Choosing an ISO 42001 Certification Auditor: What to Look For

What an ISO 42001 Certification Auditor Actually Verifies

ISO/IEC 42001:2023 specifies requirements for an Artificial Intelligence Management System, or AIMS, published by ISO in December 2023. Certification against this standard confirms that an organization has established, documented, and operated a management system for AI risk, not that a technical security assessment or penetration test has been performed. An ISO 42001 certification auditor evaluates conformity to the standard's clauses and to the control objectives in Annex A, which cover areas such as AI system impact assessment, data management, transparency, and third-party or supplier relationships. Because ISO/IEC 42001 is a management-system standard, it does not itself prescribe specific technical implementations, such as how agent permissions should be configured. Instead, it requires the organization to define, document, and operate controls appropriate to its own AI system lifecycle, and the auditor assesses those controls against the organization's stated scope and objectives. This distinction matters when evaluating auditors, since the standard's flexibility places significant weight on how rigorously an individual auditor interprets and tests conformity.

Accreditation Is the Non-Negotiable Starting Point

A certification is only as credible as the accreditation behind it. Certification bodies must be accredited by a national accreditation body to issue recognized ISO/IEC 42001 certificates, and that accreditation's credibility depends on whether the accreditation body is a signatory to the IAF Multilateral Recognition Arrangement. ISO/IEC 17021-1 establishes the general requirements for bodies providing audit and certification of management systems, including baseline auditor competence requirements that apply across ISO management system standards. The International Accreditation Forum issues mandatory documents that guide accreditation bodies on scheme-specific competence and accreditation scope for individual standards, including newer ones like ISO/IEC 42001. In practice, this means enterprises should independently verify the accreditation body's IAF MLA signatory status rather than accepting it from a certification body's own marketing materials, and should confirm that ISO/IEC 42001 is explicitly listed within the certification body's current accredited scope. Accreditation held for ISO/IEC 27001 or another standard does not by itself indicate competence to certify an AI management system.

AI-Specific Competence Beyond Generic Management-System Auditing

ISO/IEC 17021-1 competence requirements are written generically across management system standards and do not by themselves guarantee AI-specific expertise. Accreditation bodies typically require certification bodies to demonstrate scheme-specific technical competence before a new standard such as ISO/IEC 42001 is added to their accredited scope, but the depth of that competence at the individual lead auditor level varies. Enterprises should request direct evidence of the assigned lead auditor's AI or machine learning risk training and prior audit experience, rather than assuming competence from the certification body's general accreditation. Frameworks such as the NIST AI Risk Management Framework, which structures AI risk into govern, map, measure, and manage categories, are sometimes used as a comparative reference point for AI governance maturity. This framework is separate from ISO/IEC 42001 certification requirements, but an auditor's familiarity with it can indicate a working understanding of AI-specific risk categorization that goes beyond a generic management-system audit background.

Where Runtime AI Governance Fits Into Audit Scope

ISO/IEC 42001 audits follow the standard sequence used for other management system standards: a Stage 1 readiness and documentation review, a Stage 2 implementation audit, and subsequent surveillance audits. Runtime elements of AI governance, including agent permissions, autonomous action boundaries, and operational monitoring, fall under the organization's internal AIMS scope. Auditors are expected to verify evidence that these controls operate as defined, not simply that policies describing them exist. This is a meaningful evaluation point when shortlisting auditors: ask whether the audit approach extends to sampling operational logs, access control configurations, and agent permission settings, or whether it relies primarily on reviewing policy and procedure documents. Audit scope statements should also be reviewed directly to confirm whether runtime AI system boundaries, such as autonomous agents or third-party model integrations, are explicitly included in the certified AIMS scope, since gaps here can leave real operational risk unassessed even after certification is granted.

Distinguishing a Substantive Audit From a Checklist Review

• Audit depth should be evaluated on whether findings address substantive control implementation and risk treatment evidence, not just documentation completeness.
• Confirm the Stage 1 and Stage 2 audit plan explicitly addresses the organization's AI system lifecycle as defined in its own AIMS scope statement.
• Clarify surveillance audit commitments upfront, since AIMS controls, particularly runtime and operational ones, may evolve faster than annual audit cycles.
• Treat scope mismatches between the certification body's accredited scope and the organization's actual AI use cases as a certification timeline risk to resolve early.
• Distinguish certification bodies with broad multi-standard accreditation from those with demonstrated, scheme-specific ISO/IEC 42001 competence.