← Back to blog

ISO 42001: What the AI Management Standard Requires

November 10, 2026
ISO 42001: What the AI Management Standard Requires

ISO/IEC 42001 is the first international standard for an AI management system (AIMS). It sets out how an organization establishes, implements, maintains and continually improves the way it governs AI. It applies to companies that build AI and to companies that only use it. Certification is voluntary, and an independent auditor does it, not ISO. Below we cover what the standard contains, how it relates to ISO 27001, how to implement it, and what the audit looks like.

Skypher
Keep AI governance evidence ready
Skypher helps teams automate security questionnaire responses, using AI-driven tools to manage information and collaborate across teams.
Explore Skypher

TL;DR

  • ISO/IEC 42001:2023 is a certifiable management system standard for AI. It covers policy, risk, impact assessment, data, life-cycle controls and continual improvement.
  • It applies to any organization that provides or uses AI-based products or services, whatever its size or sector.
  • It is a management framework, not a law. It helps you show responsible AI governance, but it does not replace regulation.
  • Certification runs on a three-year cycle with annual surveillance audits. You can implement the framework without certifying, but the work is the same.
  • If you already run ISO 27001, you can build on the same management system structure and evidence habits.

Table of Contents

What ISO/IEC 42001 is and what it covers

Published in 2023, ISO/IEC 42001 is described by ISO as the world's first AI management system standard. It addresses challenges specific to AI, such as ethical considerations, transparency and continuous learning, and it gives organizations a structured way to manage AI risks and opportunities while balancing innovation with governance.

According to ISO's own explainer, the requirements span leadership and organizational context, AI policy and objectives, risk management, data governance and life-cycle controls, transparency, performance evaluation, monitoring and continual improvement. In plain terms, the standard asks you to decide who is accountable for AI, how you find and treat AI risks, and how you prove the system keeps working as your models and data change.

The standard follows the Plan-Do-Check-Act cycle. ISO notes that this encourages regular AI risk assessments and risk treatment activities, so oversight keeps pace with a technology that moves quickly. Think of it as a loop: you plan your policy and risk approach, operate the controls, check them through monitoring and internal audit, then act on what you find.

One point worth stating early: ISO/IEC 42001 does not replace laws or regulations. Per ISO, it provides a management framework that helps you meet compliance obligations more effectively. It does so by demonstrating responsible governance, managing risks such as bias, safety, security and misuse, and building trust with customers, partners and regulators.

What ISO/IEC 42001 is and what it covers

Who ISO 42001 applies to

The IEC describes the standard as applying to any organization, regardless of size, type and nature, that provides or uses products or services that utilize AI systems. ISO adds that it is relevant across all industries, for public sector agencies as well as companies and non-profits.

In practice, three groups tend to look at it first:

  1. AI developers and providers, such as SaaS companies that ship machine learning features to customers.
  2. Organizations that deploy third-party AI, such as banks, insurers and healthcare groups using models inside their own processes.
  3. Vendors answering customer due diligence, who are increasingly asked how they govern the AI inside their product.

The third group is where security and sales teams feel the standard first. Questions about AI governance now appear in security questionnaires next to the familiar ones on access control and encryption, and a recognized framework gives you a consistent, defensible way to answer.

Defining your role and scope

Clause 4.1 asks you to determine your role for each AI system. Modulos documentation lists examples including AI provider, AI producer, AI customer, AI partner, AI subject and relevant authority. The same source notes that the standard does not split the AIMS into an organization level and a system level. Your scope can hold more than one AI system, and you determine the opportunities and uses for each system or group of systems.

A sensible first scope is narrow and real: the AI features you sell, or the AI tools your teams already rely on. You can widen it at later audits.

How the standard is structured

The structure will feel familiar if you have worked with other management system standards. Standards Australia summarizes the requirements as policy, leadership, planning, support, operation, performance evaluation and improvement, plus normative AI-specific controls and guidance.

PartWhat it does
Clauses 4 to 10The management system requirements: context, leadership, planning, support, operation, performance evaluation, improvement
Annex ANormative AI controls you select from and document
Annex BNormative implementation guidance for every Annex A control
Annex CInformative examples of AI objectives and risk sources
Annex DInformative guidance on integrating with other management system standards

According to the same Standards Australia page, Annex D covers integration with standards such as ISO/IEC 27001, ISO/IEC 27701 and ISO 9001.

Annex A controls

Annex A holds 38 controls in nine objectives, numbered A.2 to A.10, as one training provider lays out: AI policies, internal organization, resources for AI systems, assessing 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 deliberately high-level. Mindset Cyber explains that you select the controls that apply, record that choice in a Statement of Applicability, and implement them in proportion to the risks found in your AI system impact assessment. Nothing in Annex A tells you which model architecture to use. It tells you which questions your governance must be able to answer.

ISO 27001 vs ISO 42001

The two standards answer different questions. ISO 27001 governs an information security management system: how you protect the confidentiality, integrity and availability of information. ISO 42001 governs an AI management system: how you manage the risks and impacts of AI systems themselves, including impacts on people, such as bias, transparency and misuse.

They overlap in mechanics. Both use the common management system structure, so internal audit, management review, document control and corrective action work much the same way. BSI observes that because ISO/IEC 42001 aligns with the structure of other management system standards, you can build on your existing compliance structure. Standards Australia likewise points to integrating AI governance with information security, privacy and quality management frameworks.

If you already hold an ISO 27001 certificate, a good share of the groundwork exists: a risk method, an audit program, a document system and an internal review rhythm. What is new is the AI-specific layer: impact assessments, life-cycle controls, data quality for AI, and responsibilities across AI roles. For a refresher on the security side, see our guides to ISO 27001 controls and the 93 Annex A controls, and our overview of ISO frameworks and security automation.

ISO 27001 vs ISO 42001

How to implement ISO 42001

A workable path has six steps. None needs special tooling, though good document control makes every step easier.

  1. Inventory your AI systems. List what you build, buy and use, who owns each, and what data feeds it. Include the AI features inside tools teams adopted on their own.
  2. Set scope and roles. Decide which systems the AIMS covers and what role you play for each, as Clause 4.1 requires.
  3. Write the AI policy and objectives. Get leadership sign-off, since the standard puts leadership accountability at the center.
  4. Assess risk and impact. Run an AI risk assessment, then an impact assessment per system, and decide how you will treat each finding.
  5. Select controls and record them. Work through Annex A, using Annex B for guidance, and document your choices in the Statement of Applicability.
  6. Operate, monitor, audit, review. Collect logs and monitoring results, run an internal audit, and hold a management review before any external audit.

Evidence an auditor expects

Standards Australia lists the typical artifacts: an AI policy, the AIMS scope, an AI risk assessment and treatment plan with a Statement of Applicability, AI system impact assessments, competence records, operational monitoring and event logs, internal audit reports and management review outputs.

Keep that list on one page and assign an owner to each item. Most delays we see in compliance programs come from evidence nobody owns, not from missing controls. Versioning matters too: a published, dated policy with a change history is much easier to defend than a document that quietly changed.

How ISO 42001 certification works

Certification is voluntary, and ISO does not certify organizations. Independent certification bodies do, and they may be accredited by national accreditation bodies. The certification body standard is ISO/IEC 42006. ASCB explains that it sits on top of ISO/IEC 17021-1 and adds AIMS-specific requirements, including auditor competence, scope, audit time, remote auditing, surveillance and recertification.

In the United States, ANAB launched its ISO/IEC 42001 accreditation program in January 2024 and set a transition period for accredited certification bodies that ended on 30 September 2025. When you choose an auditor, confirm that the body holds accreditation for ISO/IEC 42001 specifically. We cover that check in our six-step guide to getting ISO certified.

Stages and timeline

Per Elevate Consult, the process follows the ISO 17021 method:

  • Stage 1 reviews readiness and the design of your AIMS, typically one to two days.
  • Stage 2 collects full evidence that the system operates effectively, typically one to three weeks depending on scope.
  • Validity is three years, with surveillance audits at 12-month intervals in years two and three.

Glocert adds that surveillance audits typically run 30 to 50 percent of the initial audit duration and must cover internal audits, management review, actions on earlier nonconformities, complaints handling, AIMS effectiveness and selected operational controls. Our article on mastering ISO audits covers audit preparation in more depth.

Certification versus compliance

The difference is validation. Certification gives independent confirmation that your AIMS meets the standard. Implementing the framework without a certificate means no third-party validation, and both paths need the same work. If customers or regulators in your market ask for a certificate, certify. If you are earlier in your program, implementing first and certifying when you are ready is a reasonable sequence.

Benefits of adopting ISO 42001

BSI states that the standard helps organizations build trust in AI systems, improve quality, security, traceability, transparency and reliability, maintain compliance, promote accountability and meet stakeholder expectations. It also describes the standard as helping prevent a fragmented landscape of isolated guidelines.

For B2B teams, the practical benefit is commercial. A recognized framework turns a vague answer about responsible AI into a specific one that a buyer's risk team can evaluate. It also gives your own teams a common vocabulary: one policy, one impact assessment format, one place to look when a question comes in.

Our perspective

We see ISO 42001 as a questionnaire problem as much as a governance problem. Once your AI program is real, every enterprise buyer will ask about it, in security reviews, due diligence questionnaires and procurement portals, and the answers must match what your policies and evidence actually say.

That is why we think the strongest AI governance programs treat their documents as a living knowledge base: a versioned AI policy, current impact assessments, a Statement of Applicability, and clear owners. When that content is organized, it can feed every response consistently, instead of being rewritten by whoever is available that week.

Our platform is designed around this idea. Our proprietary retrieval AI draws answers from your own approved content, attaches a confidence score so reviewers know where to focus, and works with the formats and portals your customers already use. We built it so security, GRC and sales teams can collaborate in real time on one source of truth. Please feel free to reach out if you would like to see how that fits your ISO 42001 program.

Gaspard

Put your AI governance evidence to work

A finished AIMS only pays off commercially when buyers can see it. Our Questionnaire Automation Tool is designed to answer security and AI governance questionnaires from your approved knowledge base, with integrations for platforms such as OneTrust and ServiceNow and for the tools your documents already live in, including Confluence, Notion, Google Drive and SharePoint.

Skypher

If you want to publish your compliance posture proactively, our Trust Center gives customers one place to find it. You can also see how our smart security knowledge base keeps policies and answers versioned, or talk to us about pricing.

For related reading, see how AI is changing risk management for security questionnaires and our practical guide to the NIST AI Risk Management Framework, a natural companion to ISO 42001.

FAQ

What is ISO 42001?

ISO/IEC 42001:2023 is the first international standard for AI management systems. It gives requirements and guidance for establishing, implementing, maintaining and continually improving how an organization governs the AI it develops, provides or uses. Organizations can have their system independently certified.

Is ISO 42001 mandatory?

No. Certification is voluntary, and the standard does not replace laws or regulations. Many organizations adopt it because customers, partners or regulators expect a recognized way to show responsible AI governance.

What is the difference between ISO 27001 and ISO 42001?

ISO 27001 governs information security management, while ISO 42001 governs the management of AI systems and their risks and impacts. Both use a common management system structure, so existing processes such as internal audit and management review carry over. Many organizations run the two together.

Is ISO 42001 worth it?

It is worth it when AI is part of your product or operations and buyers ask how you govern it. The framework gives you a structured program and, with certification, independent proof. Because the work is the same with or without a certificate, your effort supports both customer trust and internal control.

How long does an ISO 42001 certificate last?

Three years, with surveillance audits in years two and three, according to Elevate Consult. Recertification follows at the end of the cycle.

Sources