logo
language iconarrow
menu icon

ISO/IEC 42001:2023

42001-01 Who the standard applies to

4 min read · V

All articles

For: V
Source: ISO/IEC 42001:2023(en), English. Introduction, Clause 1, Clause 2, 4.4

The introduction describes AI as increasingly used in all sectors that use information technology, and as an expected economic driver. Certain applications can create societal challenges in the coming years. The document aims to help organizations fulfil their role with respect to AI systems responsibly. As examples the introduction names using, developing, monitoring or providing products and services that use AI. It highlights three considerations, each with “can” or as a description, not as a list of requirements: automated decision-making, sometimes non-transparent and not explainable, can call for a control that goes beyond classical IT systems; data analysis, insight and machine learning, instead of logic written by humans alone, change development, justification and provision; systems with continual learning change their behaviour in use and need particular consideration so that responsible use continues when the behaviour has changed.

The document provides requirements for establishing, implementing, maintaining and continually improving an AI management system in the context of the organization. Organizations are expected to direct the application of the requirements towards characteristics that are peculiar to AI. Continual learning and improvement, or a lack of transparency or explainability, can justify other protective measures when they create additional concerns compared with the conventional performance of the same task. Introducing an AI management system in order to extend existing governance structures is a strategic decision.

Needs and objectives, processes, size and structure, and the expectations of interested parties influence establishment and implementation. So do the variety of use cases and the balance between control mechanisms and innovation. Organizations can apply the requirements with a risk-based approach so that the degree of control fits the use case, the service or the product within the scope. “Can”. These influencing factors are expected to change and to be reviewed from time to time.

The system should be integrated into the processes and the overall governance structure. AI-specific questions should be considered in the design of processes, information systems and controls. “Should”. As decisive examples of such governance processes the introduction names: setting organizational objectives, involvement of interested parties and organizational policy; dealing with risks and opportunities; processes for concerns about trustworthiness, for example security, safety, fairness, transparency, data quality and the quality of AI systems throughout the life cycle; processes for suppliers, partners and third parties that provide or develop AI systems. These are examples from the introduction. The document says that it provides guidelines for the use of applicable controls in support of such processes, and that it avoids specific guidance on governance processes. The organization can combine recognized frameworks, other International Standards and its own experience, for example for risk management, life cycle and data quality. Annex B is nevertheless the normative implementation guidance for the controls in Table A.1, with the limit from B.1. That is covered in 42001-21.

An organization that fulfils the requirements can generate evidence of its responsibility and accountability with respect to its role regarding AI systems. “Can”. The order of the requirements reflects neither their importance nor the order of implementation. Lists are numbered only for reference.

The section on compatibility says that the document uses the harmonized structure. The AI management system provides requirements that are specific to dealing with issues and risks arising from the use of AI. The same approach facilitates implementation and compatibility with other management system standards, for example on quality, safety, security and privacy. Annex D sets this out as informative guidance, see 42001-20.

Clause 1 specifies that the document specifies requirements and provides guidance for establishing, implementing, maintaining and continually improving an AI management system in the context of an organization. It is for an organization that provides or uses products or services which use AI systems. It is intended to help the organization develop, provide or use AI systems responsibly, pursue its objectives and fulfil applicable requirements, obligations towards interested parties and expectations from them. The English “is intended to” is purpose, not a second “shall”. It applies to any organization, regardless of size, type and nature, that provides or uses such products or services. Clause 1 does not set specific performance criteria for AI. Clause 1 does not contain a sentence that conformity is permissible only with no exclusion at all.

Clause 2 names ISO/IEC 22989:2022 as the only normative reference. Its content is not set out here.

4.4 requires an AI management system, including the processes needed and their interactions, to be established, implemented, maintained, continually improved and documented, in accordance with the requirements of this document. “Document” stands in this sentence beside the four usual verbs.

Not in this text: a definition of “AI system”. That stands in ISO/IEC 22989:2022; this edition is not available. No article on the EU AI Act. The introduction names societal challenges as background and lists no legal acts.