ISO 42001: the AI management system standard, explained

ISO/IEC 42001:2023 is the international standard for an artificial intelligence management system, and the first certifiable standard of its kind. Published in December 2023, it does for AI what ISO/IEC 27001 does for information security. It defines a management system, a structured set of policies, processes, roles, and controls, that an organization implements and that an accredited body can audit and certify.¹ The standard exists because AI raises governance questions general management standards do not answer: how an organization decides where to use AI, how it assesses the impact of those systems on people, and how it keeps that impact under control across a model's life.

For a regulated company in finance, healthcare, critical infrastructure, or SaaS, ISO 42001 matters now for two reasons. The first is commercial. Customers, partners, and procurement teams have started asking how the AI in a product is governed, and a certificate from an accredited body is independent evidence rather than a marketing claim. The second is regulatory. The EU AI Act is binding law with obligations that take effect on a staged timeline through 2027, and an AI management system gives an organization the governance backbone those obligations assume. This article explains what ISO 42001 covers, how it relates to ISO 27001, its core requirements, how it supports EU AI Act readiness, who should consider it, how certification works, and how it fits with the NIST AI Risk Management Framework, drawing on our security program and risk service and AI security service.

What ISO 42001 is, and why it matters now

ISO 42001 specifies the requirements for establishing, implementing, maintaining, and continually improving an AI management system, usually abbreviated AIMS.¹ Like other ISO management system standards, it follows the plan-do-check-act cycle and the common high-level structure shared across ISO management standards. That shared structure is deliberate, because it lets an organization run AI governance inside the same management system it already uses for information security or quality, rather than as a separate silo.

The defining point is what the standard certifies. ISO 42001 is about how an organization manages AI, not about declaring any individual AI system safe. It establishes governance, accountability, and repeatable process. Whether a specific model resists prompt injection or leaks training data is a separate, technical question answered by testing, the kind of work covered in AI penetration testing and the OWASP LLM Top 10. The two connect, because a credible management system requires that AI systems actually be evaluated, and that is where security testing supplies the evidence.

It matters now because the timing is no longer theoretical. AI is moving into production across regulated sectors, the EU AI Act has set deadlines, and buyers are writing AI governance into vendor questionnaires. An organization that builds a management system early is ready when a customer asks for proof and when a regulator asks how its high-risk systems are controlled.

How ISO 42001 relates to ISO 27001

ISO 42001 and ISO 27001 are built on the same foundation. Both use the harmonized management system structure, both run on plan-do-check-act, and both require leadership commitment, risk assessment, an annex of controls with a statement of applicability, internal audit, and management review. If you have implemented ISO 27001, the shape of ISO 42001 will be immediately familiar, and most of the governance machinery, the audit programme, the management review cadence, the corrective action process, can be reused rather than rebuilt.

The difference is scope. ISO 27001 governs information security: confidentiality, integrity, and availability of information. ISO 42001 governs the responsible development and use of AI, which adds concerns information security alone does not cover, including the impact of AI on individuals and society, the AI lifecycle, and transparency toward affected people. The practical consequence is that the two integrate cleanly. Organizations that hold ISO 27001 typically extend their existing system to cover AI rather than stand up a parallel one. Our guides to the ISO 27001 certification process and the information security management system explain the base ISO 42001 builds on.

The core requirements of ISO 42001

ISO 42001 sets management system requirements in its main clauses and lists controls in its annex. The work falls into a few connected areas, each of which a certification auditor will expect to see operating, not just written down.

• An AI policy and clear leadership accountability, stating how the organization uses AI, who owns AI risk, and what objectives the management system pursues.
• AI risk assessment and AI system impact assessment, where the impact assessment specifically considers consequences for individuals and groups, not only risk to the organization.
• Controls drawn from the standard's Annex A, covering areas such as AI policies, internal organization and roles, resources and data for AI, the AI system lifecycle, and information for interested parties, recorded in a statement of applicability.
• Lifecycle governance across the whole AI system life, from objective-setting and data selection through development, verification, deployment, operation, monitoring, and retirement.
• Transparency and information for users and affected parties, so people know when AI is in use and understand its limitations and intended purpose.
• Continual improvement through internal audit, management review, and corrective action, exactly as in other ISO management systems.

The AI impact assessment is the requirement most teams underestimate. It is not the same as a risk assessment focused on the business. It asks what an AI system could do to the people it touches, including bias, unfair outcomes, and loss of autonomy, and it expects those findings to feed back into design and controls. Done properly, it overlaps usefully with the fundamental rights impact assessment the EU AI Act expects for certain high-risk deployments.

ISO 42001 does not certify that your AI is safe. It certifies that you run a system for managing AI risk, the way ISO 27001 certifies you run a system for managing information security.

How ISO 42001 helps demonstrate EU AI Act readiness

The EU AI Act, Regulation (EU) 2024/1689, is binding law that sets obligations by risk tier, with the heaviest duties falling on providers and deployers of high-risk AI systems.² Those duties include a risk management system, data governance, technical documentation, record-keeping, transparency, human oversight, and post-market monitoring. ISO 42001 does not replace the Act, and certification is not the same as legal conformity. What it does is give an organization the management system the Act assumes you already run, covered in detail on our EU AI Act compliance page and in our plain-English explainer.

The overlap is substantial. The Act's risk management requirement maps onto ISO 42001's risk and impact assessment. Its data governance and documentation duties map onto the standard's data and lifecycle controls. Its transparency and human oversight obligations map onto the standard's transparency clauses. An organization that has implemented ISO 42001 has built most of the operational scaffolding the Act requires, which turns AI Act compliance from a standing start into a gap-closing exercise. ISO 42001 is the recognized international baseline organizations are adopting while harmonized European standards are finalized, and tying it all together is the discipline of AI governance.

Who should consider ISO 42001

ISO 42001 is relevant to any organization that develops, provides, or uses AI and wants to demonstrate responsible governance. It is most valuable in a few specific situations, where the cost of building the management system is clearly outweighed by the commercial or regulatory return.

• AI vendors and SaaS providers, for whom certification is becoming a procurement requirement and a way to answer customer security questionnaires with independent proof.
• Organizations subject to the EU AI Act, especially providers and deployers of high-risk systems, which can use ISO 42001 as a structured route toward demonstrating governance.
• Enterprises deploying AI at scale across many teams, which need a systematic way to manage AI risk consistently rather than case by case.
• Organizations that already hold ISO 27001, for which adding an AI management system is an efficient extension of an existing certification rather than a new build.
• Public-sector and critical-infrastructure bodies, which face heightened scrutiny over how automated decisions affect citizens and services.

How ISO 42001 certification works

The path mirrors other ISO certifications such as ISO 27001 and SOC 2, scaled to your starting point. The certificate is issued by an accredited certification body after an audit, not by the consultant who prepares you, which is what makes it independent evidence rather than a self-assessment.

1. 01
   Run a gap assessment

   Compare your current state against the standard to establish what already exists and what is missing. Organizations that already hold ISO 27001 usually find much of the management system machinery is reusable, which shortens this stage considerably.
2. 02
   Define scope and build the management system

   Decide which AI systems and which parts of the organization the management system covers, then build or extend it: the AI policy, the risk and impact assessment process, roles and accountability, and the Annex A controls captured in a statement of applicability.
3. 03
   Implement and operate the controls

   Put the controls into practice rather than only documenting them. This is where AI systems are actually evaluated, where lifecycle and data governance are exercised, and where evidence accumulates. An auditor checks that controls run, not just that they are written.
4. 04
   Conduct internal audit and management review

   Audit the management system against the standard, correct what the audit finds, and hold a management review so leadership confirms the system is working and resourced. The standard requires both before certification.
5. 05
   Pass the certification audit

   An accredited certification body conducts a Stage 1 documentation review and a Stage 2 audit of the system in operation, then issues the certificate. Surveillance audits follow in later years, with recertification on a multi-year cycle, so the system has to keep operating.

In our experience a first-time AI management system takes several months of focused effort for an organization that already runs ISO 27001, and longer for one starting from no management system at all, though the real estimate depends on scope, the number of AI systems, and the maturity of existing governance. Our role is to prepare the organization, connect the system to real testing, and validate that the controls hold before the certification body arrives.

The relationship to the NIST AI RMF

The NIST AI Risk Management Framework is a voluntary framework, not a certifiable standard.³ It organizes AI risk work into four functions, govern, map, measure, and manage, and it is widely used to structure the underlying analysis of how an AI system behaves and where it can go wrong. ISO 42001 and the NIST AI RMF are complementary rather than competing, and many organizations use them together. We cover the framework in detail in our NIST AI RMF guide.

A common and efficient pattern is to use the NIST AI RMF to structure the risk work, ISO 42001 to certify the management system that surrounds it, and the EU AI Act as the binding legal obligation the whole effort is aimed at. The framework gives you the method, the standard gives you the certifiable system and external proof, and the Act gives you the deadline. None of the three duplicates the others, so aligning them avoids doing the same risk analysis three times under three labels.

How Raptoric helps

ISO 42001 is most valuable when the management system rests on real security rather than paperwork, and that is the part organizations most often get wrong. We help regulated companies run the gap assessment, build or extend the AI management system on an existing ISO 27001 base, perform AI risk and impact assessments, and connect the controls to genuine evaluation of the AI systems in scope, so the certificate reflects how the AI behaves. If you want an AI management system that proves you govern AI responsibly and gives you a head start on EU AI Act readiness, see our security program and risk service and AI security service, then book a scoping call.