Artificial intelligence is sweeping across our personal and professional lives, from the facial recognition apps on our mobile phones to the robotics that help streamline business operations. The need for responsible and ethical AI standardization is vital, given the complexities of data and security—especially as AI capabilities across industries continue to grow.
Formed five years ago, the International Organization for Standardization / International Electrotechnical Commission (ISO/IEC) Joint Technical Committee (JTC) 1, Information technology, Subcommittee (SC) 42, Artificial intelligence, is the first-of-its-kind international standards committee looking at the full AI IT ecosystem. The committee has grown to include 52 national bodies representing a wide variety of regions, with as many as 250 participants in virtual meetings.
The American National Standards Institute (ANSI), which holds the Secretariat of the committee, recently spoke with its chairperson Wael William Diab about some recent highlights and the future of AI standardization work. Over the past year and a half, the group has collaborated on publications across the entire work program, encompassing topic areas such as foundational standards, data, trustworthiness, ethics, use cases, computational approaches and characteristics, and governance.
Among its recent efforts to provide confidence to AI consumers and to enable third party audit/certification, SC 42 initiated the development of a management system standard. The committee is also exploring the AI data ecosystem by developing a 5-part series that looks at AI data quality, including a variety of aspects from terminology, measures, requirements, guidelines, process framework, and governance framework.
*Interview edited for clarity and style.
ANSI: Let’s talk about AI standardization. Can you tell us about the development of the recently published standard ISO/IEC 22989:2022, Information Technology - Artificial Intelligence - Artificial Intelligence Concepts and Terminology? How many stakeholders were involved, and why is this standard important to support AI?
WD: One of the most interesting aspects about this project and SC 42 in general is the stakeholders. With 35 countries voting at the DIS ballot stage, we had quite a few stakeholders interested in the project with a diversity of stakeholder perspectives participating. We had more than just the technologists at the table but also academics, regulators, application developers, ethicists, experts with legal backgrounds, and application domain experts, to name a few. Not only does this make the standard stronger in the long term, but it is reflective of how we use technology today.
AI has the ability to change how we live, work, and play. Thus, it stands to reason that the technology developers are joined by stakeholders that will use it, regulate it, etc. Based on that premise, it becomes increasingly important that everyone speaks the same language. We often find that even between technologists of different areas, misunderstandings can happen. Thus, having a common language is that much more important. The standard also serves as a building block for application domains that want to develop their AI standards by building upon it.
ANSI: Has SC 42 considered emerging regulatory requirements around the world that are intended to safeguard consumers of AI-enabled systems from potential issues such as ethical concerns?
WD: The approach SC 42 has taken by comprehensively looking at how AI plays a role in the full ecosystem has allowed us to deliver horizontal standards that can then be used by application committees and relevant stakeholders as the building blocks for their standards/guidelines/guidance, where they can rely on their domain knowledge and expertise along with our standards.
SC 42 continues to see diversity in participants that goes beyond the traditional technology participants in IT standards such as engineers, marketeers and data scientists. We are engaging with organizations, from both the public and private sector, that are approaching the deployment of AI systems with a focus on customers’ needs, segments, services, products, and regulatory requirements. SC 42 has also built relationships with outside communities including UNESCO, Organisation for Economic Co-operation and Development (OECD), and European Commission that bring in complementary perspectives based on their stakeholders. Government bodies, like other relevant stakeholders, have the ability to utilize the work of SC 42 as they begin to consider developing guidance relative to the implementation and/or deployment of various types of Artificial Intelligence Systems.
SC 42’s approach of proactively and concurrently addressing concerns such as trustworthiness, ethics, and societal implications by developing horizontal standards that can be incorporated into application domain standards, and developing a management system standard in conjunction with the diverse stakeholders contributing to the standards, will help ensure that regulatory requirements are addressed.
Voluntary international standards, such as those being developed by SC 42, that consider global requirements, across geographies and application domains, not only address the emerging issues, but also reduce fragmentation, which in turn reduces costs of building requirements, building deployments and ultimately ensuring AI systems meet those requirements.
ANSI: Let’s look into the future of AI: What’s next for JTC 1 SC 42 standardization work?
WD: That’s a great question. With a field that is rapidly evolving, not just technologically, but also in how we interact with the technology and our “requirements” for it, it is just as important to be able to plan as it is to execute. To address this, we launched a bi-annual ISO/IEC AI Workshopseries to foster a dialogue about the AI ecosystem. The workshop series looks at emerging trends, technology, requirements, and applications, as well as the critical role international standards play to enable innovation and responsible adoption. The workshops are freely available and provide insights to what areas we may want to look at next, as well as promote the AI standardization work to a variety of stakeholders that may otherwise not engage in standards development.
In terms of its work program, SC 42 is working on a variety of projects. We are also collaborating on a number of areas with sister committees in ISO, IEC and JTC 1 on projects such as AI sustainability, governance, security, cloud, multimedia coding, functional safety, AI testing, verification, and validation to AI-enabled health informatics. We are also seeing a continued focus on certification, new application domains for AI, and how AI can help achieve sustainability and societally beneficial outcomes.
We also continue to see new work in our existing areas such as trustworthiness, computational, applications, data, and foundational aspects of AI standardization. This also means that we have to do more with the same resources, which underscores the importance of collaboration and looking at mechanisms to gather input about what is important, through our workshop, outreach, and national body input.
Looking at our work program by the numbers, we are adding roughly 10 new projects a year, have over 50 national bodies participating and have expanded our ecosystem of collaboration to over 50 liaisons. For a committee that started with two projects and just under a dozen participating national bodies, I would say we are doing quite well.
ANSI: How can stakeholders get involved?
WD: It is never a bad time to join. We have plenty of challenges and we welcome anyone who is interested. In the U.S., INCITS has been given the responsibility by ANSI to manage the Technical Advisory Group (TAG). Stakeholders are more than welcome to contact me: [email protected] or our committee manager for anything related to SC 42.
About ISO/IEC JTC 1/SC 42
Now in its 34th year, ISO/IEC JTC 1 has addressed the rapidly changing standardization requirements of the global information and communications technology (ICT) industry, speeding the developmental process and the wide deployment of relevant standards.
The U.S plays a leading role in JTC 1, with ANSI, the U.S. member body to ISO, serving as secretariat.
Charged with serving as the focal point within ISO/IEC JTC 1 for AI standardization, SC 42 provides guidance to JTC 1 and ISO and IEC committees as they look to develop AI-related standards in their sector-specific application areas.