Checked against primary sources on 18 October 2026.
Who it applies to
Organizations of any size that develop, provide or use AI-based products or services and choose a management system to govern AI risk, often to answer customer and procurement questions.
Who checks
Auditors from accredited certification bodies, whose AI-specific competence and audit requirements are set by ISO/IEC 42006:2025; customers then review the certificate and scope.
How often
Two-stage initial certification audit, surveillance audits each year, recertification every three years
Current edition
ISO/IEC 42001:2023 (Edition 1), 2023-12-18. Paid standard (CHF 225 for English PDF + ePub on the ISO Store).
Dates to know
Dates for ISO/IEC 42001
Date
What happens
Status
2025-05-28
ISO/IEC 42005:2025 AI system impact assessment guidance published
Done
2025-07-07
ISO/IEC 42006:2025 requirements for bodies auditing and certifying AI management systems published
Done
2026-01-01
Global Accreditation Cooperation Incorporated replaces IAF and ILAC as the international accreditation body
Done
What ISO/IEC 42001 asks you to write down
Each one needs an owner, an approval, and proof it reached the people it covers.
Clause 5.2 and Annex A.2.2
AI policy
Top management sets an AI policy suited to the organization's purpose that frames AI objectives, commits to meeting requirements and improvement, and is communicated internally.
Top management approval and date
Version in force and review history
Record of who received it
Annex A.2.3 and A.2.4
AI policy alignment and review
The AI policy is checked against other organizational policies it touches and reviewed at planned intervals, or sooner when changes affect its suitability.
Related policies checked for alignment
Review dates and outcome
Clause 5.3 and Annex A.3.2, A.3.3
AI roles and reporting of concerns
Roles and responsibilities for AI are defined and assigned, and a process lets people raise concerns about the organization's role with AI systems.
Named role holders and their AI duties
Concern-reporting route and handling record
Clauses 6.1.2 and 6.1.3
AI risk assessment, treatment and Statement of Applicability
Documented AI risk criteria and assessment results, a treatment plan, and a Statement of Applicability justifying included and excluded controls.
Risk criteria and dated results
Control justifications and implementation status
Clauses 6.1.4 and 8.4
AI system impact assessment
A documented process assesses potential consequences of AI systems for individuals, groups and society, with results retained and repeated at planned intervals or on significant change.
Assessment results per AI system
Date and trigger for each reassessment
Clause 7.3
Awareness of the AI policy
People doing work under the organization's control know the AI policy, their contribution to the AI management system and the implications of not conforming.
Who was briefed and when
Policy version each briefing covered
Citations are to the standard or regulation; wording is ours.
What gets asked for on the day, inferred
Inferred from the clauses and general audit practice.
AI policy approval and reach
Auditors ask for the approved AI policy, evidence of top management sign-off and records showing it reached the people who build, buy or use AI systems.
Impact assessment records
They sample AI systems in scope and ask for each impact assessment, its date and what triggered any reassessment.
Statement of Applicability
Auditors read control selections and exclusion justifications in the SoA against the AI risk assessment results.
Awareness and concern reporting
Interviews check that staff know the AI policy, their responsibilities and how to raise a concern about an AI system.
No public auditor checklist exists for ISO/IEC 42001 yet. These points are inferred from the clauses and from general audit practice, and will be replaced with published guidance when it appears.
The policy and acknowledgment layer for ISO/IEC 42001.
Write each policy, send it through approval, and send it to the people it covers. Each acknowledgment is recorded by name, with the version, the date and the time.
When a policy changes, the new version is approved, its Editor sends it out again, and everyone it applies to is asked again. Export the record as a PDF or CSV, or give a reviewer Viewer access to the documents their tags allow.
Approved before it goes outUp to four steps, with any or all approvers.
Acknowledged by nameName, email, date, time and version on every record.
Asked again on every changeEach new version goes out again to everyone it applies to.
The record on requestA PDF or CSV export, or Viewer access to the documents their tags allow.
Your region’s laws, the standards your customers ask for, and your own house rules. Write each one as a policy, approve it, and send it to the people it covers.
How does ISO/IEC 42001 sit alongside ISO/IEC 27001?
Both use the ISO harmonized structure, so clauses 4 to 10 line up and one document control process can serve both systems. Annex A differs: ISO/IEC 42001 centers on AI policy, impact assessment, data, the AI life cycle and third-party relationships.
What is ISO/IEC 42006?
Published in July 2025, ISO/IEC 42006 adds AI-specific requirements for bodies that audit and certify AI management systems, on top of ISO/IEC 17021-1. It governs auditor competence and audit conduct and adds no requirements for the audited organization.
Where does ISO/IEC 42005 fit?
ISO/IEC 42005:2025, published in May 2025, gives guidance on carrying out AI system impact assessments. It can be used to structure the impact assessment process that clause 6.1.4 asks for.
Does the standard require acknowledgment of the AI policy?
Clause 5.2 asks for the AI policy to be communicated within the organization, and clause 7.3 asks for awareness of it. A dated, version-specific acknowledgment is a common way to evidence both, though the form is left to the organization.
Is a new edition on the way?
As of October 2026, ISO lists the 2023 edition as current with no amendment or revision under development. The standard was first published in December 2023.