AI governanceGlobal

ISO 42001:2023

Build an AI management system with clear ownership

Connect AI policies, risk decisions and review evidence in a programme that can improve over time.

Full title
ISO/IEC 42001:2023 Information technology: Artificial intelligence: Management system
Issued by
ISO and IEC, through ISO/IEC JTC 1/SC 42, the subcommittee on artificial intelligence
Published
December 2023, first edition. No amendment or newer edition as of October 2026.
Applies to
Any organisation that provides, develops or uses AI-based products or services, of any size or sector
Nature
Voluntary international standard. Organisations can be certified by a certification body.
Structure
Management system requirements in Clauses 4 to 10, plus 38 reference controls under 9 objectives in Annex A, and Annexes B to D

Overview

What ISO 42001 is

ISO/IEC 42001 is the international standard for an AI management system, or AIMS. It sets out how an organisation governs the AI systems it develops, provides or uses: who is accountable, how AI risks and impacts are assessed, which controls are applied, and how the whole system is monitored and improved. It does not certify that a particular AI model is safe or lawful. It certifies that the organisation runs a management system for AI responsibly and consistently.

The standard follows the same harmonised structure as ISO/IEC 27001, so organisations that already run an information security management system will recognise Clauses 4 to 10. What is specific to AI is the planning: alongside AI risk assessment and AI risk treatment, the organisation must carry out AI system impact assessments, which look at the potential consequences of AI systems for individuals, groups of individuals and societies.

Annex A lists 38 reference controls under nine control objectives, from AI policy and internal organisation to the AI system life cycle, data, information for interested parties, use of AI systems and third-party relationships. As with ISO 27001, the organisation compares its risk treatment with Annex A and records its decisions in a Statement of Applicability. Annex B gives implementation guidance for each control, Annex C lists potential AI-related objectives and risk sources, and Annex D discusses use of the AIMS across domains and sectors.

Related standards support the programme. ISO/IEC 42005:2025 gives guidance on AI system impact assessment, and ISO/IEC 42006:2025 sets requirements for the bodies that audit and certify an AIMS. Certification to ISO 42001 does not by itself establish compliance with laws such as the EU AI Act, which need their own assessment.

Who it applies to

ISO 42001 is relevant to any organisation that provides, develops or uses AI systems, including:

  • Software and SaaS companies building AI features into their products
  • Developers and providers of AI models and platforms
  • Banks, NBFCs, insurers and fintechs using AI in credit, fraud, customer service or operations
  • Healthcare, education and public sector bodies deploying AI that affects individuals
  • Organisations buying AI services that want governance over how those services are used
  • MSSPs, vCISOs and consultancies extending their governance services to AI

The organisation decides the scope of the AIMS and its role in relation to each AI system, for example as a provider, producer, customer or user.

Structure

How the standard is organised

Clauses 0 to 3 introduce the standard. Clauses 4 to 10 contain the requirements. Annex A lists the reference controls, and Annexes B to D give guidance.

Clauses 0–3

Introduction, scope, references and terms

How the standard is used and what it covers. Many AI terms rely on ISO/IEC 22989. These clauses contain no auditable requirements.

Clause 4

Context of the organisation

4.1 Understanding the organisation and its context
Internal and external issues relevant to AI, including the intended purpose of the AI systems and the organisation’s roles in relation to them.
4.2 Needs and expectations of interested parties
Who has a stake in the AI systems and what they require.
4.3 Scope of the AI management system
The boundaries and applicability of the AIMS, documented.
4.4 AI management system
Establishing, implementing, maintaining and continually improving the AIMS.
Clause 5

Leadership

5.1 Leadership and commitment
Top management shows commitment, aligns the AIMS with strategy and provides resources.
5.2 AI policy
A documented AI policy that provides a framework for AI objectives, communicated and available as appropriate.
5.3 Roles, responsibilities and authorities
Responsibilities for the AIMS assigned and communicated.
Clause 6

Planning

6.1 Actions to address risks and opportunities
AI risk criteria, a defined AI risk assessment process, AI risk treatment with a Statement of Applicability compared against Annex A, and an AI system impact assessment process.
6.2 AI objectives and planning to achieve them
Measurable objectives, with plans for what will be done, by whom and by when.
6.3 Planning of changes
Changes to the AIMS carried out in a planned way.
Clause 7

Support

7.1 to 7.4 Resources, competence, awareness and communication
The people, skills, awareness and communication the AIMS needs, with evidence of competence.
7.5 Documented information
Creating, updating and controlling the documents and records the AIMS needs.
Clause 8

Operation

8.1 Operational planning and control
Planning and controlling the processes needed, including externally provided ones.
8.2 AI risk assessment
Performed at planned intervals or on significant change, with results retained.
8.3 AI risk treatment
The treatment plan implemented and its effectiveness verified.
8.4 AI system impact assessment
Impact assessments performed at planned intervals or on significant change, with results retained.
Clauses 9–10

Performance evaluation and improvement

9.1 Monitoring, measurement, analysis and evaluation
What is measured, how and when, and how results are evaluated.
9.2 Internal audit
Internal audits at planned intervals under an audit programme.
9.3 Management review
Top management reviews the AIMS at planned intervals and records decisions.
10.1 and 10.2 Improvement and corrective action
Continual improvement, and reacting to nonconformities with corrective action.
Annex A

38 reference controls under 9 objectives

The organisation compares its risk treatment with these controls and records inclusions and exclusions in the Statement of Applicability.

A.2 Policies related to AI (3)
An AI policy, its alignment with other organisational policies, and its review.
A.3 Internal organisation (2)
AI roles and responsibilities, and a process for reporting concerns about AI systems.
A.4 Resources for AI systems (5)
Documenting the data, tooling, system and computing, and human resources each AI system relies on.
A.5 Assessing impacts of AI systems (4)
An impact assessment process, its documentation, and assessing impacts on individuals, groups and society.
A.6 AI system life cycle (9)
Objectives and processes for responsible development, then requirements, design, verification and validation, deployment, operation and monitoring, technical documentation and event logs.
A.7 Data for AI systems (5)
Data for development and enhancement, acquisition, quality, provenance and preparation.
A.8 Information for interested parties (4)
Documentation and information for users, external reporting, communication of incidents, and information for interested parties.
A.9 Use of AI systems (3)
Processes and objectives for responsible use, and use in line with the intended use.
A.10 Third-party and customer relationships (3)
Allocating responsibilities, and managing suppliers and customers.
Annexes B–D

Guidance annexes

Annex B Implementation guidance for AI controls
Guidance on implementing each Annex A control. Organisations need not follow it exactly.
Annex C Potential AI-related organisational objectives and risk sources
Examples of objectives, such as fairness, transparency, accountability, privacy, robustness and safety, and of sources of AI risk.
Annex D Use of the AI management system across domains or sectors
How the AIMS can be applied in different sectors and integrated with other management system standards.

Implementation roadmap

How to implement it, step by step

A typical route to an operating AIMS and certification. Organisations with an ISO 27001 ISMS can reuse much of the management system.

  1. 1

    Inventory AI systems and decide your roles

    List the AI systems the organisation develops, provides or uses, including AI inside purchased services. For each, record its intended purpose and the organisation’s role.

    OutputsAI system inventoryRole determination
  2. 2

    Define the scope and the AI policy

    Decide which AI systems, units and processes the AIMS covers. Approve an AI policy that sets the framework for AI objectives and aligns with existing policies.

    OutputsAIMS scopeAI policy
  3. 3

    Assign roles and set objectives

    Allocate AI roles and responsibilities, set up a route for raising concerns, and agree measurable AI objectives.

    OutputsRoles and responsibilitiesAI objectives
  4. 4

    Assess AI risks and impacts

    Define AI risk criteria. Assess risks to the organisation and carry out AI system impact assessments of consequences for individuals, groups and society.

    OutputsAI risk assessmentAI system impact assessments
  5. 5

    Treat risks and write the Statement of Applicability

    Choose controls, compare them with the 38 Annex A controls, and record which apply, why, and the justification for any exclusion.

    OutputsRisk treatment planStatement of Applicability
  6. 6

    Embed controls in the AI life cycle

    Build the controls into design, data preparation, verification and validation, deployment, monitoring and supplier management, so that each stage produces records.

    OutputsLife cycle proceduresTechnical documentationEvent logs
  7. 7

    Monitor, audit and review

    Measure AI performance against objectives, run an internal audit across all clauses, and hold a management review.

    OutputsInternal audit reportManagement review minutes
  8. 8

    Certify and keep it running

    Engage a certification body for a Stage 1 and Stage 2 audit. Keep the AIMS operating through surveillance audits and recertification.

    OutputsCertificateSurveillance plan

Audit and evidence guide

What auditors look for, and the evidence to keep

Certification is carried out by a certification body. ISO/IEC 42006:2025 sets additional requirements for those bodies, including the AI competence of audit teams.

The certification cycle

  1. Step 1

    Stage 1 audit

    A review of the AIMS design: scope, AI policy, risk and impact assessment methods, Statement of Applicability, internal audit and management review.

  2. Step 2

    Stage 2 audit

    An assessment of whether the AIMS is implemented and effective, through interviews, observation and sampling of records across the AI life cycle.

  3. Step 3

    Nonconformities and certification

    Nonconformities are corrected or covered by accepted corrective action plans before the certificate is issued.

  4. Step 4

    Surveillance and recertification

    Surveillance audits check that the AIMS keeps operating, and a recertification audit takes place before the certificate expires.

What reviewers look for

  • An AI system inventory that matches what the organisation actually runs and uses
  • The organisation’s role for each AI system, and how the scope was decided
  • AI risk criteria and assessments that follow the documented method
  • AI system impact assessments for individuals, groups and society, and how their results were used
  • A Statement of Applicability consistent with the risk treatment plan
  • Data provenance, quality and preparation records for AI systems
  • Verification and validation before deployment, and monitoring in operation
  • Information given to users and interested parties, and how incidents are communicated
  • Responsibilities agreed with suppliers and customers of AI systems
  • Internal audit and management review covering the whole AIMS

Key timelines

Certificate cycle, with surveillance audits between3 years
AI risk assessments (8.2)Planned intervals, on change
AI system impact assessments (8.4)Planned intervals, on change
Internal audits (9.2)At planned intervals
Management reviews (9.3)At planned intervals
AI policy review (A.2.4)At planned intervals

Evidence to keep, by area

Examples to guide your evidence library. Your auditor or regulator decides what is sufficient.

Governance

  • AIMS scope
  • AI policy
  • AI roles and responsibilities
  • Concern reporting process

Risk and impact

  • AI risk criteria
  • AI risk assessments
  • AI system impact assessments
  • Statement of Applicability

AI life cycle

  • Requirements and design records
  • Verification and validation results
  • Deployment approvals
  • Monitoring and event logs

Data

  • Data provenance records
  • Data quality checks
  • Data preparation records

Interested parties

  • User documentation
  • External reports
  • Incident communications

Suppliers and customers

  • Responsibility allocations
  • Supplier assessments
  • Customer agreements

Evaluation

  • Monitoring results
  • Internal audit reports
  • Management review minutes
  • Corrective actions

Areas that often need attention

  • AI used inside purchased tools left out of the inventory
  • Impact assessments treated as a one-off form rather than repeated on change
  • No link between impact assessment results and design or deployment decisions
  • Data provenance unknown for training or fine-tuning data
  • Models deployed without recorded verification and validation
  • Statement of Applicability copied from a template, not from the risk treatment
  • Supplier responsibilities for AI components not agreed in writing

How Netra helps

Where each part of the work lives in Netra

Netra loads ISO 42001 as requirements, tasks and policies. Where the management system clauses match ISO 27001, the same work counts towards both.

Clause and control requirements owned and tracked

Task Engine

Requirements arrive as tasks with owners and due dates, and recurring tasks keep assessments and reviews current.

Status for each clause and Annex A control

Compliance Centre

Status is derived from completed tasks and evidence, never typed in.

AI policy and related policies

Policy Hub

A policy library generated at onboarding, with a score that moves as the work behind each policy is completed.

AI risk assessment and treatment

Risk Manager

A risk register with inherent and residual heatmaps, with risks reduced as the linked tasks are completed.

AI systems and the components they rely on

Asset Management

Assets linked to tasks, so teams know which AI system each piece of work covers.

Suppliers of AI components and services

Third-Party Risk

Find vendors from what Accounts and Purchase pay, assess them through a vendor portal and record each decision with a rating.

Internal and certification audits

Audit Centre

Run audits end to end with findings and corrective actions, using evidence held in a tamper-evident vault.

Management review and board reporting

Dashboard & Reports

A live Posture Score and a one-click board report for the framework.

Netra organises the work and evidence of an AIMS. It does not assess or certify AI systems, does not by itself establish conformity with ISO/IEC 42001 or compliance with any AI law, and is not a certification body.

Common questions

More questions about Netra

Related frameworks

View all frameworks
AI governance

NIST AI RMF 1.0

Turn responsible AI goals into accountable work

AI governance

EU AI ACT 2024

Keep AI governance decisions supported by evidence

Security

ISO 27001:2022 (Requirements)

Build an ISMS your team can run every day

See ISO 42001:2023 in Netra

See how Netra connects your framework requirements with tasks, evidence and audit follow-up.