Karine Perset: Building Bridges for Global AI Governance

[MUSIC] Hi, I'm Kevin Wuerbach, professor of legal studies and business ethics at the Wharton School of the University of Pennsylvania. For decades, I've studied emerging technologies from broadband to blockchain. Today, AI is promising to transform our world. But AI needs accountability, mechanisms to ensure it's developed and deployed in responsible, safe, and trustworthy ways. On this podcast, I speak with the experts leading the charge for accountable AI. [MUSIC] The global AI policy world is complex and it's not limited to national policies or even formal international agreements. My guest, Karine Persit, is the acting head of the OECD's AI and emerging digital technology division, where she oversees the OECD.ai policy observatory and the global partnership on AI as well as other initiatives, which she'll talk about in our conversation. We discuss the unique role of the OECD in AI governance, starting with what is the OECD for those of you who are not familiar. And going to detail on some of the important data-driven initiatives that it's undertaking in the area. Karine, welcome. Thanks for joining me on the podcast. >> Thank you, Kevin. It's delighted to join you today. First of all, for those who aren't familiar, what exactly is the OECD? And how is it different from other larger intergovernmental organizations? >> So the OECD stands for the organization for economic cooperation and development. It's an intergovernmental organization that helps countries better design policies for economic and social well-being or better policies for better lives, as we say. It was founded in 1961 after the Second World War II. It's basically helped implement the Marshall Plan. The OECD now has 38 member countries. So they're market-based democracies like the US, the UK, Japan, Germany, and Australia, for example. We cover nearly all policy areas, basically mirroring the structure of a national government and focus on the areas in which governments need to cooperate and find common solutions to common challenges or learn from one another or actually need to cooperate because they're interdependent. So we cover all policy areas except for defense, which we leave to our colleagues at NATO. The OECD acts as a convening form, as I said, for learning from one another, countries learning from one another in each branch of government. So that's where the coordinate action on issues like education, innovation, inclusive growth for I think those cities best doing for its data-driven reports, a global policy recommendations and benchmarking tool like things like the PISA education rankings that come out every two years that rank high school students and then they've cognate as abilities, abstract reasoning abilities and components like that and that helps schools know where they compare on the international scale. Tax policy frameworks as well are quite important and trust for the AI Alliance in our case, which I will talk about more today. Tell me how did the OECD get involved in AI policy? So our work on AI policy has been ongoing for nearly a decade since 2016. And we've convened policy makers around an OECD AI Foreseq Forum. And since then our work has culminated around two main areas. One is the OECD recommendation on AI, which was from 2019. And the other big chunk of work I'd say is the OECD.AI Policy Observatory, which helps to bring those principles to life through data, live data, work feasible trends, tracking trends worldwide and tools for finding trust for the AI tools or metrics with partners and other resources. The OECD recommendation on AI was the first intergovernmental or standard on AI. It was adopted in May 2019 and was then adopted by the G20 and the form of the G20 Air principles. It was further updated recently in last year in 2024 to consider technological development. So it was lightly updated. Yeah, I was wondering, I was going to ask you about that about the update because obviously things change very fast in this area. So what are some of the ways that those principles have developed? The recommendation overall aims to foster trust and innovation in AI by promoting the responsible stewardship of AI and ensuring respect for democratic values and human rights and complementing existing standards in areas such as digital security, risk management, the recommendation focuses on AI-specific issues and sits as a standard that's implementable and flexible to stand the test of time. The updates pertain to requests by particularly the European Commission to have a definition of an AI system that they could use and that would be a universal definition so that we didn't have a definition in Europe and a definition in North America and so on. And so we're talking about the same thing when we're talking about artificial intelligence. So that was one of the updates where we made a few updates because at the time of writing in 2019 we had to not consider AI systems that in some instances will set sub-objectives based on overall objectives provided by humans. So we had written a definition of 2019 that were human specified, but just to be technically accurate in some cases, the sub-objectives or the actions to achieve a goal are not going to be detailed by a human. They will be figured out by the AI systems. So that was one of the changes, the other changes that countries which was not really needed because it was already encompassed was really to add the word content generation even though it was included in other types of actions. We felt at that point in time in 2024 this was the language model, the Lars language model, a generative AI period that they wanted that reflected specifically and so that those I think were the two main changes. Things did not stop there last year. The OECD which is, as I mentioned, a group of 44 democracies that are market economies and the global partnership on AI which was a somewhat related initiative decided to join force forces as the numbers of bodies interested in AI and multiplied it was a big trend. We tried to consolidate to avoid duplication to create synergy. So we joined forces to advance an ambitious agenda on AI embodying these principles, the OECD principles until this newly integrated partnership initially brings together OECD members and GP countries. So 44 in total across six continents and it aims to welcome new members including developing and emerging countries that are committed to implementing these principles. So it's OECD members plus large countries like India, Brazil, Argentina and others and I hope fully we have a list of countries who are which are counted accession countries. So we will find out more later on this year they should, the group should grow. How does the OECD bring all of those parties together given that there are so many different views and so many technical issues as well in terms of doing things like defining AI? I think the core value or one of the important components that the OECD brings to the equation is basing discussions on evidence. So we're not have data statistics trends that are from national statistical institutes or from other sources that really help us make well informed decisions and advise how countries discuss the best decisions and I would say and probably even more important with that than that is engaging with a variety of stakeholders and with partners from around the globe which is at the heart of the OECDs world and this has been continues to be the case in the work on AI and from businesses to the technical community and standards bodies and civil society as well as trade unions. The value is in the exchange between governments and those that are in the lab doing the work or trying to protect citizens or that really is the value ad. We're anchored in the OECD at principle so those are very solid and that we have a very large multi-stakeholder expert community as I'm and it's today about 800 experts that feed into the discussions on the OECD and within the global partnership on it. AI and brings tremendous value to governments in terms of specific topics that they need to work on. This expert community aims to translate the principles into practical guidance. We current this expert community has 10 expert groups today that respond to the priorities of member countries, help them on complicated topics, help them understand also evolving topics where AI systems are no longer typically one system but multiple systems with consequences from that interaction. They won't go through all of them, but just to give a few examples, one of the groups looks at AI incidents and hazards. So that is one of your, one of the evidence sources. So it's not hypothetical incidents and hazards, it's actually what has gone wrong where risks have materialized into actual incidents or could have materialized into actual incidents. And that group has recently developed a common reporting framework so that incidents are reported in much the same way in all countries because if they're not reported in the same way, we can't compare them. We can't track trends globally and that means that we were bound to reproduce errors made elsewhere. So that's an important component. We have an expert group on AI Futures which conducts foresight exercises to assess opportunities and risks of AI technologies in the medium and long term. And so this is more, each group will have different, with different balances constituents. And this would be more research oriented lab, technical folks working in labs on products that we will see tomorrow or in a few years, maybe a few months because they're going very fast. The expert group on AI risk and accountability is working on practical guidance for companies on AI value chains. And then other groups consider it measures a measurement of AI compute, which is actually quite difficult to compute. It sounds like it shouldn't be, but it is. And we're the first ones that are what we will talk about the measurements in September or how AI affects you. Workers and future work and those are just a two examples. So all of these expert groups feed into the OECD.AI policy observatory, which listeners can visit at OECD.AI. It's got a wealth of resources and tools and over 250 indicators on various dimensions of the AI ecosystem as well as right today, I think 1500 policy initiatives from 71 countries. And that's bringing rather quickly because more than what countries are inputting their own national AI policies or strategies to be able to learn, develop as a group, as a group. And also, we have some of the most important things that we can do and we have to do with the AI, especially the leading countries in AI. And we have to talk about the most important things that we can do with the AI. And we also have to focus on the most important things that we can do with the AI. And we can also think about the most important things that we can do with the AI. And we have a policy of AI policies in the world. And you can, again, it's at Boise.ai and it's countries and you'll see all the countries covered there, all the policies of those countries up to date. I think the important component here is not to have a database of national AI policies popping up every day, so we have new initiatives all the time. The life data also includes indicators on AI news, scientific publications, research publications, investments, AI jobs and skills, software development, search trends, and much more. So, taking it all together, they give you a picture of where our country stands and where we're developing an index. And AI index that is very AI-specific. Many AI indexes exist, but they're not necessarily AI-specific. You take into consideration other components that are related, digital, digital related, but not necessarily AI-specific. And this is all the indicators that composite our AI-specific. So the portal really leverages this life data to show timely trends about where, how, and what AI is being developed and used in which sectors. So it's a powerful tool to keep up to date, and I didn't encourage listeners to check it out. Fantastic. And that index would track countries' maturity in AI or what exactly would it seek to measure? Yeah. It's maturity in AI based on its development in policy areas in research and development. And investment in research and development. So investment is an important part in investment in infrastructure, in infrastructure and data and the availability of data, of compute, of, of, of, well, algorithms of knowledge about those things, the availability of, the availability of an enabling policy environment is obviously critical. Because it's the availability of talented workforce. And in our case, we also, not in the scenario, but talented workforce that has de-weight skills for AI today, because if AI is evolving and is joking at the skills that were required in the AI field a couple of years ago, they're very different from those that are required today. So it's really the real-time nature of that type of data is very important. It's also looking at news and at trends and at forecasts. So forecasts, a shoot, one, two, three years down the line, where can we expect the needs to be and sometimes, well, 10 years down the line, which is harder, but it's a bit, yeah. But critically important, I can absolutely-- Very important, yeah. It also looks at international collaboration and international agreements, which is an important component to have interoperable frameworks that can work together so that could just convince that some other countries are doing in specific sectors. For example, applications in agriculture that can be exported into other countries in a useful elsewhere. What's the biggest challenge right now in terms of international coordination or even cooperation on the kind of data sharing that you're talking about? I'll start by saying that international cooperation on AI policy is important because just like the internet, AI knows the borders and so what's developed in one jurisdiction will be available to maybe in some places to advance the users and others to everyone. And this is especially true since the generative AI boom of the late 2022, but like as a general purpose technology like electricity, AI is already or will very soon be critical to pretty much-- actually, it might even never say it is pretty much every sector from agriculture to health care in most countries. And the diffusion of AI is ramping up. The diffusion rates are ramping up, which has real impacts in a world where many firms do business internationally. And that's been the default for a long time. And that's something that's very hard to-- I mean, that's very lengthy to develop and to change as the case may be. So having global standards for trustworthy AI is critical to provide businesses with stability and predictability and to give users assurances that AI applications that they're using are trustworthy, robust, and secure. And in addition, many systems around the world are using similar core AI algorithms to train AI models. This means that many countries are actually vulnerable to the same AI-related risks, like security and issues or infringement of human rights. And so we need to take these some safeguards into international guidelines and standards. The challenges to achieving that are significant. Geopolitics, study, have become more complex than they were a while back, but not insurmountable. And at the heart of the work that we do at the OECD, things are taking longer. But the discussion is still taking place. It's not as rapid. And perhaps notice rapid as it could or should be. But it's taking place and it needs to continue to take place. And those challenges also stem from what can could be seen as strengths at the same time. For example, diversity among member countries. And particularly, we talked about the global partnership on AI, which has diverse levels of diverse countries in terms of size, levels of maturity, and so on. Like India has different priorities than some European countries do or the United States. And so, and different levels of experience, different histories, cultural sensitivities, and economic characteristics, economic industrial structures that were in which. AI is being embedded that lead to understandable differences in how they'll approach things like I or AI in tax or AI in education systems. But let me continue with the real opportunity here because absolute harmonization between countries is impossible and it's not necessary the most of the time. I mean, on most issues, it's impossible with different backgrounds, but that doesn't mean that approaches should be disjointed, patchwork and work against each other. And that's, this means that the timing now, especially as geopolitics are becoming more sensitive, the timing is increasingly critical. And that AI laws, regulations and standards currently being implemented or in varying levels of development are critical for the future. So the opportunity for further alignment is what we strive for in our work at the OECD from the work I mentioned on creating a common definition of what we're talking about or providing a definition for AI incidents so that we can accurately report and compare trends across country and avoid recurrence of issues. The OECD also has a role with the G7's Hiroshima process, which has a voluntary private sector reporting component to it. Can you talk about how that fits in with the other work that you're doing? The Hiroshima AI process was begun three years ago under the Japanese presidency of the G7 and developed code of conduct for our responsible AI. And the following year, last year, 2024, the Italian G7 presidency followed up on that work by developing a reporting framework, well, by asking, tasking actually the OECD to develop a reporting framework for companies to explain how their implementing trust with the AI, what measures they're taking at different stages of the AI system life cycle. And so we've been working with big, small, largely, mostly big because the work originally targeted providers of advanced AI, I mean, developers of advanced AI systems. And so we've been working with those companies to develop voluntary reporting, transparency measures that so has proven very helpful to governments to understand what companies are doing and they're doing a lot. They're doing a lot in all of the areas where they should. And there also, it's also helping the companies in terms, because large, very large companies like Microsoft, which is doing a tremendous amount, it is so big that this kind of having one to one reporting framework and forces coordination and internal coordination, which you're always running and there are always priorities between that coordination helps everyone to understand how they fit into the bigger picture. And so all of this, the framework was launched and I said earlier in this year in Paris at the AI summit, it's 20 companies reported, but many more are interested. And also not just developers of advanced AI systems, but also employers and also analysis of different types of organizations and of different sizes. So we have SMEs versus very large companies versus media scale ups. And then we also have companies that are operating in narrow and that are operating in business to consumer B2C and companies that are operating like Salesforce operating in mostly B2B and they'll have different practices will rely also on the business models of those different companies. But we have a task force that is exchangeable practices on what they're doing on different topics and what works less well. And that's been very helpful. These results, we are facilitating transparency and compatibility of risk mitigation measures and contribute to identifying and disseminating good practices. It's now broader than a G7 initiative in terms of the companies reporting. And a number of countries are reporting and more countries wish to particularly countries from the global partnership on the AI, like India, which has many, many employers of advanced systems and a lot of developments like in emphasis and others. And so we see this as a bit of a bridging the gap while there's some regulatory movement and uncertainty. This is really what the companies are doing and they're keen to share that because they know that the businesses do rely on trust and they are doing much, very much on all fronts from security to presenting other types of farms and testing and so on. So although all the results are public, I mean, everything is publicly available on OECD.DI of course, needless to say, I'm not advertising. I'm just saying that anybody interested in. You're not saying anything, it's all for you. Anybody who's interested can see the detailed reports of the companies that have submitted. We've drafted a summary that should be online as well. Going forward, are there any particular, either technological or market developments that you're looking at that might shift the environment for AI governance? Yeah, there are some significant shifts where I think we're aware of what is aware of on how some countries approach AI regulation and GAI regulation. Some as an enabler of trust and some as an impediment as an impediment to competition. How does the OECD address issues when its members, it may change due to elections, new administrations that happens all the time. So that's not, there's nothing special about it. We're a consensus driven body. So policy shifts after elections are normal and expected and our job is to create continuity by convening countries around the table and anchoring discussions. That shared evidence that I was mentioning earlier. The long-term values that these countries cherish and peer-reviewed policy tools. So that enables the OECD and GPs more and more policy advice to stand the test of time even when the political wins our due shift. And they due shift often. Indeed. Anything else that's on your agenda for the coming year? For 2025, I think maybe two lenses at an operational level with the new integrated partnership with GPE. We're working hard to implement the annual work plan and to strengthen the merger with new countries, work with these new countries that we hadn't worked so closely with before. Like in San Agal, in Africa, Brazil, and others. At a policy level, hand in hand with the six founded membership thanks to GPE and in view of through their expansion because we have more countries which we hope will join soon that are doing incredible things in AI. So we're focusing on building tools to help countries at varying levels of AI development to implement the AI principles. They also have, but on a localized, in a localized manner. And so that's not just here as a guide for a culture developing. No, it's a lot of creation workshops. So we've been doing a lot of work with the African Union Commission with the, with ASEAN countries, with Latin America. And so that's, I think, an important direction that we're taking. We're particularly developing an AI policy toolkit which countries can use to help them set up national frameworks, programs and tools to guide national AI policy policy. That's aligned with the AI principles. We have numerous other work teams that are also priorities, but just mentioned this one. And the listeners can find a worth. They're just absolutely. That's, it's a lot, but it's great to hear about all the different work that you're doing. Kerry, thanks so much for being with me. Thank you. This has been the Road to Accountable AI. If you're like what you're hearing, please give us a good review and check out my substack for more insights on AI accountability. Thank you for listening. This is Kevin Wuerback. If you want to go deeper on AI governance, trust and responsibility with me and other distinguished faculty of the world's top business school, find out for the next cohort of Orton's strategies for a countable AI online executive education program. Featuring live interaction with faculty, expert interviews and custom designed asynchronous content. Join fellow business leaders to learn valuable skills you can put to work in your organization. Visit executive.orgon.upand.edu/acai for full details. I hope to see you there. (gentle music)