 42005 impact assessment, ISO/IEC 23894 risk guidance | AI RMF Playbook; Generative AI Profile, NIST AI 600-1 (26 July 2024) |
| Status, October 2026 | Published, stage 60.60, no amendment | Version 1.0 in force; NIST says it is being revised |
| US legal status | Voluntary | Voluntary; named in Texas TRAIGA’s liability provisions (via the GenAI Profile) |
Sources: iso.org standard and certification pages; ISO/IEC 42001:2023 official preview (table of contents); NIST AI RMF page and NIST AI 100-1; AI RMF Core on NIST’s AI Resource Center; Texas HB 149 enrolled text.
The NIST AI Risk Management Framework page says the framework is “intended for voluntary use” and that AI RMF 1.0 “is being revised as part of the White House AI Action Plan”. No AI RMF 2.0 has been published.
How do ISO 42001 and the NIST AI RMF differ in practice?
The difference between ISO/IEC 42001 and the AI RMF shows up at audit time. ISO/IEC 42001 says what you “shall” do, and an auditor checks the evidence. The AI RMF describes outcomes you choose from, and nobody issues a certificate for meeting them.
ISO/IEC 42001 makes you build a management system: a defined scope, an AI policy, roles, risk and impact assessments, a Statement of Applicability for Annex A controls, internal audits, management reviews, corrective action and continual improvement under clause 10. Each produces records. A certificate then runs on a cycle of surveillance audits. For the risk assessment, the evidence and questions involved are set out in how auditors test an AI risk method.
The AI RMF is organized around four functions. Govern sets culture, policy and accountability, and NIST calls it “a cross-cutting function” that enables the other three. Map establishes context and identifies risks. Measure analyzes and tracks them. Manage prioritizes and responds. The AI RMF Core says that, with governance in place, the functions “may be performed in any order”.
So the AI RMF is richer on what good AI risk outcomes look like, especially across its seven trustworthiness characteristics. ISO/IEC 42001 is stricter on proving that a process exists and runs.
Where do ISO 42001 and the NIST AI RMF overlap?
Each AI RMF function maps to a large set of ISO/IEC 42001 clauses, and there is a published mapping. The NIST AI RMF to ISO/IEC 42001 crosswalk maps each of the 72 AI RMF subcategories to ISO/IEC 42001 clauses and Annex B items.
Read it with three caveats. NIST’s AI Resource Center lists it as a contribution from Microsoft and says inclusion “does not imply NIST endorsement”. It was built against the FDIS, the final draft of ISO/IEC 42001, before publication. And it maps to Annex B implementation guidance numbers (B.2.2 to B.10.4), which follow the Annex A control numbering, rather than to Annex A itself.
| AI RMF function | Subcategories | ISO/IEC 42001 clauses the crosswalk cites | Annex B areas cited |
| Govern | 19 | 4.1, 4.4, 5.1 to 5.3, 6.1.1 to 6.1.4, 6.2, 7.1 to 7.4, 8.2 to 8.4, 9.1, 9.3 | All except B.7 (data) |
| Map | 18 | 4.1, 4.3, 5.1, 5.2, 6.1.1, 6.1.2, 6.1.4, 6.2, 7.2 to 7.4, 7.5.3, 8.2 to 8.4 | All except B.3 (internal organization) |
| Measure | 22 | 4.4, 6.1.1, 6.1.2, 8.2, 8.4, 9.1, 9.2, 9.3, 10.1 | All except B.10 (third parties) |
| Manage | 13 | 6.1.1 to 6.1.4, 7.1, 9.2, 9.3, 10.1, 10.2 | All except B.2 (AI policy) |
Source: NIST AI RMF to ISO/IEC FDIS 42001 crosswalk, listed on NIST’s AI Resource Center, counted by AI Training Skool. Clause numbers are as the crosswalk prints them against the FDIS.
Four patterns stand out. Govern is the only function that cites every subclause of leadership (5.1 to 5.3) and support (7.1 to 7.4), where ISO/IEC 42001 puts policy, roles and competence. Clause 10, improvement, appears only under Measure and Manage. The life cycle guidance (B.6) is cited by 36 of the 72 subcategories and impact assessment (B.5) by 22, far ahead of any other area. And 12 subcategories map to clauses only, with no Annex B item. The most-cited single items are B.5.4, assessing impact on individuals and groups (16 subcategories), and B.6.2.6, AI system operation and monitoring (15). Because impact work carries so much of the mapping, set up the impact assessment record ISO 42001 expects early.
The reference runs the other way too. ISO/IEC 42001 clause 4.1 lists AI roles (providers, producers, customers, partners, subjects, authorities) and notes that these roles and their link to the life cycle “are also described in the NIST AI risk management framework” (official preview). A US team that already uses the AI RMF’s AI actor language can reuse it in its ISO/IEC 42001 context analysis.
Is there a NIST AI RMF certification?
No. We found no certification or conformity program for the AI RMF on nist.gov or NIST’s AI Resource Center in October 2026. Training providers sell AI RMF courses with their own certificates, but those are private credentials, not NIST attestations.
NIST describes the framework as voluntary, and its NIST AI RMF Playbook FAQ says the Playbook is “neither a checklist nor an ordered list of steps”. Users are not expected to apply every suggestion. That flexibility is the point of the framework, and it is also why an outsider cannot certify you against it.
If you need a third party to attest to your AI governance, ISO/IEC 42001 certification is the route that exists today. For a person, AI RMF knowledge is shown through training and work samples, not a NIST credential.
Where do the Generative AI Profile and Texas TRAIGA fit?
NIST published the Generative AI Profile, NIST AI 600-1, on 26 July 2024 as a companion to AI RMF 1.0. It sets out 12 risks unique to or made worse by generative AI, including confabulation, information integrity, data privacy, and value chain and component integration.
ISO/IEC 42001 has no generative AI annex; none of its clause or annex titles name it. If you run generative AI, you can feed the GenAI Profile’s 12 risks into your ISO/IEC 42001 risk assessment as risk sources, alongside the informative Annex C.
Texas gives the profile legal weight. Under Texas HB 149 (TRAIGA), effective 1 January 2026, section 552.105(e) says a defendant may not be found liable if it discovers a violation through listed routes. One route is substantial compliance with the most recent GenAI Profile or “another nationally or internationally recognized risk management framework”, as an internal review process. The statute does not name ISO/IEC 42001, so whether a court treats it as such a framework is untested.
Which should a US organization adopt first, ISO 42001 or the NIST AI RMF?
Start with the NIST AI RMF if you need a shared risk language fast and nobody is asking for a certificate. Start with ISO/IEC 42001 if a customer, partner or tender asks for certification, or if you already run an ISO management system.
Choose ISO/IEC 42001 first if:
- An enterprise customer or contract asks for third-party certification of your AI governance.
- You already hold ISO/IEC 27001 or another ISO certificate, so document control, internal audit and management review exist. Our guide to the first international AI management system standard compares it with ISO/IEC 27001.
- You sell AI products or services outside the US and want an international reference.
Choose the NIST AI RMF first if:
- You are at an early stage, with a handful of AI systems and no external demand for a certificate.
- Your work is with US federal agencies or regulated US sectors that already cite NIST.
- You operate generative AI in Texas and want the GenAI Profile’s risk catalog in place.
Use both if:
- You want the AI RMF’s risk detail and an auditable system around it. Map each AI RMF subcategory you adopt to the ISO/IEC 42001 clause or Annex B item from the crosswalk, then build in clause order. The phases are set out in taking an organization through ISO 42001 implementation.
How to run ISO 42001 and the NIST AI RMF without duplicating work
To run ISO/IEC 42001 and the AI RMF together, keep one set of records and tag each record with both references. The crosswalk lets you do that line by line.
- One AI system register. Use it for ISO/IEC 42001 clause 4 scope and for the AI RMF’s Map function.
- One risk method and register. Score risks under clause 6.1.2 and tag each with the AI RMF subcategories it addresses.
- One Statement of Applicability. Add a column for the AI RMF subcategories each Annex A control supports. The control areas are summarized in ISO 42001 control objectives in plain terms.
- One monitoring plan. Clause 9.1 monitoring can carry the AI RMF Measure outcomes you adopt.
- One management review. Use clause 9.3 to review both, so leadership sees one picture.
The crosswalk was drafted against the FDIS, so check clause references against your licensed copy of ISO/IEC 42001:2023 before you rely on them in audit evidence.
Frequently Asked Questions
Is the NIST AI RMF mandatory?
No. NIST says the AI Risk Management Framework is intended for voluntary use, and AI RMF 1.0 (NIST AI 100-1, January 2023) carries no legal force on its own. Some laws refer to it: Texas TRAIGA, effective 1 January 2026, names the NIST Generative AI Profile in a liability provision, alongside other recognized AI risk frameworks.
Is there a NIST AI RMF 2.0?
No. As of October 2026, the current version is AI RMF 1.0, NIST AI 100-1, released on 26 January 2023. NIST’s AI RMF page says version 1.0 is being revised as part of the White House AI Action Plan, but no 2.0 has been published. The Generative AI Profile, NIST AI 600-1 from July 2024, is a companion profile, not a new version.
Does ISO 42001 certification prove NIST AI RMF compliance?
No. There is no NIST AI RMF compliance certificate to prove. ISO/IEC 42001 certification shows that an accredited certification body has audited your AI management system for a stated scope. The crosswalk listed on NIST’s AI Resource Center maps the two, so an ISO/IEC 42001 system can cover many AI RMF outcomes, but you should document which subcategories you adopted.
Is there an official mapping between ISO 42001 and the NIST AI RMF?
The public mapping is a crosswalk listed on NIST’s AI Resource Center, contributed by Microsoft and built against the final draft (FDIS) of ISO/IEC 42001. It maps all 72 AI RMF subcategories to ISO/IEC 42001 clauses and Annex B items. NIST states that listing a crosswalk does not imply its endorsement.
Which is better, ISO 42001 or the NIST AI RMF?
Neither is better in general. ISO/IEC 42001 suits organizations that need a third-party certificate or already run ISO management systems. The NIST AI RMF suits teams that need a detailed, free AI risk vocabulary without certification. You can also use the AI RMF’s risk detail inside an ISO/IEC 42001 management system, with the crosswalk linking the two.
Is the NIST AI RMF the same as the NIST RMF?
No. “NIST RMF” usually means the Risk Management Framework in NIST SP 800-37 Rev. 2 (December 2018), which applies security and privacy controls to information systems. The NIST AI RMF is a separate document, NIST AI 100-1 (January 2023), for managing the risks of AI systems. The crosswalk on NIST’s AI Resource Center maps the AI RMF, not SP 800-37, to ISO/IEC 42001.
What is the NIST AI RMF Playbook?
The NIST AI RMF Playbook is a free companion resource on NIST’s AI Resource Center that suggests actions for each AI RMF subcategory. NIST says it is neither a checklist nor an ordered list of steps, and users are not expected to apply every suggestion. NIST updates it from public feedback, so there is no final version.
