ISO/IEC 42001:2023 AI management systems

What it is

ISO/IEC 42001:2023 is the first international management system standard for artificial intelligence. It was published in December 2023 by ISO and IEC through their joint committee on AI, JTC 1/SC 42. It specifies requirements for establishing, implementing, maintaining and continually improving an AI management system, usually shortened to AIMS.

A management system standard is about the machinery of governance rather than any single AI system. It asks whether the organisation has set a policy, assigned responsibility, assessed its risks, chosen controls, measured whether they work and improved when they do not. ISO/IEC 42001 applies that model to AI, adding AI-specific requirements such as the AI system impact assessment and a set of AI reference controls.

The standard is paid and is sold by ISO and by national standards bodies, many of which have adopted it identically under their own prefix.

Who it applies to

Any organisation that develops, provides or uses AI systems can adopt it, in any sector and at any size. It applies only when an organisation chooses it, or when a customer or buyer makes certification a condition of doing business. The organisation sets the scope itself, which can be the whole organisation or a defined part, such as one product line or one business unit.

What it requires

ISO/IEC 42001 follows the ISO harmonised structure shared with ISO/IEC 27001 and ISO 9001. The requirements sit in clauses 4 to 10.

• Context (clause 4). Understand internal and external issues, the needs of interested parties, the organisation’s role with respect to AI, and the scope of the AIMS.
• Leadership (clause 5). Top management commitment, an AI policy, and assigned roles, responsibilities and authorities.
• Planning (clause 6). AI objectives and the three core assessments. These are the AI risk assessment, AI risk treatment, and the AI system impact assessment of consequences for individuals, groups and society. Risk treatment produces a Statement of Applicability.
• Support (clause 7). Resources, competence, awareness, communication and documented information.
• Operation (clause 8). Running the planned processes, repeating risk assessment, treatment and impact assessment at planned intervals and when significant changes occur.
• Performance evaluation (clause 9). Monitoring and measurement, internal audit and management review.
• Improvement (clause 10). Continual improvement, nonconformity and corrective action.

Annex A reference controls

Annex A lists 38 controls under nine control objectives. They cover policies related to AI, internal organisation, resources for AI systems, assessing the impacts of AI systems, the AI system life cycle, data for AI systems, information for interested parties, use of AI systems, and third-party and customer relationships.

The controls are reference controls, not a mandatory checklist. The organisation compares its risk treatment against Annex A, decides which controls are necessary, and justifies every inclusion and exclusion in the Statement of Applicability. It may also add controls from other sources. Annex B gives implementation guidance for each control, Annex C lists potential AI objectives and risk sources, and Annex D covers use of the management system across domains and sectors.

Certification

Certification is carried out by accredited certification bodies through a two-stage audit, followed by surveillance audits and recertification on the usual three-year cycle. ISO/IEC 42006:2025, published in July 2025, sets the additional requirements those bodies must meet, including auditor competence specific to AI. It supplements ISO/IEC 17021-1 and gives buyers a firmer basis for trusting a certificate.

Getting ready to certify

Readiness is the distance between the management system you run today and what the standard requires. Work through each clause and each Annex A control and ask two questions. Do you do what it asks, and can you show that you do? An auditor tests both, so practice without records counts as not yet ready.

The findings that recur most often are an incomplete picture of the AI in scope, governance that happens but leaves no reliable record, an internal audit and management review that have not yet run, and assessments done once and never repeated after the system changed.

Readiness checklist

• The AI systems in scope are identified, including embedded features and tools adopted informally by teams.
• An AI policy and defined roles and responsibilities are in place and current.
• AI risk assessment, risk treatment and AI system impact assessment run across the lifecycle, with owners and decisions recorded.
• A Statement of Applicability explains which Annex A controls apply and why, and the chosen controls are implemented.
• Documented information is controlled, so its integrity can be shown as well as its contents.
• Internal audit of the AIMS and management review are defined and have run at least once.

A readiness review, whether your own or an adviser’s, tells you how close you are. It does not certify you. Only the certification audit does that.

Key dates

The standard was published in December 2023 and the 2023 edition is current. ISO/IEC 42005, the companion guidance on AI system impact assessment, followed in May 2025, and ISO/IEC 42006 for certification bodies in July 2025.

How it relates to other frameworks

ISO/IEC 42001 is the management system spine that other frameworks often plug into. ISO/IEC 23894 gives deeper guidance on the AI risk process and ISO/IEC 42005 on the impact assessment. The NIST AI Risk Management Framework is commonly used as the working vocabulary for risk activities inside a 42001 management system, and the two are compatible.

For organisations in scope of the EU AI Act, 42001 does not create a presumption of conformity on its own. The Act relies on harmonised European standards for that. Its risk management, documentation, human oversight and monitoring practices overlap heavily with the Act’s requirements, so an existing AIMS is a strong base for that work. Organisations that already hold ISO/IEC 27001 or ISO/IEC 27701 can share the management system elements and focus new effort on the AI-specific requirements.

Working with it in Aicura

ISO/IEC 42001 is available in Aicura from the Pro tier. An administrator enables it in Settings under Frameworks, and chooses which governance policy reviews apply it under Reviews.

Once enabled, risk and impact assessments for each AI system and agent draw on it alongside any other frameworks you have turned on, producing rated risk items with suggested controls. Those risks and controls are published as a versioned set in Risks and Controls, where each control tracks whether it is applicable, implemented and tested. Governance policy reviews, such as the AI Risk Assessment Policy and the AI Incident Response Plan, return recommendations against the standard, and the Policy review report shows where each policy stands against the frameworks it was reviewed under. The AI Register, together with Discovery, which finds AI in use that is not yet on the register, keeps the scope of the management system current.

Aicura provides guidance and keeps the snapshots. It does not audit you, judge your readiness or certify you against the standard.