AI Compliance Vendors
Voluntary standardVoluntary standardAI management system standard

ISO/IEC 42001: the complete guide

ISO/IEC 42001 is the international standard for governing AI across an organization. It defines an AI management system (AIMS) that an independent auditor can certify, whether you build AI or only use it [2][3]. This guide covers what it requires, lists all 38 Annex A controls, explains certification, and sets out what it does and does not do under the EU AI Act.

Reviewed 15 Oct 2026 · next review by 13 Jan 2027 · 28 sources · report an error

What ISO 42001 is

ISO/IEC 42001:2023 was published on 18 December 2023 by ISO and the IEC, through their joint committee on artificial intelligence (ISO/IEC JTC 1/SC 42) [1]. It sets out requirements for establishing, implementing, maintaining and continually improving an AI management system: the policies, roles, processes and controls an organization uses to develop, provide or use AI responsibly [3].

It is a management-system standard, in the same family as ISO/IEC 27001 for information security. A certificate says how an organization governs AI; it does not certify any particular AI product [8]. Because it uses ISO’s harmonized structure, with the same clause numbers and core text as other management-system standards, it can be run alongside ISO 27001 or ISO 9001 as one integrated system [3].

Certification is voluntary and is carried out by independent certification bodies; ISO itself does not certify anyone, and the standard does not replace any law [2].

Who it applies to

The standard is written for any organization that provides or uses products or services that rely on AI systems, of any size, type or sector [3][2]. A company that only deploys third-party AI can be certified: controls that don’t apply to it, such as some development controls, are left out of its Statement of Applicability with a written justification [22].

Clause 4.1 asks each organization to decide its role in relation to its AI systems, using the roles defined in ISO/IEC 22989: for example AI provider, AI producer, AI customer or AI partner [3][12]. The role shapes which controls matter most.

Organizations that have announced certification, from our verified list of ISO 42001 certified companies, include:

  • AI platforms and model developers, such as AWS, Microsoft, Google Cloud, Anthropic and IBM (for its Granite models).
  • Software companies with AI features, such as Salesforce, Workday, ServiceNow, Zendesk and Datadog.
  • Professional-services firms, such as KPMG, PwC Taiwan and EY Greece, and compliance platforms such as Vanta and Drata.

No authority publishes a count of certificates. One independent tracker of public announcements counted 220 certified organizations in 29 countries as of 11 October 2026; it does not check accreditation registers [21].

What the standard requires: clauses 4 to 10

The requirements sit in clauses 4 to 10, which follow the same pattern as every modern ISO management-system standard [3]. The table paraphrases each clause; the standard’s own wording is copyrighted and has to be bought from ISO or a national standards body.

ISO/IEC 42001 clauses 4–10
ClauseWhat it covers
4 Context of the organizationInternal and external issues, interested parties, the scope of the AIMS, and your role(s) in relation to AI systems [3]
5 LeadershipTop-management commitment, an AI policy (5.2), and roles, responsibilities and authorities (5.3) [3]
6 PlanningAI risk assessment (6.1.2), AI risk treatment and the Statement of Applicability (6.1.3), AI system impact assessment (6.1.4), AI objectives and planning of changes [3][7]
7 SupportResources, competence, awareness, communication and documented information [3]
8 OperationOperational planning and control, and carrying out the risk assessment, risk treatment and impact assessment at planned intervals or when things change significantly (8.2–8.4) [3][7]
9 Performance evaluationMonitoring and measurement, internal audit and management review [3]
10 ImprovementContinual improvement, nonconformities and corrective action [3]

Two requirements are specific to AI. The AI risk assessment looks at risks to meeting your AI objectives. The AI system impact assessment (6.1.4) looks outward: the potential consequences of an AI system for individuals, groups of individuals and societies. Its results are documented and feed into the risk assessment [7]. ISO/IEC 42005:2025 gives detailed guidance on running impact assessments [10]; our AI impact assessment template is a practical starting point.

Annex A: all 38 controls

Annex A lists 38 reference controls in 9 areas, A.2 to A.10. Annex B gives implementation guidance for each control under matching numbers (B.2.2 for A.2.2, and so on). Both annexes are normative. Annex C lists potential AI-related objectives and risk sources, and Annex D covers using an AI management system across domains and sectors; both are informative [3][4].

ISO/IEC 42001 Annex A control areas and controls
AreaControlsCount
A.2 Policies related to AIA.2.2 AI policy · A.2.3 Alignment with other organizational policies · A.2.4 Review of the AI policy3
A.3 Internal organizationA.3.2 AI roles and responsibilities · A.3.3 Reporting of concerns2
A.4 Resources for AI systemsA.4.2 Resource documentation · A.4.3 Data resources · A.4.4 Tooling resources · A.4.5 System and computing resources · A.4.6 Human resources5
A.5 Assessing impacts of AI systemsA.5.2 AI system impact assessment process · A.5.3 Documentation of AI system impact assessments · A.5.4 Assessing AI system impact on individuals or groups of individuals · A.5.5 Assessing societal impacts of AI systems4
A.6 AI system life cycleA.6.1.2 Objectives for responsible development of AI system · A.6.1.3 Processes for responsible design and development of AI systems · A.6.2.2 AI system requirements and specification · A.6.2.3 Documentation of AI system design and development · A.6.2.4 AI system verification and validation · A.6.2.5 AI system deployment · A.6.2.6 AI system operation and monitoring · A.6.2.7 AI system technical documentation · A.6.2.8 AI system recording of event logs9
A.7 Data for AI systemsA.7.2 Data for development and enhancement of AI system · A.7.3 Acquisition of data · A.7.4 Quality of data for AI systems · A.7.5 Data provenance · A.7.6 Data preparation5
A.8 Information for interested partiesA.8.2 System documentation and information for users · A.8.3 External reporting · A.8.4 Communication of incidents · A.8.5 Information for interested parties4
A.9 Use of AI systemsA.9.2 Processes for responsible use of AI systems · A.9.3 Objectives for responsible use of AI system · A.9.4 Intended use of the AI system3
A.10 Third-party and customer relationshipsA.10.2 Allocating responsibilities · A.10.3 Suppliers · A.10.4 Customers3
Area titles as published by certification body LRQA [6]; control numbers derived from the Annex B numbering in the NIST-hosted crosswalk, which was prepared on the final draft (FDIS) text, so check the published standard for exact wording [4].

You don’t have to implement all 38. During risk treatment (6.1.3) you decide which controls you need, compare them with Annex A so nothing necessary is missed, and record the result in a Statement of Applicability that justifies each inclusion and exclusion [22].

ISO 42001 checklist: what auditors look for

A practical checklist of the evidence certification auditors typically examine, mapped to the clauses and controls above [3][7][18]:

  1. 01

    Scope and context (4.1–4.3)

    A written AIMS scope, the issues and interested parties you considered, and your role(s) for each AI system.

  2. 02

    AI policy (5.2, A.2)

    An approved AI policy, aligned with your other policies and reviewed on a schedule.

  3. 03

    Roles and reporting (5.3, A.3)

    Named owners for AI governance and a channel for staff to report concerns.

  4. 04

    AI system inventory and resources (A.4)

    Documentation of each AI system’s data, tooling, computing and people.

  5. 05

    Risk assessment and treatment (6.1.2–6.1.3, 8.2–8.3)

    A defined method and risk criteria, current results, a treatment plan and the Statement of Applicability.

  6. 06

    Impact assessments (6.1.4, 8.4, A.5)

    Documented assessments of effects on individuals, groups and society for each in-scope AI system.

  7. 07

    Life cycle, data and transparency controls (A.6–A.9)

    Evidence the controls you selected operate: requirements, validation, deployment approvals, monitoring, event logs, data provenance and quality, user information and incident communication.

  8. 08

    Suppliers and customers (A.10)

    Responsibilities allocated with AI suppliers and customers, and supplier checks.

  9. 09

    Internal audit and management review (9.2–9.3)

    At least one completed internal audit of the AIMS and a management review before the certification audit.

  10. 10

    Corrective action (10.2)

    A log of nonconformities found and how they were fixed.

How certification works

An independent certification body audits your AIMS in two stages. Stage 1 reviews the design and documentation; Stage 2 tests whether it operates as described, including clause 8 activities, monitoring, internal audit and management review. After a positive decision, the certificate runs for three years, with surveillance audits every calendar year and a full recertification audit before it expires [18][17].

Choose a certification body that is accredited for ISO/IEC 42001 by a national accreditation body. ANAB in the US opened its program in January 2024, and in September 2024 Schellman announced it was the first ANAB-accredited body [8][20]; UKAS granted its first accreditation, to BSI, in January 2026 [19]. ISO/IEC 42006:2025 sets auditor competence rules and how audit time is calculated for AI management systems, and accreditation bodies such as the Standards Council of Canada are moving certification bodies onto it [9][28].

ISO 42001 vs ISO 27001, NIST AI RMF and SOC 2

How ISO 42001 relates to the frameworks it is most often compared with
FrameworkWhat it isCertifiable?Relationship to ISO 42001
ISO/IEC 27001Information security management system standard [23]YesISO 42001 uses the harmonized structure shared by ISO management system standards, so one integrated management system can cover both; whether audits are combined depends on the certification body. Its Annex A covers information security, 42001’s covers AI governance, and neither certificate covers the other [3]
NIST AI RMF 1.0Voluntary US framework (Govern, Map, Measure, Manage), January 2023; NIST says it is being revised under the White House AI Action PlanNoNIST’s AI Resource Center hosts a crosswalk to 42001 provided by Microsoft and based on the final draft [16][5][4]
SOC 2AICPA attestation report on the Trust Services Criteria (security, availability, processing integrity, confidentiality, privacy)No (attestation report)Covers organization-wide controls against the Trust Services Criteria, not AI governance [24]

More detail: NIST AI RMF vs ISO 42001 and SOC 2 vs ISO 42001.

ISO 42001 and the EU AI Act

ISO 42001 is not a harmonised standard under the EU AI Act, so certification does not give a presumption of conformity. The European Commission says the standard’s goals and definitions are not aligned with the quality management system the AI Act requires, which is why it asked European standards bodies for a new one [13]. The European adoption, EN ISO/IEC 42001:2026, carries no annex linking it to the Act [15].

ISO 42001 is still useful groundwork. An analysis of the draft EN 18286 by the Cloud Security Alliance found that the draft maps its requirements to ISO 42001’s Annex A controls, so parts of an existing AIMS can be reused [25]. But the AI Act regulates AI systems, not organizations: providers of high-risk systems still need system-level technical documentation, testing and a conformity assessment, which a 42001 certificate does not replace [15]. After the Digital Omnibus, high-risk obligations apply from 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems [26]; see our AI compliance timeline.

Where to get the standard

ISO sells ISO/IEC 42001:2023 for CHF 225 (51 pages) [1]; national standards bodies sell their own adoptions, such as EN ISO/IEC 42001:2026 in Europe [15]. ISO’s official sample shows the foreword, contents and scope [3]. Free full-text “ISO 42001 PDF” copies circulating online are not licensed by ISO.

Software and help

Compliance platforms can hold your AI inventory, policies, risk and impact assessments and evidence in one place, and map them to Annex A. Compare them in our independent ranking of ISO 42001 software, or browse every vendor that supports the standard below.

Vendors that support ISO/IEC 42001

Sorted by coverage level. Full coverage shown first.

37 vendors

VendorHQFoundedSizePricingCoverageLast checked
LatticeFlow AIZurich, Switzerland202011-50No public pricing. Enterprise platform sold via direct sales. Contact sales for demo and pricing.FullSep 21, 2026
EnzaiBelfast, United Kingdom20211-10SaaS platform, enterprise subscription. No public pricing listed. Contact sales via enz.ai.FullSep 21, 2026
Fairly AIKitchener, Canada202011-50On-premises or private-cloud deployments; quote-based.FullSep 21, 2026
Collibra AI GovernanceNew York, United States20081000+Enterprise subscription; contact sales for custom quote based on users, assets, modules.FullSep 21, 2026
TrustibleArlington, United States202311-50Contact sales for enterprise pricing; no public plans listedFullSep 21, 2026
ModelOpChicago, United States201811-50No public pricing listed; contact sales for enterprise quotes.FullSep 21, 2026
Modulos AI GovernanceZurich, Switzerland201811-50Contact for pricingComprehensiveSep 21, 2026
Scrut AutomationPalo Alto, United States202151-200Contact for pricingComprehensiveSep 21, 2026
BABL AIIowa City, United States201811-50Contact for pricingComprehensiveSep 21, 2026
NaaiaLouveciennes, France202111-50No public pricing tiers; demo and quote requested via website.ComprehensiveSep 21, 2026
ScytaleIsrael202151-200No public prices. Platform packages (Build, Scale, Enterprise), startup bundles (Build Starter, Build DFY, Build Stronger) and consulting packages (LaunchReady, StayReady, ComplianceShield) are all quoted after a demo.ComprehensiveOct 15, 2026
SecureframeSan Francisco, United States2020201-500Three plans: Fundamentals (starting at $7,500/year, one compliance framework), Complete (quote only; adds advanced versions of third-party risk, risk management, user access reviews, Trust Center and questionnaire automation, plus SSO/SCIM and custom integrations) and Defense (quote only; CMMC tooling such as SSP and POA&M). Additional workspaces are an add-on.ComprehensiveOct 15, 2026
SprintoIndia2020—No public prices. Three plans (Foundation, Growth, Platform) are quoted by sales. Foundation includes 25+ automated frameworks, 20 AI-powered security questionnaires a year and an AI systems inventory. Add-ons include additional frameworks, AI Governance, Enterprise Risk Management, Enterprise Trust Management and Enterprise Vendor Risk Management.ComprehensiveOct 15, 2026
DrataUnited States2020—GRC plans (Foundation, Advanced, Enterprise) and Assurance plans (Trust Center and questionnaires) are quoted; drata.com shows no prices. Its AWS Marketplace listing charges a $25,000 platform fee for a 100-FTE organization plus $7,500 per framework, for a 12-month contract.ComprehensiveOct 15, 2026
VantaSan Francisco, United States2018501-1000Four plans: Essentials, Plus, Professional and Enterprise. No prices on vanta.com; its AWS Marketplace listing gives 12-month starting prices for 1-20 employees: Essentials $14,000, Plus $21,500, Professional $23,000. Enterprise is quoted. Third-party risk management is a separate product ($13,600 a year for up to 50 vendors on AWS).ComprehensiveOct 15, 2026
SaidotHelsinki, Finland201811-50No public pricing listed; contact sales implied via demos and sign-ups.PartialSep 21, 2026
HiddenLayerAustin, United States202251-200Enterprise-only, contact sales for pricing. No public pricing listed on website.PartialSep 21, 2026
OneTrust AI GovernanceAtlanta, United States20161000+Enterprise platform; contact sales for quote, no public pricing listedPartialSep 21, 2026
IBM watsonx.governanceArmonk, United States—1000+SaaS: Free trial (Essentials, max 200 resource units, max 3 use cases); Standard plan at USD $0.60 per resource unit (up to 50K records/evaluation, 500 inventories). Software (on-prem) priced by virtual processor core (VPC).PartialSep 21, 2026
Credo AIPalo Alto, United States202051-200Contact sales for enterprise subscription quote.PartialSep 21, 2026
Holistic AILondon, United Kingdom202051-200Enterprise platform; contact sales for quote.PartialSep 21, 2026
FairNowMcLean, United States202311-50Contact sales for quote; no public pricing listedPartialOct 15, 2026
Lasso SecurityTel Aviv, Israel202311-50Enterprise pricing only. Not publicly listed.PartialSep 21, 2026
CraniumShort Hills, United States202351-200Enterprise pricing not publicly listed; annual subscription model referenced in Microsoft Marketplace listing.Microsoft MarketplacePartialSep 21, 2026
CitrusˣTel Aviv, Israel202111-50Enterprise pricing not publicly listed; demo available upon request.PartialSep 21, 2026
2021.AICopenhagen, Denmark201651-200Pricing is custom and tailored to use case, organization size, deployment model (on-prem or cloud), and selected modules. Contact sales required per their pricing page.PartialSep 21, 2026
QL SecurityUnited Kingdom——Contact for pricingPartialSep 21, 2026
ComplyAgentEuropean Union——Contact for pricingPartialSep 21, 2026
SetAIComplyEuropean Union——Contact for pricingPartialSep 21, 2026
LandingRedItaly——Contact for pricingPartialSep 21, 2026
Adeptiv AIIndia——Contact for pricingPartialSep 21, 2026
MeilynxUnited States——Contact for pricingPartialSep 21, 2026
HyperproofBellevue, United States2018—No public prices or plan names on its pricing page; quoted on request. The AI-native third-party risk product includes a baseline of AI credits, with more sold at discounted tiers.PartialOct 15, 2026
ThoropassNew York, United States2019—No public prices or plan tiers. Services (Thoropass Audit, Audit-Ready Pentesting, Vulnerability Scanning, Compliance Suite) are sold on a tailored quote.PartialOct 15, 2026
Noma SecurityNew York, United States2023—No prices on noma.security. AWS Marketplace lists the Noma Security Platform at $200,000 for a 12-month contract but doesn't say how many units that covers.PartialOct 15, 2026
Fiddler AIPalo Alto, United States2018—Three plans: Free (real-time guardrails for hallucinations, toxicity, PII/PHI, prompt injection and jailbreaks, powered by Fiddler's Centor models); Developer at $0.002 per trace (adds unified observability with tests and experiments for agentic and predictive systems, custom evaluators and bring-your-own judge, role-based access control and SSO; SaaS only); and Enterprise, priced by sales (adds enterprise-grade guardrails and scale, SaaS, VPC or on-premises deployment, and a named customer success manager). The pricing page lists no trace or user limits.AdjacentOct 15, 2026
WitnessAIMountain View, United States2023—No prices on witness.ai. Its AWS Marketplace listing gives 12-month prices: WitnessAI for Enterprise at $180 per user (minimum 1,000 users, $180,000), WitnessAI Agentic Visibility at $24 per agent (minimum 2,500 agents, $60,000) and WitnessProtect model protection guardrail at $300,000.limitedOct 15, 2026

Buyer’s guide

Independent ranking with documented criteria.

See our top picks for ISO/IEC 42001

Looking for an audit firm?

Compare 12 independent ISO/IEC 42001 audit firms by accreditation, region, and services.

See all ISO 42001 audit firms

Compare across industries

See which vendors support ISO/IEC 42001 in your sector.

Frequently asked questions

+Is ISO 42001 mandatory?

No. Certification is voluntary and the standard does not replace any law [2]. It is not a harmonised standard under the EU AI Act either [13]. Customers can still require it in contracts.

+Can you get ISO 42001 certified if you only use AI and don’t build it?

Yes. The standard applies to organizations that provide or use AI systems [3][2]. A deployer excludes the controls that don’t apply, such as some development controls, in its Statement of Applicability with a justification [22].

+How many controls are in ISO 42001?

Annex A has 38 controls in 9 areas, A.2 to A.10, from AI policy to third-party and customer relationships [4][6]. Organizations select the controls they need and justify any exclusions in a Statement of Applicability [22].

+How long does ISO 42001 certification take?

There is no official benchmark. Compliance vendor Vanta puts it at 6 to 12 months for programs run on manual processes and about 3 to 6 months with automation [27], and Stage 1 and Stage 2 audits each take several days depending on scope [18]. Our certification guide breaks down the steps.

+How much does ISO 42001 cost?

The standard costs CHF 225 from ISO [1]. Certification cost depends mainly on audit days, which certification bodies calculate under ISO/IEC 42006 [28], plus any consulting and internal time. Use our cost calculator for an estimate for your organization.

+Who can certify you for ISO 42001?

An independent certification body. Pick one accredited for ISO/IEC 42001 by a national accreditation body such as ANAB or UKAS, and check its accreditation in that body’s directory [8][19]. ISO does not issue certificates [2].

+Does ISO 42001 certification satisfy the EU AI Act?

No. It is not a harmonised standard and gives no presumption of conformity [13][15]. The first harmonised AI Act standard is EN 18286:2026 for quality management, cited in the Official Journal on 14 October 2026 [14]. An existing AIMS is good groundwork, but it does not replace an AI system’s conformity assessment [15].

+What is the difference between ISO 42001 and ISO 27001?

They share the same management-system structure, so they can run as one integrated system [3]. ISO 27001 governs information security; ISO 42001 governs AI, adding AI risk and impact assessments and 38 AI-specific controls [3][4]. One certificate does not cover the other.

Sources

  1. [1]ISO, ISO/IEC 42001:2023 — Information technology — Artificial intelligence — Management system (catalog page) (2023-12).
  2. [2]ISO, ISO/IEC 42001 explained: what it is (2026-08-31).
  3. [3]ISO/IEC via iTeh Standards, ISO/IEC 42001:2023 official sample (foreword, contents, introduction, scope, clause 4).
  4. [4]Microsoft, hosted by NIST AI Resource Center, NIST AI RMF to ISO/IEC FDIS 42001 AI Management System Crosswalk.
  5. [5]NIST AI Resource Center, AI RMF crosswalks.
  6. [6]LRQA, ISO 42001 checklist.
  7. [7]Schellman, What are the ISO 42001 requirements?.
  8. [8]ANAB, ISO/IEC 42001 AI management systems accreditation program (2024).
  9. [9]ISO, ISO/IEC 42006:2025 — Requirements for bodies providing audit and certification of AI management systems (2025-07).
  10. [10]ISO, ISO/IEC 42005:2025 — AI system impact assessment (2025-05).
  11. [11]ISO, ISO/IEC 23894:2023 — Guidance on risk management (2023-02).
  12. [12]ISO, ISO/IEC 22989:2022 — Artificial intelligence concepts and terminology (2022-07).
  13. [13]European Commission, Understanding standardisation under the AI Act (FAQ) (2026-03-10).
  14. [14]Official Journal of the EU, Commission Implementing Decision (EU) 2026/2274 on harmonised standards for high-risk AI systems (EN 18286:2026) (2026-10-14).
  15. [15]LRQA, What ISO 42001 does and does not get you under the EU AI Act (2026-10-11).
  16. [16]NIST, AI Risk Management Framework (2026-08-13).
  17. [17]Bureau Veritas, Certification process (three-year cycle, surveillance and recertification).
  18. [18]Schellman, Guide to ISO 42001 certification.
  19. [19]UKAS, UKAS grants first AIMS accreditation (2026-01-15).
  20. [20]Schellman, Schellman becomes first ISO 42001 ANAB-accredited certification body (2024-09-24).
  21. [21]ISO42k, State of ISO 42001 (tracker of public certification announcements) (2026-10-11).
  22. [22]Konfirmity, ISO 42001 Statement of Applicability.
  23. [23]ISO, ISO/IEC 27001:2022 — Information security management systems (catalog page) (2022-10).
  24. [24]AICPA & CIMA, System and Organization Controls: SOC suite of services.
  25. [25]Cloud Security Alliance, Research note: EU AI Act prEN 18286 and ISO 42001 (2026-04-28).
  26. [26]Official Journal of the EU, Regulation (EU) 2026/1744 (Digital Omnibus on AI) (2026-07-24).
  27. [27]Vanta, ISO 42001 certification cost.
  28. [28]Standards Council of Canada, Bulletin: transition to ISO/IEC 42006:2025 (audit time and competence requirements) (2025-08-28).

How we keep this page current

Every claim links to its source. We re-check this page at least every 90 days, and sooner when a standard, law or vendor changes; vendor names, acquisitions and shutdowns update automatically from our directory. Vendors cannot pay to be included or ranked.

Informational summary only — not legal or certification advice.