Explanations of ISO/IEC 42001:2023
Source is ISO/IEC 42001:2023(en). Title: Information technology — Artificial intelligence — Management system. It was prepared by the joint…
Read articleISO/IEC 42001:2023
For: V, IA
Source: ISO/IEC 42001:2023(en), English. A.4, B.4
The control objective A.4 is to ensure that the organization accounts for the resources of the AI system, including components and assets, in order to understand and address risks and impacts. Five reference controls, each to the extent it is necessary under 6.1.3.
A.4.2: determine and document the relevant resources required for the activities in given life cycle stages and for other AI-related activities of the organization.
A.4.3: as part of this determination, document information about the data resources used.
A.4.4: document information about the tool resources used.
A.4.5: document information about the system and computing resources used.
A.4.6: document information about the human resources and their competences that are used for development, deployment, operation, change management, maintenance, transfer and decommissioning, and for verification and integration of the AI system.
B.4 formulates the same sentences with “should”. Documentation of the resources is said to be decisive for understanding risks and possible positive and negative impacts, and can inform the impact assessment. Resources can comprise components, data, tools, system and computing resources and people, and originate from the organization, from customers or from third parties. If resources are not available, the organization should revise the design specification or the requirements for deployment.
B.4.3 names topics that the documentation on data should include, but is not limited to: provenance, the date of the last change, for machine learning the categories training, validation, test and production, categories of data according to ISO/IEC 19944-1, labelling, intended use, quality with a reference to the ISO/IEC 5259 series, retention and deletion, known or potential bias, preparation. ISO/IEC 19944-1 and the 5259 series are not available.
B.4.4 names possible tools, including algorithms and models, preparation, optimization, evaluation, deployment tools, and hardware and software for design, development and deployment. ISO/IEC 23053 is named as guidance and is not retold.
B.4.5 names possible information on system and computing resources, including resource needs, the location of the resources, processing and the effect of the hardware, as an example an environmental effect through use or manufacture, or the cost of use. Different resources can be needed for continual improvement and for development, deployment and operation. The note refers to resource considerations in ISO/IEC 22989.
B.4.6: the organization should consider diverse expertise. Necessary people can include, among others, data science, roles of human oversight, trustworthiness professionals such as for safety, security and privacy, researchers and domain experts. Different resources can be needed in different life cycle stages. The example of including demographic groups connected with training data is introduced with “for example” and “can”.
Not in this text: a resource list as a form, an obligation for every example category in B.4, and the content of ISO/IEC 23053, ISO/IEC 19944-1 and ISO/IEC 5259. The requirements, to the extent the controls are necessary, are the five documentation sentences in A.4.