A customer asks whether you are "ISO 42001 certified." Or a vendor sends you a certificate and presents it as the answer to your AI questions. Either way, the useful question is the same: what did the auditor actually examine, and what does the certificate let anyone conclude?
The short answer is that ISO/IEC 42001:2023 certifies a management system. It says an organization has an audited, working way of governing the AI systems it develops, provides, or uses. It does not say that any particular model is accurate, fair, or safe.
This article sets out what sits inside that boundary and what sits outside it. It ends with a short checklist for reading a certificate without overstating it in either direction.
What the standard is
ISO/IEC 42001:2023 is the international standard for an artificial intelligence management system, usually shortened to AIMS. ISO and IEC published it in December 2023. It was developed by ISO/IEC JTC 1/SC 42, the subcommittee of their joint technical committee that works on artificial intelligence.
It uses the same harmonized structure as ISO/IEC 27001 and ISO 9001. Clauses 4 to 10 hold the requirements, and an annex holds reference controls. If you already run an information security management system, the shape will be familiar. Our plain-terms explanation of an ISMS covers the management system idea itself, and all of it carries over.
The requirements, in outline:
Clause 4, context. The organization's situation, its interested parties, the scope of the AIMS, and its roles with respect to the AI systems in scope.
Clause 5, leadership. Top management commitment, an AI policy, and assigned roles and responsibilities.
Clause 6, planning. AI risk assessment, AI risk treatment, AI system impact assessment, and AI objectives.
Clause 7, support. Resources, competence, awareness, communication, and documented information.
Clause 8, operation. Running the planned processes, and repeating risk and impact assessments when things change.
Clause 9, performance evaluation. Monitoring, internal audit, and management review.
Clause 10, improvement. Nonconformity, corrective action, and continual improvement.
Annex A lists 38 reference controls grouped under nine control objectives. Annex B gives implementation guidance for those controls. Annexes C and D are informative: one lists possible AI objectives and risk sources, the other discusses using the AIMS across domains and sectors.
What a certificate attests
A certification body issues the certificate after auditing the organization against the clause 4 to 10 requirements. In plain terms, it attests that within the stated scope, at the time of audit:
The organization has defined the boundary of its AIMS and its role toward each AI system inside it, for example as a developer, a provider, or a user.
Top management has approved an AI policy and assigned responsibility for the system.
There is a documented method for assessing AI risks and the impacts of AI systems, with criteria the organization set, and it has been applied.
Risk treatment decisions are recorded, including a Statement of Applicability that says which Annex A controls apply and why any are excluded.
The system is monitored, internally audited, and reviewed by management, and nonconformities are corrected.
The auditor tests that these exist and operate. That means sampling records: risk assessments, impact assessments, the Statement of Applicability, training records, internal audit reports, management review minutes, and corrective actions. A well-written policy with no records behind it does not pass.
What a certificate does not attest
This is where most misreadings happen, in both sales decks and procurement reviews.
It is not a product certification. The standard sets no accuracy threshold, no bias metric, and no test suite a model must pass. It requires the organization to decide how it will verify and validate its AI systems, and to follow that decision. The quality of the decision is the organization's own.
It is not a finding of legal compliance. The standard requires the organization to identify the legal and regulatory requirements that apply to it and to address them. A certification audit checks that this identification happened and feeds the system. It is not a legal opinion. Whether a system meets a specific law is a question for counsel.
It covers only what is in scope. A certificate scoped to one product says nothing about the company's other products. A scope that covers only the use of third-party AI tools says nothing about models the company builds.
It describes a period, not a promise. The certificate reflects what the auditor saw. Surveillance audits recheck the system during the certificate's life, but they sample. They do not watch every change.
How certification works
The process follows the same model as other ISO management system certifications.
Stage 1 reviews the documented system: scope, policy, risk and impact assessment methods, the Statement of Applicability, and readiness for the next stage.
Stage 2 tests whether the system is implemented and operating, by interviewing people and sampling records.
The certificate is typically valid for three years. Surveillance audits take place in the intervening years, and a recertification audit precedes expiry.
Certification bodies generally expect at least one internal audit and one management review to be complete before stage 2. That is the part teams most often underestimate. You cannot audit a system that has not yet run through one full cycle.
Accreditation matters as much as it does for ISO/IEC 27001. Certification bodies are accredited by national accreditation bodies, against ISO/IEC 17021-1 for management system certification generally. ISO/IEC 42006 adds requirements specific to bodies that audit and certify AI management systems. A certificate from an unaccredited body is a document, not an assurance.
Reading a certificate someone sends you
When a vendor offers an ISO/IEC 42001 certificate as evidence, read it with the same care you would give a SOC 2 report. Check these, in order:
- Not done:
Standard and edition. It should name ISO/IEC 42001:2023.
- Not done:
Scope statement. Which products, services, sites, and AI systems are covered. Does it include the system you are buying?
- Not done:
Role. Is the vendor certified as a developer or provider of AI systems, or only as a user of AI tools? These are very different scopes.
- Not done:
Certification body and accreditation. Is the body accredited for this standard? Many certification and accreditation bodies publish registers you can check.
- Not done:
Dates. Issue date, expiry date, and whether the certificate is current.
- Not done:
Statement of Applicability. Ask for it, or a summary of exclusions. Some organizations share it under NDA. An exclusion that matters to your use case is worth a follow-up question.
The certificate does not replace your own due diligence on how the vendor handles your data. It tells you there is a governed process behind the answers they give you.
Where it sits next to other frameworks
ISO/IEC 42001 does not replace ISO/IEC 27001. Information security risks to AI systems are real, but 27001 says nothing about impact on the people an AI system makes decisions about. 42001 covers that, and does not try to be a full security control set. Many organizations run the two together.
The NIST AI Risk Management Framework is a voluntary US framework. It is useful for structuring AI risk work, and it is not certifiable. Laws such as the EU AI Act set legal obligations of their own, which phase in over time; if one may apply to you, read the current text and take advice rather than assuming a certificate covers it. Our ISO 42001 framework page summarizes the standard's requirements, and the glossary defines the terms used here.
How TrustLynce helps
TrustLynce maps ISO/IEC 42001 requirements alongside your existing frameworks, so controls you already operate for ISO/IEC 27001 or SOC 2 are reused rather than rebuilt (compliance automation). The readiness dashboard shows gaps per framework before your stage 1 audit. The certification decision stays with your certification body.