ISO 42001 certification is an independent audit result saying an organization runs an AI management system that meets ISO/IEC 42001:2023. It matters most to companies that build, sell or embed AI, because their buyers ask for evidence of how that AI is governed. A company that only uses other people's AI tools usually has a cheaper way to answer the same question.
One disclosure first. I'm not an auditor or a lawyer, and I haven't taken an organization through an ISO 42001 certification. I work inside an ISO 27001 program, so this is a read of the standard and published material from that side. Every fact below links to its source, and I checked the pages on November 14, 2026.
My view: if you already run an ISO 27001 information security management system, ISO 42001 is a delta, not a second program. Both follow ISO's Harmonized Structure, clauses 4 to 10 (context, leadership, planning, support, operation, performance evaluation, improvement), so the management system machinery you already built carries over: scope, policy, internal audit, management review, corrective action and document control. The new work is narrow. It's (1) an inventory of the AI systems in scope and your role for each one, (2) an AI system impact assessment for the people affected, and (3) the AI-specific Annex A controls across the system lifecycle and its data. Certification is worth it when buyers or regulators ask for it, which is mostly organizations that build or sell AI. For a company that only uses AI tools, a good AI acceptable use policy with receipts, plus a risk register, may be enough.
What the standard actually is
ISO/IEC 42001:2023 was published on December 18, 2023, as edition 1. ISO describes it as a standard that sets out requirements for establishing, implementing, maintaining and continually improving an AI management system, for organizations that provide or use AI-based products or services, of any size or sector. It takes a Plan-Do-Check-Act approach to AI risks and opportunities.
Its introduction says it uses the harmonized structure shared with other management system standards, with identical clause numbers, clause titles, common terms and core definitions. Advisera's walkthrough lists the clauses as the same seven you know from ISO 27001: context (4), leadership (5), planning (6), support (7), operation (8), performance evaluation (9) and improvement (10).
Annex A holds the reference controls. By the count in third-party breakdowns from Konfirmity and Modulos, there are 38 controls in nine groups, numbered A.2 to A.10.
- A.2 Policies related to AI
- A.3 Internal organization
- A.4 Resources for AI systems
- A.5 Assessing impacts of AI systems
- A.6 AI system life cycle
- A.7 Data for AI systems
- A.8 Information for interested parties of AI systems
- A.9 Use of AI systems
- A.10 Third-party and customer relationships
As in ISO 27001, the organization decides which controls its AI risk treatment needs and can add others. I only had ISO's public preview of the table of contents, not the full text, so I'm leaning on those third-party counts for the control numbers. If your scope depends on them, read the standard itself.
The delta table
This is how I'd map an existing ISO 27001 system onto 42001. The mapping is my judgment, and your auditor may slice it differently.
| ISO 27001 artifact you have | What it becomes for 42001 | What's new |
|---|---|---|
| ISMS scope statement | AIMS scope statement, often the same organizational boundary | A list of the AI systems in scope and your role for each: whether you develop it, provide it to others, or use it |
| Information security policy | AI policy, covered by Annex A group A.2 | Stated AI objectives and intended uses. My rule, gate, receipt test applies cleanly here |
| Risk assessment and treatment | Add AI risks and your control selection against Annex A | The AI system impact assessment (clause 6.1.4, repeated in 8.4) covering effects on individuals, communities and society |
| Asset and data inventory, classification | Extend to AI systems and their data | Annex A group A.7: data for AI systems, which is where provenance, quality and preparation of data get controls |
| Secure development and change management | Extend to the AI system life cycle | Annex A group A.6: life cycle controls specific to AI |
| Supplier management | Add AI vendors and models you depend on, plus your customers' expectations | Annex A group A.10, and shadow AI use that bypasses your register |
| Awareness and competence records | Add AI competence and responsible use | Annex A groups A.4 and A.9: resources and use of AI systems |
| Incident and communication procedures | Extend to AI incidents and what you tell interested parties | Annex A group A.8: information for interested parties |
| Internal audit, management review, corrective action, document control | Same machinery, with AIMS items added to the agenda | Nothing structurally new |
The last row is the point of the thesis. Most of the effort in an ISO program is the machinery, and you've already paid for it. My three-year cost view of ISO 27001 shows where that effort sits.
The three new pieces
1. The AI system inventory, with your role
You can't scope what you haven't listed. The inventory names each AI system in scope, who owns it, and what your organization's role is for it. Building a model, offering it to customers and using someone else's model inside your product are different positions, and they pull in different controls. I'd start from observed use, not procurement records, because use runs ahead of policy.
2. The AI system impact assessment
Clause 6.1.4 asks the organization to set up a method for assessing the possible consequences of its AI systems. As one summary of the clause puts it, it covers the full life cycle, including intended use and foreseeable misuse, and looks at effects on individuals, communities and society. The results are documented and feed the AI risk assessment. This is the piece with no ISO 27001 equivalent, because a security risk register asks what could harm the organization, and this asks what the system could do to other people.
3. The AI-specific Annex A controls
Lifecycle and data are where AI differs most from ordinary software, so groups A.6 and A.7 are where I'd expect the real new work. The same discipline applies here as with any policy. For each control you select, ask what rule it states, what gate enforces it, and what receipt proves it ran. If you can't evidence a control, expect the auditor to treat it as missing. An auditor's Stage 1 review, per DQS's description, looks at your documentation, AI impact assessments, management review results and internal audits to judge readiness.
How certification works
Organizational certification comes from an audit by a certification body. Per DQS, an ISO 42001 certificate is valid for three years, key system components are audited at least once a year in surveillance audits, and recertification happens before expiry. Treat that as one body's description and confirm the cycle in your own contract.
In the US, the accreditation body for these certification bodies is ANAB. ANSI's blog says ANAB launched its ISO/IEC 42001 accreditation program in January 2024. Schellman announced on September 24, 2024 that it was the first body accredited by ANAB for the standard. SGS followed on April 29, 2025, and DQS on February 26, 2026.
A related standard, ISO/IEC 42006:2025, was published in July 2025. It adds requirements, beyond ISO/IEC 17021-1, for bodies that audit and certify AI management systems, so that audits are consistent. Its practical use for you is as a question to ask a certification body: which accreditation do you hold for this standard?
Some larger companies publish their own certification status. As of November 14, 2026:
- Microsoft's compliance page (last updated September 11, 2026) lists AI services in scope, including Microsoft Foundry, Microsoft Security Copilot and GitHub Copilot, and points to its Service Trust Portal for scope statements. Its wording also mentions progress toward certification, so read the scope statement before citing it.
- AWS announced on November 25, 2024 a certification from Schellman covering Amazon Bedrock, Amazon Q Business, Amazon Textract and Amazon Transcribe. A February 26, 2026 post reports a first surveillance audit completed in November 2025 with no findings.
- Anthropic announced on January 13, 2025 certification issued by Schellman Compliance, an ANAB-accredited body.
These are vendors that build and sell AI, which fits the thesis.
Individual certificates are not organizational certification
This is where search results get muddy. Many pages ranking for “ISO 42001 certification” sell training. PECB's ISO 42001 page, for example, offers Foundation, Lead Implementer and Lead Auditor tracks, with the exam and certification included. Those credentials are for a person, and they say that person has learned to implement or audit an AI management system.
They don't certify your company. If a buyer asks whether you're ISO 42001 certified, a staff member's Lead Implementer certificate isn't the answer. The answer is an audited certificate for your management system, with a stated scope, from a certification body. The individual credential can still be useful, for example if you want someone on staff who can run the internal audit or speak the auditor's language.
Is it worth it?
I'd run the decision in this order.
- Has a buyer, partner or regulator asked for it? If yes, and the deal is large enough, that's the answer. Ask what evidence they accept, since some questionnaires want the certificate and others settle for a described program.
- Do you build, sell or embed AI? If yes, expect the question to arrive. Certification gives you one artifact to hand over in place of a long questionnaire every time.
- Do you already run ISO 27001? If yes, the incremental cost is lower, because the machinery exists. If no, you're building the machinery and the AI layer together, which is a bigger project. My ISO 27001 vs SOC 2 comparison covers how to tell which attestation your buyers actually want.
- Do you only use AI tools? Then start smaller: an acceptable use policy where every rule has a gate and a receipt, a register of the AI tools in use, and a risk register. If a customer later asks for certification, that work is not wasted, because it is the same evidence the audit would want.
On regulation, I'm not a lawyer, and I'd be careful about assuming certification answers a legal question. The European Commission's standardization FAQ, last updated March 10, 2026, says ISO/IEC 42001 helps organizations set up an AI management system, but that its goals and definitions do not match the EU AI Act's quality management system requirements. The Commission asked CEN and CENELEC for separate standards aimed at the Act, and the first has arrived. EN 18286:2026, a quality management system standard written for the Act, is cited in Commission Implementing Decision (EU) 2026/2274, adopted October 12, 2026 and in force from October 14. CEN-CENELEC's JTC 21 says it gives providers of high-risk AI systems a presumption of conformity with Article 17, and that using it is voluntary. It's a separate standard from ISO/IEC 42001, so don't tell anyone a 42001 certificate gets them covered.
What it costs, as components
I'm not going to put a dollar figure on it, because the published ranges I found are marketing pages, not data. The cost has three components, and you can price each one yourself.
- Certification body audit days. Stage 1, Stage 2, annual surveillance audits and recertification. Get quotes from accredited bodies against your scope, and price all three years.
- Internal time. The impact assessments, the inventory, the lifecycle and data controls, and the evidence collection. If you run ISO 27001, the shared machinery is mostly already staffed.
- Tooling and outside help. Optional, and the part most sensitive to how much you already have.
Frequently asked questions
What is ISO 42001 certification?
It's an independent audit result showing that an organization runs an AI management system that meets ISO/IEC 42001:2023, the international management system standard for AI. A certification body audits the organization, not a single product or person. The certificate is valid for three years with annual surveillance audits, according to one certification body's published description.
Is ISO 42001 worth it?
It's worth it when buyers or regulators ask for it, which in practice means organizations that build, sell or embed AI. If you only use AI tools, a clear AI acceptable use policy with enforcement and records, plus a risk register, may cover what your customers need. Ask your five biggest buyers what evidence they want before you start.
What is the difference between ISO 42001 and ISO 27001?
ISO 27001 certifies an information security management system. ISO 42001 certifies an AI management system. Both use ISO's harmonized structure (clauses 4 to 10), so scope, policy, internal audit, management review and corrective action carry over. The new work in 42001 is an AI system inventory with your role for each system, an AI system impact assessment, and the AI-specific Annex A controls.
How long does ISO 42001 certification take and what does it cost?
I don't have a trustworthy number, so I won't invent one. Time depends on how much of an ISO 27001 style system you already run. Cost comes in three parts: the certification body's audit days, your team's internal time, and any tooling or outside help. Get quotes from accredited certification bodies for your scope, and price the full three-year cycle.
Is an individual ISO 42001 certificate the same as organizational certification?
No. Lead Implementer, Lead Auditor and Foundation certificates are credentials for a person who passed a course and exam. They don't certify your organization. Organizational certification comes from an audit by a certification body, and the result is a certificate for your management system and its scope.
Does ISO 42001 certification satisfy the EU AI Act?
I'm not a lawyer, so check with one. The European Commission's standardization FAQ, last updated March 10, 2026, says ISO/IEC 42001 helps an organization set up an AI management system, but that its goals and definitions do not match the AI Act's quality management system requirements. The first harmonized standard for the Act, EN 18286:2026 on quality management systems, took effect in October 2026, and it's a separate standard from ISO/IEC 42001. So treat certification as useful governance evidence, not as proof of AI Act compliance.
Sources
All pages checked November 14, 2026.
- ISO, ISO/IEC 42001:2023 standard page (published 2023-12-18)
- ISO, ISO/IEC 42006:2025 standard page (published July 2025)
- ISO/IEC 42001:2023 public preview, table of contents and introduction (via iTeh Standards)
- Advisera, ISO 42001 requirements: clauses and structure (published November 14, 2025; modified February 10, 2026)
- Konfirmity, ISO 42001 controls (third-party summary of Annex A)
- Modulos, ISO 42001 Annexes A to D (third-party summary of Annex A)
- Cyberday, ISO 42001 clause 6.1.4 summary (third-party)
- DQS, ISO 42001 certification (three-year validity, annual surveillance, audit stages)
- ANSI Blog, ISO/IEC 42001 AI management systems (ANAB program launched January 2024)
- Schellman, first ISO 42001 ANAB-accredited certification body (September 24, 2024)
- SGS, ANAB accreditation for ISO/IEC 42001 (April 29, 2025)
- DQS, ANAB accreditation for ISO/IEC 42001 (February 26, 2026)
- Microsoft Learn, ISO/IEC 42001 offering (last updated September 11, 2026)
- AWS, AWS achieves ISO/IEC 42001:2023 accredited certification (November 25, 2024)
- AWS, first surveillance audit for ISO 42001:2023 with no findings (February 26, 2026)
- Anthropic, Anthropic achieves ISO 42001 certification for responsible AI (January 13, 2025)
- PECB, ISO 42001 training and certification tracks
- European Commission, Understanding standardisation and the AI Act, FAQ (last updated March 10, 2026)
- European Commission, Implementing Decision (EU) 2026/2274 of 12 October 2026 on harmonised standards for high-risk AI systems
- CEN-CENELEC JTC 21, EN 18286 becomes the first harmonised standard under the EU AI Act (October 14, 2026)