Harmonizing High-Tech: The role of AI standards as an implementation tool

International leaders discuss the future of AI standardisation at panel event

During a panel discussion on AI governance and standardisation, leaders from the International Electrotechnical Commission (IEC), the International Organization for Standardization (ISO), and the International Telecommunication Union (ITU) came together to elucidate their roles in the evolving landscape of AI standards. Philippe Metzger, CEO of IEC, highlighted the longevity and complementarity of the three organisations, which are all based in Geneva and have been instrumental in shaping the standardisation framework for various technologies over the years. Emphasising principles such as transparency, inclusivity, and consensus-based decision-making, Metzger underscored the importance of these values in the development of AI standards. He also touched upon the areas of AI-specific standards being developed, including AI trustworthiness, sustainability, functional safety, and data quality.

Sergio Mojica, CEO of ISO, stressed the necessity for coordination between the three organisations to provide clear and unified standards for AI, which would prevent confusion and duplication in the field. He mentioned the World Standards Cooperation (WSC) as a platform for technical collaboration and for promoting a positive message about standardisation. Mojica also highlighted the need for inclusivity and active engagement with developing countries to ensure that they have a voice in the standardisation process, which is particularly relevant for AI.

Sezio Onoe, Director of the Telecommunications Standardization Bureau, underscored how standards support public-private collaboration and the achievement of policy objectives. He provided examples of ITU’s standards in various domains such as network orchestration, multimedia coding, energy efficiency, and collaborations with other organisations in health, agriculture, and natural disaster management. Onoe also addressed the dual nature of emerging technologies, acknowledging the need to mitigate the negative aspects of AI while promoting its positive potential for sustainable development.

As the discussion concluded, the panelists collectively reassured the audience that the World Standards Cooperation is effectively coordinating the AI standards landscape. They affirmed their commitment to continue this collaboration, cooperation, and to be inclusive of other efforts around AI standards. The panelists’ consensus was a clear indication of their dedication to ensuring that AI development and governance are underpinned by robust, globally recognised standards that foster safety, interoperability, and trust.

Bilel Jamoussi:
cooperation. The CEO of IEC, Mr. Philippe Metzger, the CEO of ISO, Mr. Sergio Mojica, and the Director of the Telecommunications Standardization Bureau, Onoe san. Please join me at the stage. As I load my next program. Welcome. This is the final stretch, but it’s a very important element of the whole discussion this morning. We’ve been talking about the governance of AI, lots of discussions on having standards to base that governance on and to help in moving forward with AI. We’re very pleased to have the World Standards Cooperation leaders with us today. The three international standards organizations based here in Geneva, IEC, ISO, and ITU. I’ll start with Filip Metzger, the CEO of IEC. Filip, if I may, to help set the scene for our discussion. Could you please explain the position that the IEC, ISO, and ITU hold in the diverse standardization landscape?

Philippe Metzger:
Yeah. Good evening, everyone. And thanks very much, Bilel. I think when I answer that question, I’m still under the impression also of the discussion we had this morning, a fantastic discussion on the governance challenges I think that we have. And I think many of the hallmarks of the three organizations that we are representing here will hopefully also help address those challenges. And I think the starting point is that we are all quite old institutions. We certainly consider ourselves leading standards development organizations. If I look at ITU, probably the oldest global international organization founded in 1865. ISO, 1947. IEC, 1906. So I think we have a long track record. We consider ourselves also sister organizations because we are, I think, quite complementary in what we are doing. And all of these contributions will play a very important role, and already do, and will more do so in the future in the AI context. We are based on sound principles. If I again refer to the discussion this morning, very important aspects. What principles we use? Transparency, diversity, also inclusivity, and equal opportunities. I think that was a sort of a current underlying theme that we had in the early discussions on governance. We are consensus-based, so these standards that we are producing are really reflecting global involvement of global communities, and also are based on consensus. The processes we apply, they are, I would say, tried and tested. Of course, in this day and age, it’s a bit of a difficult thing to say. It sounds very sort of self-content. That’s not at all the case. We are trying to adapt also our way of working to new methods. But ultimately, we have processes which are stable and which are also reliable. I would also emphasize, in terms of the principles, that what we are doing, the way we function, is really designed for global impact. We are all-encompassing, and we, of course, are also very aware of the challenges of including, especially developing countries and communities, even more going forward. All in all, I think we are very well-positioned for fostering collaboration in the standardization world, and that’s obviously many more other organizations, regionally and globally, or locally, that contribute to that. We are also very open to that collaboration beyond our own remit. If you look at the history of technology and come to AI, as today’s main topic, I think our systems have proven to be quite adaptable, despite the long or the old age our organizations have. We started, in our case, with electrical safety in 1906. It then developed into efficiency of also energy production. Certainly, then IT came into the fore, and AI, of course, is a next big dimension, stemming also from all those foundations that we already have been working on. I think we managed to keep going and addressing those new challenges and contributing to that as well. Quite importantly, we’re also important drivers for trade and for prosperity globally. We are recognized by the World Trade Organization as indispensable for fostering the elimination of technical barriers to trade. We turn out to be very relevant also for regulation and legislation. If you look at how our standards are used, they are voluntary standards, of course, but they are of much interest to regulators and also to legislators. If I take the example of Europe as the most advanced integrated regional market in Europe, 80 percent or more of the standards are actually IEC standards in our area or based on IEC standards. I think that also proves the relevance also with a view to regional or local specificities. What I really would like to say as a penultimate point, we are very open to collaboration and interactions with others. We have tools for that. We call it liaisons in our case. That means it’s not only our members who already are, I think, globally represented, but also new players, for instance, who can work with us. To end my brief overview, fundamentally, I think, ITU, ISO, IEC, we are enabling safety and security. We are enabling interoperability, a very important dimension, of course, also in the AI discussion governance. We are enabling trust and hopefully all that leads in the context of our discussion today. to responsible adoption. When I speak about responsible adoption, I speak about AI technologies

Bilel Jamoussi:
Fantastic. Thank you, Philippe, for that excellent introduction of the three organizations. Between ISO, IEC, and ITU, there are hundreds of years of experience in developing standards. The due process that you mentioned is ready and has been embarking on AI standards. So, it’s just another topic, another technology, another frontier that the three organizations are very well equipped to handle. Which brings me to Sergio Mojica, the and promote coordinated action from other bodies? ISO CEO. What are some of the practical ways that the three organizations coordinate their work

Sergio Mujica:
Very well, thank you, Bilel, and good afternoon to all of you. I think the most important point here is the why. Why do we need to coordinate ourselves? And the point is that we all believe, as mentioned by Philippe and as discussed this morning, is that international standards can be really instrumental for us to address global challenges and also to enable us to capture the opportunities, and in particular those emerging from AI. But it’s also true that this is a very crowded field, and a lot of organizations, governments, and companies, they’re really confused about what are the relevant standards that they should use. And I think our responsibility as leading organizations in the area of standardization is to provide something that makes sense to everyone. Because at the end of the day, people don’t care that much whether this is an ITU, an IEC, or an ISO standard. What they want to know is what is it what they need to address the challenges and opportunities related to AI. So that is our main responsibility. So we are different organizations, but we do believe in the same values. And the level of collaboration will also depend on the topic that we’re talking about. In some areas, it’s enough with exchange of information. In other areas, we need to be more ambitious to look for opportunities for interoperability or mutual recognition. And probably the ultimate goal, when appropriate, is to have a co-creation of international standards that could have the logo of the three organizations. So one single solution. So all of this looks very nice, but it means nothing if we do not create concrete tools for us to operationalize our collaboration. So that’s the reason why the World Standard Cooperation, WSC, exists. And we focus on two main areas. The first one is the technical area. Because the issue is as follows. We do not create solutions looking for problems. I do not tell you from Geneva what is it what you need. We work in a very decentralized manner, and it is our members in 170 countries that will tell us, I need a solution. I need a new standard. Please initiate a standard development process. But the issue with that, which is great, actually, because it is really bottom up, is that there is not necessarily a good level of coordination or overarching approach in all those standardization activities. So we have asked our experts, the technical management boards of the three organizations, to get together and to coordinate our work programs. In some times, that will end up with an exchange of information, as I mentioned before. In other times, it will end up in co-creation of standards. And I think we have a wonderful example when we talk about the high efficiency video coding, which is the fancy solution that we have created to compress the ultra high definition TV videos, which are really heavy. And if you compress them, you need to use a similar code in order not to lose quality. Well, we did that many years ago, and we’ve been together recognized with Emmy Awards. So that is a very good example of what can work well, but I think it’s also important. Two Emmy Awards. Two Emmy Awards, that’s true, absolutely. But I think it’s equally important to highlight that sometimes this collaboration has ended up with a lack of duplication of standards, which I think is equally valuable when it comes to final user. So that’s the first area, technical collaboration. The second one is promoting a positive message about standardization. Because I can give you a new standard in two weeks, if I hire two, three experts, high level experts, I bring them to my office in Geneva, and in two weeks, I have a new standard. It doesn’t work like that. We need, as mentioned by Philip, proper governance, and we need to respect some core values we believe in, because we work with full inclusivity, transparency, by consensus, and making it sure that developing countries have a voice. This cannot be a conversation among rich people or highly sophisticated organizations. We need to make it really inclusive, and that is particularly relevant for AI. So we promote together the importance of standardization in this inclusive manner.

Bilel Jamoussi:
Fantastic. Muchas gracias. Onoe-san, how do standards support public-private collaboration and the achievement of policy objectives?

Sezio Onoe:
Thank you, Philippe. Good afternoon, everyone. I can talk within two minutes. Actually, my belief that standardization itself is the outcome of the public-private collaboration. So actually, in the process of the standardization, we have many involvement from the many different stakeholders, governments, regulators, companies. So it’s a public-private collaboration. So I believe that standards can support public-private collaboration. So actually, it works as follows. Voluntary specification, voluntary standards can offer valuable support to the achievement of policy objectives. And then they can provide policymakers with a powerful tool to accelerate sustainable development, for example. So ISO, IEC, ITU, they want to advocate such activities in the event in our discussion in the G20 event and the COP28 and COP29 plan. So that happens. So then I can give another example in the ITU. I would say that this is just an example. Regulators promoting good service quality work together with companies specializing in service quality KPIs and monitoring tools to develop standards that capture the common understanding between companies and regulators overseeing that industry. That’s one example. And the second point is that policy and regulation can establish rules. And technical standards can provide powerful tools, practical tools to uphold the rules. Then I can give you some example for safety and privacy. It’s not limited to the technology. It’s beyond technology. But the security controls set by standards are practical tools to ensure safety and privacy. As such, standards are technical tools supporting key privacy policy objectives.

Bilel Jamoussi:
Thank you very much, Onoe-san. And now the next question, and we have about maybe 10 minutes or so, is to all the panelists. What are some of the issues that your standards on AI are addressing? So now we get into more of the AI-specific standards. Maybe I’ll start with you, Philippe. Thank you.

Philippe Metzger:
Thank you, Bilel. Maybe to be as succinct as possible, just would like to mention four areas, which I think all are also relevant in the context of the governance discussion. The first body of work, area of work, where I think we already have quite some track record now and where work is also ongoing, is the entire area of AI trustworthiness. We have, for instance, and this is essentially also work done through a joint technical committee between ISO and the IEC, Joint Technical Committee 1, Subcommittee 42, which I think has now a global recognition as well, has really done an incredible progress and achievements over the past few months and years. So for instance, there is guidance on addressing societal concerns, ethical considerations, when systems are developed that use AI. So it’s how you bring the non-technical aspects that we are discussing in the governance discussion into the technical work and how you can, by also exemplifying constellations, create an understanding, a global understanding of the experts, of the engineers, how they take those societal and ethical concerns into account when they do their technical work. Another example is work guidance on explainability and interpretability of machine learning models and AI systems. So you go really to the heart of how AI develops and of course also in a way to address dimensions that raise fundamental questions in terms of trustworthiness by the different stakeholders in society. We have work ongoing on treatment of unwanted bias. We know bias is a big topic in this context. And also there is guidance that has been elaborated and of course all this will continue to develop on human oversight of AI systems. I think a key topic in terms of the ultimate question of how good or how bad AI will be for humanity in the future. Another area of work in the same context of AI and the same technical committee is sustainability. We’re looking at the question of how AI systems can be crafted, created and operated in a sustainable way, looking at different aspects, including of course carbon impact and the like. And I think a further area that’s really important is functional safety. Functional safety, if you look at sliding doors, automatic doors, that’s precisely where you see functional safety at work. How do electronically controlled products, devices react when a constellation of danger is occurring and how can you rely that the door will really keep open when it should keep open. And there the work focus is now in the AI context, of course on systems that will use AI for functional safety and make sure that in the new era of AI using that, the functional safety is still reliable and still trustworthy. That’s an important aspect as well. And of course data in a sense as the lifeblood for all these AI systems is a big area of work as well, looking at characteristics and also at properties and quality of data and how that can really be used in AI processes so that at the end you get meaningful analytic results. I would like to just close with one point. IEC has one specificity as a standards organization. We are also operating so-called conformity assessment systems. And I think the question of conformity will be very key as well in the future when we speak about AI standards. And so this is, I think, forward-looking stuff that I’m telling you, I can’t offer you already solutions or services in that respect, but our experts are really thinking very much about how in the future the conformity can be assessed of products that use AI because that’s not really a sort of a constellation that’s cracked at the moment where we have already recipes we can apply, but we will of course want to make sure in the future that conformity assessment also stays relevant and produces the test results that we can do today for non-AI systems. And the other one, in a sense, the flip side of the coin is how can we use AI in conformity assessment? So in the process of testing and then issuing certificates, how can we leverage the AI technology? Those are sort of two main angles at which our conformity assessment experts are looking at the moment.

Bilel Jamoussi:
Great, thank you. Thank you, Philippe. Sergio, some issues that ISO is working on to resolve AI issues?

Sergio Mujica:
ISO and IC are working together on this because we have a joint technical committee on AI and Philippe has already shared a lot of concrete examples. Perhaps I will take a different angle. How can we help to address the issues and opportunities related to AI? How standards in particular can help? Well, first, it’s a trusted platform. So when it comes to developing new standard, this is a place where we can have a conversation by experts that take decisions by consensus and they have that conversation in a completely transparent and inclusive manner, invited to the table even those who disagree with what we are proposing. And then it’s the aspect that I mentioned before on developing countries. It cannot, it’s not enough to say developing countries are invited to participate in the conversation. You need to actively engage with them to make it sure that they have a voice. So that is one element. The second element is that we have already quite a robust portfolio of AI-related standards have been developed in this joint technical committee, subcommittee 42 that Philippe referred to. Two things to highlight there. First, the so-called foundational standards. Perhaps one area where we can make a huge contribution is in the definitions and terminologies. So we all talk about AI, advanced AI, risk, but we don’t mean the same thing. So if we could agree on the meaning of those terms, that is already a huge contribution. And the second one is the management standard on AI that we recently launched at the end of the last year, which is the first one of that nature worldwide, where we have, where we provide practical guidance to all kind of organization, SMEs, large corporation, even government about how to address AI. So it’s a very practical guidance about the implementation of AI in their own companies. It’s about how to address the risk related to AI. What is the role and responsibility in the various layers of the organization, including the management responsibilities and so on.

Bilel Jamoussi:
Thank you, Sergio. Onoue-san, what are some of the ITUT standards in AI?

Sezio Onoe:
Yeah, actually the ITU has published over 100 standards and also the 120, around 120 now ongoing. So actually the AI technology can be applied, naturally applied to the widely available technologies, like network orchestration and multimedia coding and energy efficiency for the networks and data centers. So in addition, we have some activities with a collaboration with other organization and in the other areas that create a new breakthroughs. We have a collaboration with, for example, the AI for health collaboration with WHO and now WIPO. And also we have AI for agriculture with collaboration with the FAO. And we have AI for natural disaster management with WMO and UNEP. As such, we have some area for this. But just I’d like to add one thing, for the, in general, emerging technology has always both sides, negative side and positive side. Actually the AI technology is a promising technology for to accelerate the sustainable development. But on the other hand, we need to effort to mitigate the negative side. So actually nowadays, AI-generated contents may cause a serious problem, like defects and misinformation. So now we have a new work item to verify the authenticity and provenance of multimedia content. So during the summit, we have a workshop on AI watermarking. That brings everyone together to identify the start on the right foot. So as such, we need a collaboration. I see ITU collaborate to provide unified frameworks for AI standard development.

Bilel Jamoussi:
Thank you very much, Onohe-san. I promised our Deputy Sec-Gen and the UNESCO ADG that we will finish on time. It is right on time. But before we ask for a round of applause, as we conclude this discussion during the governance day of AI and all the discussions we had this morning, could we reassure the governors of AI that the World Standards Cooperation is coordinating the AI standards landscape and they can rely on your leadership as the three international organizations going forward to continue this collaboration, cooperation, and be inclusive of other efforts around AI standards? It’s a yes, no. Of course, of course. Okay, a big round of applause. Thank you very much and sorry to put you on the spot, but I thought the answer was going to be yes. Thank you.