ISO/IEC 42001: a management system for AI, explained

What it is

ISO/IEC 42001:2023  (opens iso.org, external site), published in December 2023, specifies requirements for establishing, implementing, maintaining and improving an artificial intelligence management system. It applies to any organisation that develops, provides or uses AI, of any size. Unlike guidance documents, it is certifiable: an accredited certification body can audit an organisation against it.

How it is structured

The standard follows the harmonised structure shared by ISO management system standards, so organisations already certified to ISO/IEC 27001 or ISO 9001 will recognise it.

• Clauses 4 to 10: context of the organisation, leadership, planning, support, operation, performance evaluation and improvement.
• Planning includes an AI risk assessment, an AI risk treatment plan and an AI system impact assessment, which looks at consequences for individuals and society.
• Annex A sets out reference controls across nine control objectives, from AI policy and internal organisation to data, the AI system life cycle, use of AI systems, and third-party relationships.
• Annex B gives implementation guidance for those controls; Annexes C and D cover objectives, risk sources and use across domains.

As with ISO/IEC 27001, an organisation selects the controls that apply and records its choices in a statement of applicability, with the reasons for including or excluding each.

Related standards

• ISO/IEC 23894:2023 gives guidance on AI risk management.
• ISO/IEC 42005:2025 gives guidance on AI system impact assessment.
• ISO/IEC 42006:2025 sets requirements for bodies that audit and certify AI management systems.

What it asks of agents that act

The standard is technology-neutral, but several control areas land squarely on agents that change records.

• Use of AI systems: defined objectives and processes for responsible use, including where a person must be involved.
• AI system life cycle: documented requirements, verification and validation before deployment, and event logs during operation.
• Information for interested parties: the ability to explain to users, customers and auditors what the system did.
• Third-party relationships: clarity about responsibilities when AI components come from suppliers.

An auditor assessing these controls will ask for evidence. A policy that says "risky actions require approval" is a start; a record showing which actions were approved, by whom, under which rule version, is what makes the control demonstrable.

How elan1 maps to it (draft)

• Operational controls as rules: elan1 Govern turns approval policy into rules evaluated before every agent action.
• Evidence of operation: every action, allowed, approved or refused, leaves a signed, hash-chained record that can be exported for audit.
• Verification before deployment: elan1 Work tests ERP changes against a copy of your data before they go live; elan1 Decide runs in shadow before any automation.
• Supplier clarity: elan1 Run places each AI component behind a versioned contract, so third-party responsibilities are named.

elan1 is not certified to ISO/IEC 42001, and using elan1 does not make an organisation compliant. ISO/IEC 42001 certification is also not, on its own, conformity with the EU AI Act.