ISO/IEC 5338 will help to improve understanding of the functions and performance of AI systems during development, and promote mutual comprehension between developers and users, according to IEC.
It is based on existing standards, ISO/IEC/IEE 15288 and ISO/IEC/IEEE 12207, for system and software life cycle processes, with adjustments to support the definition, control and improvement of AI systems. ISO/IEC 5388 covers everything from key concepts to risk management, stakeholder needs, verification, maintenance and disposal.
Observations of processes based on use cases found in ISO/IEC TR 24030 to help identify AI-specific aspects are included.
AI systems such as machine learning, natural language processing, speech recognition and information retrieval use a combination of traditional software elements, such as code and databases, alongside those specific to AI, noted IEC.
ISO/IEC 5338 Project editor Yuchang Cheng said: “The standard expands on current established and widely-used life cycle processes for traditional software so that AI systems can benefit too. It enables developers to continually improve, resulting in more effective and efficient AI products, and instilling greater trust in users.”
Wael William Diab, Chair of the joint ISO and IEC committee for AI that developed the standard, SC 42, stated: “Having a common language and framework is key to ensuring responsible design, development and adoption of AI.”
SC 42 develops international standards for AI, taking a holistic approach to consider the entire AI ecosystem. It looks at technology capability and non-technical requirements, such as business, regulatory and policy requirements, application domain needs, and ethical and societal concerns.