ISO/IEC 42001 · AI management system

ISO 42001 certification: governing the AI you build, buy and ship

What ISO 42001 certification involves, how it relates to the EU AI Act and NIST AI RMF, and how to build an AI management system on top of the security work you already have.
Book a readiness call
Start the free self-assessment

In short

ISO/IEC 42001 is the first certifiable international standard for an AI management system (AIMS). It sets requirements for how an organisation governs the AI it develops, provides or uses: an inventory of AI systems, risk and impact assessments, lifecycle controls and clear accountability. ISO 42001 certification is a two-stage audit by an accredited certification body, and the standard shares its structure with ISO 27001.

2023

year the standard was published

4–6 wks

to audit-ready from an ISO 27001 base

2 stages

certification audit, like ISO 27001

1 register

for AI and security risk together

What ISO 42001 covers that ISO 27001 does not

ISO 27001 protects information. ISO 42001 governs how AI systems are chosen, built, deployed, monitored and retired, which brings in risks a security standard never had to consider. It certifies the management system around your AI, not a model, and it applies whether you train models yourself or build features on someone else’s.

Bias and fairness

Whether outcomes differ unfairly across groups, in training data and in production.

Transparency and explainability

Whether you can tell people when AI is involved and justify a decision it contributed to.

Data provenance

Where training and fine-tuning data came from, and on what basis you use it.

Drift and lifecycle

How performance is monitored after release, and how systems are retrained or retired.

Human oversight

Where a person can intervene, and who is accountable when they do.

Third-party AI

Models and AI features you buy or embed, governed as carefully as the ones you build.

Who is being asked for it

Demand is concentrated but growing quickly. Enterprise security questionnaires now include AI sections on training data, bias testing and human oversight that a SOC 2 report alone cannot answer, and procurement teams increasingly point to ISO 42001 as the evidence they would accept.

  • Vendors selling AI features into regulated industries such as finance, health and HR
  • Companies in scope of the EU AI Act that want a demonstrable governance baseline
  • Suppliers to enterprises with published AI procurement standards
  • Teams embedding third-party foundation models in customer-facing products

What an AI management system contains

ISO 42001 follows the same clause structure as ISO 27001: context, leadership, planning, support, operation, performance evaluation and improvement. What changes is the subject. Instead of information assets you inventory AI systems; instead of security risks you assess AI-specific ones; instead of the security Annex A you apply Annex A controls for AI covering data, models, lifecycle and accountability.

AI system inventory

Every model, agent and third-party AI service you build or use, with its purpose, data, owner and risk tier. Everything else depends on it.

AI impact assessment

For each system: who is affected, what could go wrong, how likely and how severe, and what you do about it.

Lifecycle controls

Data provenance, model documentation, testing before release, monitoring in production, human oversight and handling AI incidents.

Accountability

Named owners, an AI policy leadership signs, training for the people who build and use AI, and records that decisions were made and reviewed.

Start with the inventory

The hardest first step is knowing what AI you actually run. Most organisations underestimate it by a wide margin, because AI arrives through product features, vendor updates and experiments as often as through planned projects.

Commonly missed

  • Third-party models called through an API inside product features
  • AI embedded in SaaS tools your teams already use
  • Internal experiments that quietly reached production
  • Vendor features switched on by default after an update

For each system, record the owner, the purpose, the data it uses, who is affected by its output and a risk tier. Our guide on how to build an AI system inventory walks through it field by field.

AI system inventory with risk tiers and impact assessmentsIllustrative AI system inventory: four systems, whether each is built, bought or embedded, its risk tier and the status of its impact assessment.AI system inventory4 systems · 1 high-riskAI SYSTEMRISK TIERIMPACT ASSESSMENTSupport reply draftingThird-party model via APILimitedAssessedCredit-risk scoringBuilt in-houseHighIn reviewMeeting transcriptionEmbedded in a SaaS toolMinimalAssessedCode assistantVendor tool, internal useMinimalDueAI system inventory with risk tiers and impact assessmentsIllustrative AI system inventory: four systems, whether each is built, bought or embedded, its risk tier and the status of its impact assessment.AI system inventory4 systemsAI SYSTEMTIERIMPACTSupport reply draftingThird-party model via APILimitedAssessedCredit-risk scoringBuilt in-houseHighIn reviewMeeting transcriptionEmbedded in a SaaS toolMinimalAssessedCode assistantVendor tool, internal useMinimalDue

The inventory is the first document an auditor, regulator or enterprise buyer will ask to see. Illustrative data.

AI risk and impact assessments

ISO 42001 asks for two related assessments. The AI risk assessment looks at risks to your organisation and its objectives. The AI system impact assessment looks outwards, at the effects a system could have on individuals, groups and society. Systems with a high risk tier need the most depth.

In Certifyi AI risks sit in the same register as security risks, with the same scoring, owners, treatments and review dates, so the board sees one picture. Acceptance above your risk appetite needs a second approver who is never the person who recorded the risk, and acceptances expire so old decisions get looked at again.

AI impact assessment for one systemIllustrative AI impact assessment for a credit-risk scoring system: fairness, transparency, human oversight and data provenance, each with its control and residual rating.AI impact assessmentCredit-risk scoringDIMENSION AND CONTROLRESIDUALFairness across groupsQuarterly outcome testing by segmentMediumTransparency to applicantsPlain-language reasons on every declineLowHuman oversightAn analyst reviews every declineLowTraining data provenanceSources and lawful basis recordedLowResidual risk above appetite needs a second approver, never the person who recorded itAI impact assessment for one systemIllustrative AI impact assessment for a credit-risk scoring system: fairness, transparency, human oversight and data provenance, each with its control and residual rating.AI impact assessmentDIMENSION AND CONTROLRESIDUALFairness across groupsQuarterly outcome testing by segm…MediumTransparency to applicantsPlain-language reasons on every d…LowHuman oversightAn analyst reviews every declineLowTraining data provenanceSources and lawful basis recordedLowAbove appetite: second approver

One assessment per system, each dimension tied to the control that treats it. Illustrative data.

ISO 42001, the EU AI Act and NIST AI RMF

What it isApplies whenOutput
ISO/IEC 42001Certifiable AI management system standard (2023)You develop, provide or use AI and want independent proof of governanceCertificate from an accredited body
EU AI ActRegulation with obligations by risk tier: prohibited, high-risk, limited and minimalYour AI is placed on the EU market or its output is used in the EULegal compliance; conformity assessment for high-risk systems
NIST AI RMFVoluntary US framework: Govern, Map, Measure, ManageUS customers or public-sector buyers ask for itDocumented practices, no certificate

They are not the same thing, and ISO 42001 certification does not by itself prove EU AI Act compliance. But an ISO 42001 management system produces much of the governance evidence the Act expects, such as risk management, documentation, human oversight and record-keeping, which is why teams facing the Act often use the standard as their route to it.

On timing: the Act’s prohibited practices applied from February 2025 and general-purpose AI obligations from August 2025. Most high-risk obligations were scheduled for August 2026, with some product-safety categories in 2027. The European Commission has proposed changes to that timetable, so check the current dates before planning around them.

In Certifyi one control library is mapped to all three, so a control is evidenced once and answers wherever it applies.

One control library mapped to ISO 42001, the EU AI Act and NIST AI RMFIllustrative mapping: AI risk assessment, impact assessment, human oversight and logging controls each mapped to an ISO 42001 clause, an EU AI Act article and a NIST AI RMF function.Control mappingReferences abbreviatedISO 42001EU AI ActNIST AI RMFAI risk assessment6.1.2Art. 9MapAI impact assessment6.1.4Art. 27MapHuman oversightAnnex AArt. 14ManageLogging and recordsAnnex AArt. 12MeasureOne control, evidenced once, answers all three frameworksOne control library mapped to ISO 42001, the EU AI Act and NIST AI RMFIllustrative mapping: AI risk assessment, impact assessment, human oversight and logging controls each mapped to an ISO 42001 clause, an EU AI Act article and a NIST AI RMF function.Control mappingAI risk assessmentISO 6.1.2EU Art. 9NIST MapAI impact assessmentISO 6.1.4EU Art. 27NIST MapHuman oversightISO Annex AEU Art. 14NIST ManageLogging and recordsISO Annex AEU Art. 12NIST MeasureOne control, evidenced once

A single control mapped to an ISO 42001 clause, an EU AI Act article and a NIST AI RMF function. Illustrative mapping.

Building on ISO 27001

If you hold ISO 27001, a large part of an ISO 42001 management system already exists: the governance clauses, risk methodology, document control, internal audit and management review. What you add is the AI inventory, AI-specific risk and impact assessments, and the Annex A controls for AI.

That is why teams starting from ISO 27001 typically reach ISO 42001 audit-ready in four to six weeks. Many certification bodies run integrated audits, so a surveillance visit for ISO 27001 and Stage 1 for ISO 42001 can happen together, with shared controls examined once. Certifyi keeps both standards on one audit calendar.

You don’t need ISO 27001 first, but it helps considerably. Without it, allow for building the management system itself as well as the AI-specific parts.

Integrated audit calendar for ISO 27001 and ISO 42001Illustrative audit calendar: a joint management review and internal audit, an ISO 27001 surveillance visit combined with ISO 42001 Stage 1, then ISO 42001 Stage 2.Audit calendarOne calendar, both standardsEVENTWHENManagement reviewISO 27001 and ISO 42001 togetherJanInternal auditShared clauses audited onceFebISO 27001 surveillanceCertification body, year 1MarISO 42001 Stage 1Integrated with the surveillance visitMarISO 42001 Stage 2Implementation auditMayIntegrated audit calendar for ISO 27001 and ISO 42001Illustrative audit calendar: a joint management review and internal audit, an ISO 27001 surveillance visit combined with ISO 42001 Stage 1, then ISO 42001 Stage 2.Audit calendarEVENTWHENManagement reviewISO 27001 and ISO 42001 togetherJanInternal auditShared clauses audited onceFebISO 27001 surveillanceCertification body, year 1MarISO 42001 Stage 1Integrated with the surveillanc…MarISO 42001 Stage 2Implementation auditMay

Both standards on one calendar, with shared clauses audited once. Illustrative data.

The path to certification

1

Weeks 1–2

Inventory and tiering

Every AI system catalogued with its owner, purpose and data, and classified by risk tier. Scope agreed.

2

Weeks 3–8

Governance build

AI policy, risk and impact assessments, lifecycle controls and human oversight in place, with evidence collected on a schedule.

3

Weeks 9–12

Internal audit and review

Internal audit of the AIMS, management review and findings closed. From an ISO 27001 base this phase is much shorter.

4

Then

Certification audit

Stage 1 documentation review and Stage 2 implementation audit, often integrated with your ISO 27001 cycle.

With an existing ISO 27001 system, Certifyi gets you audit-ready in four to six weeks, and three to five months to the Stage 2 audit is realistic once the certification body’s schedule is included. Without one, allow six to nine months, most of it spent establishing the management system. The certification process guide covers each stage in detail.

How Certifyi helps

AI governance on the programme you already run

A named compliance lead

Weekly sessions with someone who has built management systems before, from the inventory to the certification audit.

One register for AI and security risk

The same scoring, owners, treatments and review cycles, so the board sees one picture rather than two.

One control library, three frameworks

ISO 42001, the EU AI Act and NIST AI RMF mapped together with ISO 27001 and SOC 2.

Integrated audits

ISO 27001 and ISO 42001 visits scheduled on one calendar, with the auditor working inside the platform.

Go deeper

AI governance guides from the compliance team

Understand

ISO 42001 vs the EU AI Act

Where they overlap, and where the Act asks for more.

Read →

Understand

The EU AI Act, explained

Risk tiers, obligations and timing.

Read →

Understand

NIST AI RMF, explained

Govern, Map, Measure and Manage in practice.

Read →

Plan

ISO 42001 certification process

Each stage, with realistic timings.

Read →

Plan

EU AI Act compliance

How Certifyi maps your systems to the Act.

Read →

Plan

SOC 2 for AI companies

Where AI features change a SOC 2 scope.

Read →

Do

How to build an AI system inventory

The first document an auditor asks for.

Read →

Do

ISO 42001 checklist

The AI management system, clause by clause.

Read →

Do

NIST AI RMF with Certifyi

Documented practices mapped to your controls.

Read →

Questions

ISO 42001, answered

ISO/IEC 42001 is the first certifiable international standard for an AI management system. It sets requirements for governing the AI an organisation develops, provides or uses, covering risk and impact assessment, data, lifecycle controls, transparency and human oversight.

Any organisation that develops, provides or uses AI systems and wants independent proof of governance. Enterprise buyers and the EU AI Act increasingly point to it.

Yes. If AI is part of your product or operations, the governance obligation is yours regardless of who trained the model. Your controls focus on selection, data handling, monitoring and human oversight rather than training.

No, but it helps considerably. The clause structure and much of the governance machinery are shared, so an ISO 27001 base shortens the work to four to six weeks to audit-ready.

From an ISO 27001 base, four to six weeks to audit-ready and three to five months to Stage 2 including the certification body’s schedule. Without one, allow six to nine months.

Not by itself. It demonstrates a conforming AI management system, which covers much of what the Act expects, but the Act’s specific obligations, such as conformity assessment for high-risk systems, still apply.

Yes. Many certification bodies run integrated audits, and Certifyi schedules both on one calendar so shared controls are examined once.

An AI system inventory, then an impact assessment for each system. Everything else in the standard depends on those two.

The risk assessment looks at risks to your organisation and its objectives. The impact assessment looks outwards, at the effects a system could have on individuals, groups and society.

Yes, if your AI system is placed on the EU market or its output is used in the EU, in the same way GDPR applies regardless of where you are based.

Shipping AI features?

Bring your product and your customer list. A compliance lead will tell you which framework your buyers will ask for first and what your inventory needs to contain.
Book a readiness call
Start the free self-assessment

Checklists

Compare

Company

Offices

Pulchowk, Lalitpur, Nepal
342 Waterloo Rd, Glenroy VIC 3046, Australia

Email

[email protected]

Copyright © 2026 CERTIFYI. All Rights Reserved by Certifyi AI

Scroll to Top