ISO 42001 vs NIST AI RMF: which one to choose

ISO 42001 vs NIST AI RMF: which one does your customer actually need?

ISO/IEC 42001:2023 is a certifiable management system. An external certification body can audit an organization against it and issue a certificate, and that certificate is what a customer usually means when it asks for ISO 42001. The NIST AI Risk Management Framework 1.0 is different: NIST describes it as intended for voluntary use, and it has no certification. Before you answer a questionnaire or contract clause, decide which of three things the requester actually wants: a certificate from an independent body, a governance structure you build and run, or a framework you can cite in risk documentation. The answer determines whether you need an audit at all. Ask the requester to name the document or certificate they will accept as evidence, and get that wording in writing.

A regulator reference is often less specific than it looks. A reference to the NIST AI RMF usually expects a documented mapping of your AI risks to the Govern, Map, Measure and Manage functions, with named owners and review cycles. A reference to ISO 42001 usually expects a statement of applicability under clause 6.1.3, which justifies each Annex A control you include or exclude, plus evidence that the management system runs. Your organization owns its controls and the decisions behind them. An implementation partner can help prepare the system and its records, but only an independent auditor or certification body issues any report or certificate. Confirm with the requester whether they accept self-declared alignment or require third-party assessment.

The two are also used together. NIST hosts a crosswalk, contributed by Microsoft, that maps AI RMF subcategories to ISO/IEC FDIS 42001 clauses and Annex B guidance, and NIST states that inclusion implies neither endorsement nor full coverage. Treat the crosswalk as a starting point to check against your own environment, not as proof that one framework satisfies the other. Because the AI RMF 1.0 is being revised as part of the White House AI Action Plan, and ISO/IEC 42001 has no amendment in progress, confirm the current versions with NIST and ISO before you cite either one in a contract.

Which framework fits which request

This table matches four requests buyers commonly send, such as a certificate demand or a NIST alignment questionnaire, to the framework that answers each one. Read each row to see what evidence satisfies the request, whether an independent body issues a result, and the usual next step.

Request	What satisfies it	Independent result, and next step
Certificate demand ("Are you ISO/IEC 42001 certified?")	An ISO/IEC 42001 certificate covering a defined AI management system scope, backed by a statement of applicability and internal audit and management review records	Yes, a certification body accredited or not issues the certificate. Next step: ask for the certificate scope and verify it with the issuing body before accepting it
Request for a governance framework	A documented AI management system built on clauses 4 to 10 and the Annex A controls, or a NIST AI RMF program organized around GOVERN, MAP, MEASURE and MANAGE	No body issues a result for the NIST AI RMF, since it is voluntary with no certification. For ISO/IEC 42001, certification is optional. Next step: agree which framework the contract names
Questionnaire asking for NIST alignment	Evidence mapped to AI RMF subcategories, such as the NIST-hosted crosswalk to ISO/IEC 42001, plus records showing each mapped control operates	No independent result is issued for NIST alignment. Next step: confirm whether the buyer accepts a self-assessment or wants third-party evidence, and state that crosswalk inclusion does not imply full coverage
EU AI Act reference ("Do you meet Article 17 quality management?")	A quality management system that meets the AI Act requirements, with harmonised standard references such as EN 18286 where they apply	The European Commission's view is that ISO/IEC 42001 helps set up an AI management system but does not match the AI Act quality management system. ISO/IEC 42006:2025 governs bodies certifying ISO/IEC 42001, not AI Act conformity. Next step: check the AI Act obligations with legal counsel
Illustrative: a customer asks for both ISO 42001 and NIST alignment	An ISO/IEC 42001 management system, with the NIST crosswalk used to show where AI RMF subcategories map to clauses and Annex A guidance	A certification body issues the ISO certificate. The NIST mapping stays a self-prepared document. Next step: send the certificate and the mapping together, with the crosswalk's no-endorsement caveat

How the structures differ: clauses and Annex A versus four functions

ISO/IEC 42001 is built as a management system, and NIST's AI RMF is built as a set of outcomes. ISO/IEC 42001 applies the harmonized structure shared by ISO management system standards, with requirements in clauses 4 to 10 covering context, leadership, planning, support, operation, performance evaluation and improvement. On top of those clauses sits Annex A, which lists 38 reference controls across nine objective areas, from A.2 policies related to AI through A.10 third-party and customer relationships. Annex B gives implementation guidance for each control. Organizations do not have to implement every Annex A control, but clause 6.1.3 requires a statement of applicability that lists the necessary controls and justifies including or excluding each one. Your team should expect to write that justification and keep it current.

The NIST AI RMF core is organized differently. It has four functions, GOVERN, MAP, MEASURE and MANAGE, broken into 72 subcategories. Those subcategories describe outcomes to achieve, not controls to implement and document one by one. NIST's framework has no required statement of applicability, no defined internal audit cycle and no management review step that an outside party tests. Organizations choose how to meet each subcategory, which gives flexibility but also means nothing forces a documented exclusion decision or a recurring governance review. The practical question for a CISO is whether the internal process you already run would produce evidence an outside party could test against each.

Several parts of ISO/IEC 42001 exist mainly to make the system self-correcting, and a certification body audits them directly. Internal audit under clause 9.2, management review under clause 9.3 and corrective action under clause 10.2 are the clauses that show the AI management system is operated and improved over time, not just designed. The NIST AI RMF has no equivalent requirement to demonstrate those loops. Before choosing a path, confirm with your certification body which evidence it will sample for each of these clauses, and check whether your existing internal audit and management review calendars already cover AI systems in scope.

Where the four NIST functions land in ISO/IEC 42001

NIST AI RMF organizes its 72 subcategories under four functions, and the crosswalk NIST hosts maps those subcategories to ISO/IEC 42001 clauses and Annex A guidance. Each card below shows one function and one example row, so you can see where the management system and the control set meet the voluntary framework.

GOVERN: 19 subcategories

GOVERN covers the policies, accountability and risk culture around AI. In the crosswalk, GOVERN 1.6, the inventory of AI systems, maps to the implementation guidance for A.4.2 to A.4.6, which describes the resources each AI system depends on. Check that your inventory records the data, tooling, system, computing and human resources those controls name, and that the inventory is kept current enough to serve as audit evidence.

MAP: 18 subcategories

MAP covers establishing the context of each AI system, including its purpose, stakeholders and potential impacts. These subcategories feed the impact assessment and risk identification work that ISO/IEC 42001 places in clauses 6.1 and 8.4 and in the A.5 controls. Confirm that the context you record for each system is the same context your AI risk and impact assessments use, so an assessor can trace one to the other.

MEASURE: 22 subcategories

MEASURE covers testing, monitoring and evaluating AI systems against their risks. In the crosswalk, MEASURE 2.4, production monitoring, maps to clause 9.1 on monitoring and measurement and to the guidance for A.6.2.6 and A.6.2.8 on operation and monitoring. Confirm that your production monitoring produces records that show performance was checked against defined criteria, not only that dashboards exist.

MANAGE: 13 subcategories

MANAGE covers prioritizing and responding to AI risks, including risks that come from outside the organization. In the crosswalk, MANAGE 3.1, third-party AI risks monitored, maps to the implementation guidance for A.10.2 and A.10.3, which address managing suppliers of AI products and services. Confirm that supplier reviews, contract terms and monitoring records for each third-party model or service are kept with the AI system they support.

Reading a crosswalk row

A crosswalk row is a mapping, not a finding that your organization complies. NIST states that inclusion of a crosswalk implies neither endorsement nor full coverage, and the mapping was contributed by an outside party. Treat each row as a pointer to evidence you must produce yourself, and confirm each mapped clause or Annex A guidance item against your own records before you cite it to a customer or assessor.

Status and revisions: what is stable and what is changing

ISO/IEC 42001:2023 is the stable side of this comparison. It is edition 1, published on 18 December 2023, and as of October 2026 ISO lists no amendment or revision in progress. The companion standard that governs the bodies auditing against it, ISO/IEC 42006:2025, was published on 7 July 2025 and adds requirements on top of ISO/IEC 17021-1 for certification bodies. Confirm the current edition status on the ISO standard page before you write it into a contract, and confirm the certification body's accreditation scope separately, since accreditation of AI management system certification is voluntary and not every body holds it.

The NIST AI RMF is the side that is moving. NIST released AI RMF 1.0 on 26 January 2023 and describes it as intended for voluntary use. NIST's AI Risk Management Framework page now says that AI RMF 1.0 is being revised as part of the White House AI Action Plan. Before you cite the framework in a policy, customer questionnaire or regulator response, check which version NIST currently publishes and cite that version by its identifier and date. Two companion documents also sit on NIST's side: the Generative AI Profile, NIST AI 600-1, published on 26 July 2024, and a concept note released on 7 April 2026 for an AI RMF profile on trustworthy AI in critical infrastructure. A concept note is a proposal, not a finished profile, so do not treat it as a set of requirements.

Recheck the crosswalk against the revised framework. The NIST-hosted crosswalk to ISO/IEC 42001 maps AI RMF subcategories to ISO/IEC FDIS 42001 clauses and Annex B guidance, and NIST states that inclusion implies neither endorsement nor full coverage. Its row references were built against the current AI RMF subcategory numbering. If NIST renumbers or rewrites subcategories in the revision, any mapping in your evidence file will need to be redone. Assign an owner to check the NIST AI Resource Center each time either framework changes, and record the version and date of both documents in your statement of applicability and your governance records so an assessor can see which versions you relied on.

How organizations use both frameworks without duplicating work

This sequence lets one AI inventory and one risk process serve both NIST AI RMF alignment and an ISO/IEC 42001 management system, so each piece of evidence is produced once and mapped to both frameworks. It ends with an independent certification body only if a certificate is actually required.

Confirm which framework the requester named
Start by asking the customer, contract owner or regulator which framework they actually require: an ISO/IEC 42001 certificate, a self-declared alignment with the NIST AI RMF, or both. The answer sets scope and the endpoint. A certificate needs an accredited or recognized certification body and a defined management system scope, while NIST AI RMF alignment is a self-directed claim with no certification. The output is a short written requirement note that names the framework, the requester, the systems in scope and whether a certificate is needed. Confirm the exact wording of the requirement with the requester before any mapping work starts.
Build the AI inventory and risk process once
Create one register of AI systems, their owners, purposes, providers, data sources and deployment context, along with one AI risk assessment method that records how risks are identified, rated and accepted. This single set of records answers NIST GOVERN 1.6 on inventory and the risk process clauses 6.1.1 to 6.1.3 of ISO/IEC 42001. Assign a named owner for each system and confirm that leadership has approved the risk criteria. The output is an approved AI system inventory and a documented risk method, both version-controlled, that every later step draws from.
Map NIST subcategories to ISO clauses and Annex A controls
Use the NIST-hosted crosswalk to AI RMF to ISO/IEC 42001, which was contributed by Microsoft, to link each AI RMF subcategory to the ISO clauses and Annex A guidance it relates to. Read each row against your own evidence, because NIST states that inclusion implies neither endorsement nor full coverage. Where a subcategory has no matching control, record a gap rather than forcing a match. The output is a mapping register that lists each subcategory, its ISO reference, the evidence location and a reviewer's initials. Have a competent reviewer check the register against the source documents.
Record exclusions in the statement of applicability
Clause 6.1.3 of ISO/IEC 42001 requires a statement of applicability that lists the Annex A controls the organization needs and justifies including or excluding each one. Populate it from the mapping register so every included control points to evidence and every excluded control has a documented reason tied to the AI systems in scope. Do not exclude a control only because the crosswalk has no matching row. The output is an approved statement of applicability, signed by the accountable executive, that the internal audit and certification body will both read.
Run internal audit against both mappings
Plan internal audits of the AI management system under clause 9.2 of ISO/IEC 42001 and test the mapped NIST evidence in the same sampling. The audit checks whether the documented controls operate as written, whether records exist for the sampled AI systems, and whether gaps are logged. Findings feed corrective action under clause 10.2. The output is an internal audit plan, the audit records with findings, and a corrective action log that shows how each finding was closed or accepted. Confirm the audit frequency and effort in a written plan, because both depend on scope.
Hold management review and decide on changes
Management review under clause 9.3 of ISO/IEC 42001 brings internal audit results, AI risk changes, incidents and objectives in front of top management. The crosswalk ties GOVERN 2.3, on executive responsibility for AI risk decisions, to clauses 5.1, 5.2 and 9.3.1 to 9.3.3, so the review also serves the NIST governance expectation. Record decisions on resources, risk tolerance and system changes. The output is signed management review minutes with action owners and a list of changes to the inventory, risk method or statement of applicability.
Hand evidence to an independent certification body if required
Only if the requirement calls for a certificate, select an independent certification body and give it the management system documentation, the statement of applicability, the mapping register and the internal audit and management review records. ISO does not certify organizations, and the certification body alone issues the certificate. DeployOpen can prepare systems, controls and evidence for this step, but it does not issue audit reports or certificates. The output is an audit-ready evidence package and a written engagement with the certification body, with its accreditation status confirmed before work begins. Ask the certifier to confirm how it treats the NIST mapping, since it is not a certification requirement.

Where the NIST crosswalk misleads teams

The NIST-hosted crosswalk between AI RMF and ISO/IEC 42001 is a useful starting point for mapping, but it is not a compliance shortcut. The errors below share one root cause: treating a mapping row as proof of coverage, endorsement or permanence.

Treating a row as proof of coverage

A crosswalk row shows where a NIST subcategory and an ISO/IEC 42001 clause or control are related. NIST states that inclusion implies neither endorsement nor full coverage, so a row can point to guidance that only partly addresses the subcategory. Before you cite a row as evidence, read the ISO clause and its Annex B guidance yourself and record the gap, if any, in your own gap analysis.

Assuming NIST mappings satisfy internal audit

Certification bodies audit requirements that have no AI RMF counterpart, including internal audit under clause 9.2, management review under clause 9.3 and corrective action under clause 10.2. A team that satisfies AI RMF subcategories may still have no documented internal audit programme or management review records. Check each management system clause against its own evidence, independent of the NIST mapping.

Mapping GOVERN 1.3 and 1.4 to Annex A

GOVERN 1.3 and GOVERN 1.4 concern risk management processes and risk tolerance. The crosswalk points them to clauses 6.1.1 to 6.1.3 and 8.3 rather than to Annex A controls. Teams that look for an Annex A control for these items will build a control record that the crosswalk does not support. Document the risk process in the clause 6 and 8 records and reflect it in your statement of applicability.

Overstating the crosswalk's origin

The crosswalk was contributed by Microsoft and hosted by the NIST AI Resource Center. It is not an official NIST endorsement of a particular mapping, and it does not make Microsoft's interpretation authoritative for your organization. Note its contributor and status in your documentation, and treat it as one input that your own assessors review, not as a document NIST or an auditor has approved.

Skipping recheck after NIST revisions

NIST describes AI RMF 1.0 as being revised as part of the White House AI Action Plan, so subcategory wording and numbering may change. A mapping that was accurate when your team built its evidence register may no longer match. Record the crosswalk version and the AI RMF revision you used, subscribe to NIST's AI RMF page, and recheck every mapped row before each surveillance or internal audit cycle.

Who owns the controls, who prepares the evidence, and who certifies

The organization owns the AI management system in full. Its leadership sets the AI policy, decides how much AI risk it will accept, chooses which Annex A controls apply and which do not, and signs off on the statement of applicability. These decisions are business and governance decisions, and no outside party can make them on the organization's behalf without creating an accountability gap. Under the NIST AI RMF, the same people answer for the GOVERN function, including who is accountable for AI risk decisions.

Policy and risk tolerance: Executive leadership approves the AI policy and the level of AI risk the organization will accept.
Control decisions: The team documents which Annex A controls are necessary, and why each excluded control is excluded, in the statement of applicability.
Internal audit and management review: Staff run the internal audit and management review cycles required by the management system clauses.
Corrective action: Owners track nonconformities and show how they were fixed.
Ongoing operation: Product, data and security teams keep AI system records current after the audit.

Before engaging any partner, confirm which named executive will own each decision and that this person can defend the statement of applicability to an auditor.

Questions buyers ask before choosing between the two

Does aligning with the NIST AI RMF satisfy an ISO 42001 certificate request?

No. Aligning with the NIST AI RMF does not by itself satisfy an ISO/IEC 42001 certificate request. The NIST AI RMF is a voluntary framework with no certification, while ISO/IEC 42001 is a certifiable management system standard. A certification body audits the management system clauses, including internal audit under clause 9.2, management review under clause 9.3 and corrective action under clause 10.2, and it reviews your statement of applicability. Use the NIST AI RMF as useful input for your risk and governance work, then confirm with the certification body which evidence it will accept.

Can we run the NIST AI RMF and ISO 42001 together?

Yes, and many organizations do. The NIST-hosted crosswalk maps AI RMF subcategories to ISO/IEC FDIS 42001 clauses and Annex B guidance, which lets one set of AI risk records serve both. For example, GOVERN 2.3, on executive responsibility for AI risk decisions, maps to clauses 5.1 and 5.2 and the management review clauses. NIST states that inclusion in the crosswalk implies neither endorsement nor full coverage, so treat it as a starting point and check each mapping against your own evidence. A single risk process with one register and one owner usually works best.

How much effort does mapping the two frameworks take, and what drives it?

The effort depends on scope and evidence maturity more than on the framework text itself. The crosswalk covers 72 AI RMF subcategories (19 GOVERN, 18 MAP, 22 MEASURE and 13 MANAGE), so the work grows with how many of those your organization already evidences. The main drivers are the number of AI systems in scope, how many third parties supply models or data, how well your inventory and monitoring records exist today, and how many documents need rewriting into management system form. Ask any implementation partner for a written breakdown of these drivers before agreeing a fixed scope.

Which parts of the mapping and the management system can we handle internally?

Most of the work can be done internally if you have people with the right competence. Clause 7 requires you to determine and document the competence of people affecting AI performance, and clause 9.2 requires internal audit of the management system. Your team can own the AI policy, the inventory, the risk process and the audit program. An outside partner is most useful for preparing systems, controls and evidence, configuring tooling or running a readiness review. The organization keeps ownership of its controls, and the internal audit must be performed by people independent enough to test them.

Can an implementation partner issue an ISO 42001 certificate?

No. Only an independent certification body can issue an ISO/IEC 42001 certificate. ISO does not certify organizations, and accreditation of certification bodies is voluntary. A partner such as DeployOpen can plan, implement, document and prepare evidence for your AI management system, but it does not issue audit reports, attestations or certificates. Keep the implementation partner and the certification body separate so the auditor's independence is not in question. Check the certification body's scope and accreditation status before you sign.

Should we choose the NIST AI RMF or ISO 42001 first?

Choose based on what your customers, contracts or regulators ask for. If a buyer requests a certificate, ISO/IEC 42001 is the certifiable route, and the NIST AI RMF can inform the risk work underneath it. If no one requires a certificate, the NIST AI RMF may be the lighter starting point for governance, since it is voluntary and organized around Govern, Map, Measure and Manage. Many teams start with the NIST AI RMF to build risk practices and then add the ISO/IEC 42001 management system clauses where certification becomes necessary.

Do revisions to the NIST AI RMF change the crosswalk?

They can. NIST's page states that AI RMF 1.0 is being revised as part of the White House AI Action Plan, so subcategory numbering and text may change. ISO/IEC 42001 has no amendment or revision in progress according to ISO as of October 2026. Record the AI RMF version you mapped against, check NIST's crosswalk page for updates before each audit cycle, and ask your certification body how it will treat mappings built on a superseded framework version.

How DeployOpen scopes the mapping and prepares evidence

DeployOpen is a Chainscore Labs company that plans and implements self-hosted open-source platforms inside the client's own infrastructure. For ISO 42001 vs NIST AI RMF, we help teams map their existing AI governance work to both frameworks and prepare the records an independent certification body or assessor would review, while the organization keeps ownership of its controls.

How we scope

Engagements are fixed-scope, with the mapping boundary, AI systems covered and deliverables agreed in writing before work starts.
We confirm which NIST AI RMF subcategories and which ISO/IEC 42001 clauses and Annex A controls fall inside the written scope.
We agree with the client which crosswalk rows are reviewed as a starting point and which are checked against the organization's own records.

How we build

We set up the governance records a mapping depends on, such as AI system inventories, risk registers and management review minutes.
We build a traceable matrix linking each GOVERN, MAP, MEASURE and MANAGE subcategory to the evidence that supports it.
We test that evidence for gaps, including clause 9 internal audit and clause 10 corrective action, before anything is handed over.

How we hand over

You receive the mapping matrix, evidence index and documented gap list in formats your assessor and internal teams can read.
The crosswalk is delivered with its limits recorded, since NIST states that inclusion implies neither endorsement nor full coverage.
Ownership of policies, risk decisions and sign-off stays with your organization, and certification decisions rest with the independent certification body.

How we support

Ongoing support, upgrades and continuous testing of the governance tooling are available as a retainer.
Retainers can keep the mapping current when NIST revises AI RMF 1.0 or publishes new profiles.
Ask any provider to confirm in writing which parts of the mapping they prepare and which parts your organization or an auditor must own.

ISO 42001 vs NIST AI RMF