Why organizations pursue ISO 42001.
- The AI-governance question. Enterprise buyers increasingly ask how vendors govern AI before approving them. 42001 is the answer procurement recognizes.
- EU AI Act alignment. The AIMS maps closely to the Act’s risk-management, data-governance, and documentation duties: useful evidence as enforcement ramps.
- Board & investor assurance. A certified management system gives directors and investors a defensible answer on how AI is overseen.
- Independent, registered proof. A CB-issued certificate resolves on the IAF database: third-party verification, not a self-claim.
- A first-mover trust signal. As the first certifiable AI standard, 42001 still differentiates: a credible governance signal ahead of the market.
Context
A 42001 certificate does not certify that any model is safe, unbiased, or correct. It demonstrates that you have a management system capable of governing AI risk across the lifecycle. For most organizations, that distinction is exactly what buyers and regulators are asking them to prove.
One management system.
Every AI risk it governs.
- AI Risk Management
- Impact Assessment
- Data Governance
- System Lifecycle
- Human Oversight
- Transparency
- Evidence & Logging
- Audit & Improvement
ISO 42001 exists because buyers and regulators need confidence that AI risk, impact assessment, data governance, and the model lifecycle operate as one accountable system.
What ISO 42001 actually is, in plain English.
An accredited certificate that you operate a working AI Management System — it certifies the system, not any single model. Excellent evidence of governance maturity; not proof a model is safe.
The standard, in full
ISO/IEC 42001:2023 is the first internationally-recognized management system standard for artificial intelligence, published on 18 December 2023 by ISO/IEC Joint Technical Committee 1, Subcommittee 42. It is structured like ISO 27001:2022: ten High-Level Structure clauses describing the management system, a normative Annex A of 38 controls under 9 objectives, and informative annexes (B implementation guidance, C risk sources, D sectoral use). If you have ever stood up an ISMS, the architecture is familiar within ten minutes.
The unit of certification is the AIMS: not a model, not a product. You define a scope, demonstrate that the AIMS addresses the Annex A controls you declare applicable in your Statement of Applicability, then operate it long enough for an accredited Certification Body to sample evidence at a Stage 2 audit. The certificate is issued by the CB, not ISO; the CB is accredited by an IAF MLA signatory (ANAB, UKAS, DAkkS, JAS-ANZ) and the certificate number resolves on IAF CertSearch.
What it is not: a product certification. 42001 does not certify that a model is safe, unbiased, or correct. It certifies you have a management system capable of governing AI risk across the lifecycle: excellent evidence of governance maturity, and almost none of system-level conformance. Use it for what it is.
AI governance, answered.
A qualified prospect wants to buy. Then their risk team asks how you govern AI.
You point to your 42001 certificate. The governance review opens instead of stalling.
They sample your AIMS evidence: impact assessments, lifecycle records, logs. It is already there.
Procurement has independent, registered assurance to proceed. No questionnaire marathon.
The deal advances. Answered.
All cleared. A live 42001 certificate compresses the AI-governance review from weeks to a link.
A live 42001 certificate compresses this review from weeks to a link.
A boutique firm, not a compliance factory.
- Canned SOW, identical for every client
- Junior associate does the work, partner sells it
- Scope padded with line items you don't need
- "You don't qualify" if you don't fit the template
- Surprise change-orders mid-engagement
- SOW built around your business, not a template
- Senior practitioner from day one: no junior hand-off
- We cut unnecessary cost: you pay for what moves the audit
- We edit SOW language together until both sides agree
- Fixed fee, written: no surprise change-orders
We never say "you don't qualify."
Our approach
If a framework isn't right for you yet, we'll tell you: and tell you what is. When it is right, we sit down and shape the statement of work to your reality: your stack, your team, your timeline, your budget. The deliverable bends to your business. You do not bend to a canned deliverable.
Does this apply to you?
A quick filter we run on every intake call. The more of these that point to “yes,” the more likely ISO 42001 belongs on your roadmap inside twelve months.
ISO 42001 fit check
AIMS designed.
Not yet operating.
Evidence begins
accumulating.
Internal audit confirms it works.
Certified.
Surveillance begins.
Stage 1 reviews the AIMS on paper: scope, AI policy, Statement of Applicability, and the Clause 6 risk and impact assessments. It proves the system is designed — not that it runs.
The AIMS goes live. Controls must now operate: impact assessments, lifecycle records, and event logs are captured continuously — not assembled before the audit.
Internal audit and management review test effectiveness across the period. Nonconformities are raised and closed — the loop is turning before the CB arrives.
Stage 2 passes. The CB issues a certificate that resolves on the IAF database, then returns for annual surveillance across the three-year cycle.
A first-time AIMS, realistically: 9 to 15 months.
From kickoff to a certificate that resolves on the IAF database. The chokepoint is rarely the audit; it is standing the AIMS up and operating it long enough to show evidence of effectiveness.
Mo 0 – 1Gap & scope
Mo 1 – 4Build the AIMS
Mo 4 – 8Operate & evidence
Mo 7 – 9Stage 1 audit
Mo 9 – 15Stage 2 + certificate
Built on your 27001 fabric. Senior partner from day one.
Partner-led from kickoff through Stage 2. Built on top of your existing management-system fabric. Audit-room delivered before the CB arrives.
Methodology detail
Most consultants hand you a policy template pack and a portal, then stop calling. We don’t. Every Nexurion 42001 engagement is led by a senior practitioner: the person on the engagement letter is the one reading your impact assessments, sitting in your management reviews, and on the call when the CB samples evidence. Read our methodology.
If you already operate an ISO 27001 ISMS, we build the AIMS on top of it: shared Clauses 4–10, one internal-audit program, one management review. We instrument the Clause 6 assessments and Annex A evidence from day one and hand the CB a read-only audit-room with every artifact pre-mapped to the clause or control it satisfies. More on our evidence model »
The goal is not only a certificate. It is an AIMS that is sustainable: cheaper and lower-effort to operate in year two than in year one, and durable across surveillance audits as your AI footprint grows. See engagement outcomes.
Independent of the certification body: by design.
We are not a Certification Body and we do not issue 42001 certificates. That independence matters: a readiness partner whose firm also certifies has a conflict no engagement letter can paper over. We work alongside the IAF-accredited CB of your choice, or introduce you to bodies whose AI competence we’ve verified. Certification bodies »
Six places a 42001 goes sideways.
After standing these up, the failure modes are consistent. Almost none are technical: they are about planning, scope, and ownership.
Certify it, don’t decorate it.
Why it happens
Annex A first, planning never.
Why it happens
Certify every AI system at once.
Why it happens
Models nobody owns.
Why it happens
“Same ISMS, new logo.”
Why it happens
A body that doesn’t understand AI.
Why it happens
Certification bodies we’ve walked clean engagements with.
How we work with certification bodies
We are CB-neutral by policy and refer on fit, AI competence, and IAF accreditation: never on referral fees (we accept none). After multiple engagements you learn which bodies field auditors who actually understand model lifecycles, and which treat 42001 as ISO 27001 with a new cover. We’ll introduce you to bodies whose AI competence and accreditation we’ve verified. Buyers and regulators check the accreditation behind the certificate: we make sure it holds up.
If you already hold a CB relationship from your ISO 27001 certificate, we’ll sit with their lead auditor and align scope, evidence format, and Stage 1 / Stage 2 cadence: often the same body can certify both. We’ve never had to disengage from a body mid-audit.
ISO 42001 against the rest of the AI stack.
42001 is rarely the only AI framework your buyers and regulators name. The good news: most of the work overlaps. The bad news: the gaps are not where you think.
ISO 42001 against the rest of the AI stack.
42001 is the core, not the whole stack.
| Framework | Overlap with 42001 | What you still need to do |
|---|---|---|
| NIST AI RMF | ~70%: Govern / Map / Measure / Manage map onto the AIMS. | Nothing to certify: the RMF is voluntary. Use it as the risk backbone; 42001 makes it auditable. |
| EU AI Act | ~60%: AIMS evidence supports the Act's duties. | Conformity assessment, CE marking, and risk-tier-specific obligations the AIMS does not itself satisfy. |
| ISO 27001 | ~55%: shared HLS, internal audit, management review. | Information-security Annex A controls; the AIMS adds the AI-specific control set. |
| ISO 23894 | ~50%: feeds the Clause 6 AI risk assessment. | Guidance only, not certifiable: it informs the AIMS rather than replacing it. |
| OWASP LLM Top 10 | ~35%: touches Annex A lifecycle & security. | Application-layer testing: prompt injection, insecure output handling, model DoS. |
| SOC 2 | ~25%: risk & change-management evidence carries over. | Trust Services Criteria security controls and a CPA attestation. See SOC 2 » |
The ten clauses & Annex A controls.
42001 is built on the ISO High-Level Structure. Clauses 4–10 are the management system an auditor tests; Annex A is the AI-specific control set you justify in your Statement of Applicability. Every Annex A control defaults to applicable: you justify exclusions, not inclusions.
| Clause | Title | What an auditor expects |
|---|---|---|
| 4 | Context of the organization | Internal/external issues, interested parties, AIMS scope statement, inventory of AI systems with role (developer/deployer) and lifecycle stage. |
| 5 | Leadership | AI policy signed by top management, assigned AIMS roles, evidence of leadership engagement: minutes, objectives, resourcing. |
| 6 | Planning | The hardest clause. AI risk assessment, AI system impact assessment (per ISO 42005), Statement of Applicability for the 38 controls, measurable AI objectives. |
| 7 | Support | Resources, AI competence (training records), awareness, internal/external communication, documented information. |
| 8 | Operation | Operational planning, AI risk treatment in practice, impact assessments, change control over models & data. |
| 9 | Performance evaluation | Monitoring & measurement, internal audit program, management review: quarterly minimum. |
| 10 | Improvement | Nonconformity & corrective action, continual improvement: a track record of NCs closed. |
Annex A defines 38 controls across 9 objectives. Click each objective for the controls a CB will sample.
Annex A.4Resources for AI systems
A documented inventory of the resources the AIMS needs, and an explicit assessment of resources required across the lifecycle. Most clients confuse this with capacity planning: it is accountability.
Annex A.5AI system impact assessment
The most-cited objective in EU AI Act conversations. A.5 demands a process for assessing impacts on individuals, groups, and society, with companion standard ISO 42005 giving the methodology. CBs sample these heavily.
Annex A.6AI system lifecycle: the heart of the standard
Ten controls covering objectives, design & development, verification & validation, deployment, operation & monitoring, technical documentation, and event logging. Mature MLOps already does ~70% of this; the gap is documentation.
Annex A.7Data for AI systems
Six controls on the data that feeds models: provenance, quality, and preparation, plus the governance an auditor expects before trusting any output.
Annex A.8–A.10Information, use & third parties
The outward-facing controls: what you tell users and interested parties, how the system is responsibly used, and how you govern the third-party AI you build on.
42001 in its second year.
42001 was published in December 2023, and adoption accelerated through 2025 as the EU AI Act timeline firmed up. Four things to watch in 2026:
- EU AI Act harmonization. CEN-CENELEC is drafting harmonized European standards for the Act. 42001 is widely expected to inform them: a 42001 AIMS is strong positioning, not a substitute for conformity assessment.
- CB accreditation maturing. More IAF-accredited bodies now offer 42001 and auditor AI competence is rising. The gap between a credible certificate and a rubber-stamp is narrowing: but still real.
- ISO 42005 in force. The AI system impact-assessment standard gives Clause 6 and Annex A.5 a concrete methodology. CBs are sampling impact assessments more rigorously as a result.
- Buyer demand. Enterprise and public-sector buyers increasingly name 42001 in AI-governance questionnaires. See our AI governance practice »
Read our deeper take in Nexurion Field Notes.
42001 is rarely the only AI framework.
What we typically scope alongside it: in order of how often the question comes up.
Frequently asked.
Are we certified when the report is issued? +
How long does the first Type 2 take? +
What does it cost? +
Should we run SOC 2 and ISO 27001 at the same time? +
Can we self-attest? +
Does our existing HITRUST or ISO 27001 cert satisfy a SOC 2 ask? +
What’s the difference between SOC 1, SOC 2, and SOC 3? +
Field notes on SOC 2.
Pieces from Nexurion Field Notes directly relevant to the standard.