With the rapid development of artificial intelligence (AI) technology, its application areas are changing and becoming increasingly. Whether in the European Union, the United States, or other advanced countries, preliminary work on the standardization of AI technology has already begun. ISO/IEC JTC1 SC42 is formed by the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC), that is a joint technical committee dedicated to developing standards and methods for AI. The structure of ISO/IEC JTC1 SC42 is composed of multiple working groups (WGs), which are also called technical working groups. There are currently five groups from WG1 to WG5 and co-working with each other. Each of groups has its own field responsible for the standards and applications related to AI technology, AI software/hardware, and AI supervision, respectively.
ISO/IEC JTC1 SC42 Committee: Responsible for the direction and indicators of the overall AI standard formulation, including guidance, instruction and other framework-related work in AI-related fields, and part of the work is to integrate the progress and connections of WG1~WG5, striving to strengthen the communication and AI infrastructure between WGs.
WG1: Responsible for the discussion of basic AI technologies and the formulation of relevant standards and specifications, including the standardization of basic AI concepts, AI keyword’s definition/compilation, and AI basic framework.
WG2: Responsible for the formulation of standards and specifications for quality management and quality governance of AI data and materials. The content mainly includes data measurement, big data management, data quality management, data architecture analysis and AI life cycle standard.
WG3: Responsible for the formulation of relevant standards, including AI models, model-building, trustworthiness, privacy, permissions and security of AI modeling methods, respectively.
WG4: Responsible for the formulation of standards and frameworks for mutual understanding between AI, stakeholders, users and application scenarios, including the formulation of relevant standards and specifications for adaptation and global application.
WG5: Responsible for the definition of the core architecture of AI hardware and software, including an overview of AI hardware, training efficiency, and the formulation of overall standards.
ISO/IEC JTC1 SC42 has developed and published a number of standards, including ISO/IEC 42001, ISO/IEC 23053, ISO/IEC 22989, and is scheduled to complete ISO/IEC 42005 and ISO/IEC 42006, etc. To connect with the international standard committee, based on the self-developed AI evaluation technology, Taiwan will attend international standards organization conferences continuously to discuss AI-related standards with other countries.
The ISO/IEC JTC1 SC42 committee will hold a meeting at the end of 2024. Based on ISO/IEC 42001, WG1 will begin to develop an audit guide for AI management systems. WG2 will gradually build a process and data management framework for big data analysis. WG3 will focus on making the NIST RMF comprehensive, and trying to establish a common vocabulary. WG4 has built and developed the projects based on the feasibility of human-machine collaboration standards to research the potential risks. WG5 focuses on standardizing the computing core and framework of the AI system including the model applicability and performance evaluation process. Based on these plans, the ISO/IEC JTC1 SC42 committee will continue its work on AI regulation, analysis and standardization.
