Zum Hauptinhalt springen
Standard · ISO

ISO/IEC 42001: AI Management System

Updated on 6 min Reviewed by: Cenedril Editorial
ISO 42001EU AI ActEN 18286

An insurer uses a model that pre-sorts claims. Six months later, internal audit asks: who decided to deploy the model? With which data was it tested? Who monitors whether it disadvantages certain customer groups? Without a management system for AI, the answers depend on individual people, and every staff change takes part of them away. ISO/IEC 42001 is the standard that turns these questions into fixed procedures.

ISO/IEC 42001 specifies the requirements for a management system for artificial intelligence (AI management system, AIMS). It describes how an organisation develops, provides or uses AI systems responsibly: with clear roles, an assessment of risks and impacts, defined controls and regular review. The current edition is ISO/IEC 42001:2023; the German version is published as DIN EN ISO/IEC 42001:2026-08.

What does the standard cover?

ISO 42001 follows the high-level structure of all ISO management system standards. If you know an ISMS under ISO 27001, you will find your way around clauses 4 to 10 immediately. New are the parts that relate to AI systems.

The clauses in the main body

  • Clause 4, Context of the organisation: internal and external issues, interested parties, scope of the AI management system. This is also where you determine your role in the AI landscape, for example as provider, producer or customer.
  • Clause 5, Leadership: top management commitment, AI policy, roles and responsibilities.
  • Clause 6, Planning: AI risk assessment (6.1.2), AI risk treatment with the Statement of Applicability (6.1.3), AI system impact assessment (6.1.4), AI objectives and planning of changes.
  • Clause 7, Support: resources, competence, awareness, communication, documented information.
  • Clause 8, Operation: risk assessment, risk treatment and impact assessment recur during operation, at planned intervals and when significant changes occur.
  • Clause 9, Performance evaluation: monitoring and measurement, internal audit, management review.
  • Clause 10, Improvement: continual improvement, nonconformity and corrective action.

Annex A: the 38 controls

Annex A is a reference catalogue. You decide which controls apply in the Statement of Applicability and justify every exclusion.

TopicContentControls
A.2 Policies related to AIAI policy, alignment with other policies, review of the AI policy3
A.3 Internal organisationAI roles and responsibilities, reporting of concerns2
A.4 Resources for AI systemsDocumentation of data, tooling, system and computing resources and human resources5
A.5 Assessing impacts of AI systemsProcess, documentation, impact on individuals and groups, societal impacts4
A.6 AI system life cycleObjectives and processes for responsible development, requirements, design, verification and validation, deployment, operation and monitoring, technical documentation, event logs9
A.7 Data for AI systemsData for development and enhancement, acquisition, data quality, data provenance, data preparation5
A.8 Information for interested partiesDocumentation for users, external reporting, communication of incidents, information for interested parties4
A.9 Use of AI systemsProcesses and objectives for responsible use, intended use3
A.10 Third-party and customer relationshipsAllocating responsibilities, suppliers, customers3

Annex B gives implementation guidance for each control. Annex C lists potential AI-related objectives (for example fairness, transparency, robustness) and risk sources, and Annex D describes the use of the management system across domains and sectors.

Certification

Certification works as for ISO 27001 in two stages: a review of the documentation (stage 1) and an on-site review of effectiveness (stage 2), followed by annual surveillance audits and recertification after three years. Since July 2025, ISO/IEC 42006 sets out the additional requirements that certification bodies for AI management systems must meet. When choosing a body, make sure it is accredited for ISO 42001.

Relationship to the EU AI Act and EN 18286

The EU AI Act is a regulation with obligations per risk class and role. ISO 42001 is a voluntary management system standard. The two interlock:

  • Common framework. Risk management, documentation, human oversight, post-market monitoring and incident handling can be run as procedures within the AI management system.
  • Quality management system for providers. Providers of high-risk AI systems need a quality management system under Art. 17 AI Act. For this, the Commission has published EN 18286 as a harmonised standard in the Official Journal (Implementing Decision (EU) 2026/2274). Applying it gives a presumption of conformity within the scope of its Annex ZA. ISO 42001 remains the management system around it.
  • No automatic conformity. An ISO 42001 certificate demonstrates the management system. You fulfil the AI Act obligations (classification, technical documentation, conformity assessment, registration) in addition.

Relationship to other standards

StandardRelationship to ISO 42001
ISO/IEC 27001Same high-level structure; the ISMS covers the information security of AI systems, both systems can be run in an integrated way
ISO/IEC 27701Privacy information management system; relevant as soon as AI systems process personal data
ISO/IEC 23894Guidance on risk management for AI, building on ISO 31000
ISO 31000General principles and process of risk management
ISO/IEC 42005Guidance on AI system impact assessment
ISO/IEC 42006Requirements for bodies certifying AI management systems
EN 18286Harmonised standard for the quality management system of providers under Art. 17 AI Act

Free templates

The AI Compliance Starter Kit contains the documents ISO/IEC 42001 requires as free templates: the AI policy (clause 5.2), the AI risk register and the AI risk treatment plan (6.1.2, 6.1.3), the AI Statement of Applicability with all 38 controls in Annex A, the fundamental rights impact assessment for A.5 and the AI communication plan (7.4). The overview shows for each control where the kit implements it.

What Cenedril covers

Cenedril’s AI assistant is built on ISO 42001. It holds the control catalogue with all 38 Annex A controls in ten groups, plus:

  • determining your role and classifying each AI system under the AI Act,
  • the AI register with purpose, data and suppliers per system,
  • the AI risk register with risk treatment plan and the Statement of Applicability with a justification per exclusion,
  • the AI system impact assessment and the fundamental rights impact assessment,
  • AI policy, internal audit and management review for the AI management system,
  • incidents, observations and corrective actions (CAPA) for AI systems.

For providers of high-risk AI, the quality management system under EN 18286 is added.

  • EU AI Act: the regulation with obligations per risk class and role.
  • EN 18286: quality management system for providers of high-risk AI.
  • ISO/IEC 27001: information security management system with the same high-level structure.
  • ISO 31000: principles of risk management.

Sources

Frequently asked questions

Do I need ISO 42001 if I have to comply with the EU AI Act?

The AI Act does not require ISO 42001 certification. The standard does give you the framework in which you manage AI risks, impact assessments, roles and evidence on an ongoing basis. Many AI Act obligations (risk management, documentation, post-market monitoring) can be run within it. For the quality management system of providers of high-risk AI, the harmonised standard EN 18286 applies; it builds on a management system such as ISO 42001.

Can I combine ISO 42001 with my ISO 27001 ISMS?

Yes. Both standards follow the same high-level structure for management systems (clauses 4 to 10). Context, leadership, document control, internal audit, management review and corrective action can be run jointly. On top come the AI-specific parts: AI system impact assessment, AI risk assessment and the Annex A controls.

Does ISO 42001 only apply to companies that develop AI?

No. The standard addresses every organisation that provides, develops or uses AI systems. In clause 4.1 you determine your role (for example provider, producer, customer or partner), and that role determines which Annex A controls apply.

No Cookies!

This wiki collects nothing, bakes nothing and leaves nothing behind. There's nothing to consent to. Privacy doesn't get better than this.