ISO/IEC 42001: the certifiable standard for managing AI responsibly
ISO/IEC 42001 is to AI what ISO 27001 is to information security: a management-system standard an organisation can be certified against. Published in December 2023, it is becoming the credential enterprise buyers and regulators ask for from AI vendors and from organisations deploying AI at scale. This guide explains the structure, the controls and the certification path, and how it fits with ISO 27001 and the EU AI Act.
ISO 42001 asks an organisation to define which AI systems it develops, provides or uses, to understand the risks and impacts of each, to decide and document the controls, and to operate the whole thing as a managed, audited, improving system. The familiar clauses 4 to 10 apply: context, leadership, planning, support, operation, performance evaluation and improvement. What is new is the subject matter and two assessments: an AI risk assessment and an AI system impact assessment that looks at consequences for individuals, groups and society.
Annex A’s 38 controls cover AI policies, internal organisation, resources for AI systems, assessing impacts, the AI system life cycle, data for AI systems, information for interested parties, use of AI systems and third-party relationships. Because the structure mirrors ISO 27001, organisations already certified can integrate the two into one management system and one audit.
What the standard requires you to produce
Context and scope (clause 4)
Which AI systems and which roles (provider, developer, user, partner) are inside the AI management system, and the interested parties and their requirements, including legal ones.
AI policy and leadership (clause 5)
A top-level AI policy aligned with organisational objectives, with roles, responsibilities and authorities for AI assigned.
AI risk assessment and treatment (clause 6)
A repeatable process to identify, analyse and evaluate AI risks, select Annex A controls or others, and record them in a statement of applicability.
AI system impact assessment (clause 6 and control A.5)
For each AI system, an assessment of potential consequences for individuals, groups and society, considering fairness, safety, privacy, transparency and environmental effects, with results feeding risk treatment.
Resources and competence (clause 7)
Data, tooling, computing and human resources documented per system; competence of the people who build and operate AI; awareness across the organisation.
AI system life cycle (clause 8 and control A.6)
Objectives, requirements, design, verification and validation, deployment, operation, monitoring, event logging and decommissioning for each AI system, documented.
Data management (control A.7)
Provenance, quality, preparation and acquisition of data for AI systems, recorded.
Transparency and use (controls A.8 and A.9)
Information to users and affected parties, incident reporting, responsible-use objectives and intended-use boundaries.
Third parties (control A.10)
Allocation of responsibilities with suppliers and customers of AI systems and components.
Evaluation and improvement (clauses 9 and 10)
Monitoring, internal audit, management review and corrective action, with records.
The certification path
- Scoping and gap assessment. Inventory AI systems, decide the roles the organisation plays, assess against clauses and Annex A.
- Build the AIMS. AI policy, risk and impact assessment methods, statement of applicability, life-cycle and data documentation, controls operating.
- Internal audit and management review. Required before certification, with records.
- Stage 1 and Stage 2 audits. Documentation review, then implementation testing; integrated with ISO 27001 where both are held.
- Surveillance and recertification. Annual surveillance, recertification in year three.
Who it applies to
AI vendors and platform providers that need a credential for enterprise procurement; organisations deploying AI in regulated decisions such as credit, insurance, recruitment, healthcare and public services; and any organisation that must show EU AI Act readiness. It is designed to be used by providers, deployers and developers alike, at any size.
From scope to evidence, in one system
ISO 42001 as a native framework next to ISO 27001
All 38 Annex A controls and the clause requirements are carried with owners, status and evidence, sharing the ISMS records, risk methodology, internal audit and management review already used for ISO 27001, so one management system serves both.
AI system inventory and impact assessments
Each AI system is recorded with role, purpose, data, model and users, and its impact assessment is run in the Risk & Quantification module with the standard’s criteria and a history of decisions.
Statement of applicability generated
Control decisions and justifications produce the SoA, updated as systems and risks change.
Policies and awareness
The AI Policy Builder generates the AI policy set; the Policy Register tracks acknowledgement; the Awareness module records training, which closes clause 7.
Crosswalk to NIST AI RMF and the EU AI Act
ISO 42001 controls are mapped to AI RMF subcategories and to EU AI Act obligations, so the same evidence supports a certificate, a US customer questionnaire and an EU regulatory file.
Frequently asked questions
Do we need ISO 27001 first?
No, but it helps. The clauses are the same structure and most organisations integrate the two. An AI management system without information security underneath it would struggle with the security-related Annex A controls.
Does ISO 42001 certification mean EU AI Act compliance?
No. The Act has specific legal obligations by risk tier that a certificate does not discharge. But a certified AIMS produces most of the risk-management, documentation, logging and human-oversight evidence the Act requires, and the European Commission has signalled that harmonised standards will build on it.
What is an AI system impact assessment?
An assessment, per AI system, of the potential consequences for individuals, groups and society of developing, providing or using it, covering fairness, safety, privacy, transparency, accountability and wider effects. It is distinct from the organisation-level AI risk assessment and is required by control A.5.
How long does certification take?
Comparable to ISO 27001: six to twelve months for an organisation with an existing ISMS and a clear AI inventory, longer where AI use is undocumented. The audit itself is days; the work is in the inventory, assessments and life-cycle records.
Related frameworks
Reviewed November 2026. This guide is general information about the published framework and is not legal advice; always check the current official text from the publisher before relying on it.