[Panel Discussion] Global AI Policy Coordination

3 Oct 2025 17:00h - 17:40h

Session at a glance

Summary

This panel discussion focused on developing a global AI policy framework through international cooperation and historical perspectives on technology governance. The conversation began with Avinash Dadhich emphasizing that globalization and interconnectedness are not new phenomena, drawing from ancient Sanskrit wisdom about universal human connection. Jovan Kurbalija reinforced this historical perspective, arguing that many current AI governance challenges can be addressed using existing legal principles rather than creating entirely new regulatory frameworks.

The panelists explored Europe’s proactive role in establishing human-centric policies, with comparisons drawn to historical precedents like the abolition of slavery. Frederic Werner from ITU highlighted numerous AI-for-good applications while identifying key bottlenecks including data sharing challenges, global connectivity gaps, and governance coordination issues. Alexandra Baumann presented Switzerland’s three-pillar approach emphasizing inclusive multilateral cooperation, bridging global norms with local realities, and strengthening Geneva as a digital governance hub.

A significant debate emerged around Global South representation, with Kalika Bali arguing for greater inclusion of underrepresented voices in AI policy discussions, while Avinash Dadhich suggested moving beyond North-South divisions toward a more unified human-centric approach. The discussion addressed intellectual property challenges in AI development, with Hu Ling examining the tension between innovation needs and copyright protection. Lucia Velasco provided updates on UN initiatives, including the newly launched scientific panel and global dialogue mechanisms for AI governance.

Alexandre Mirlesse concluded by presenting the concept of “open sovereignty” as a third way between competing AI models, emphasizing that countries want to deploy AI beneficially while maintaining decision-making autonomy. The overall consensus pointed toward the need for inclusive, multilateral cooperation that respects cultural differences while establishing shared human-centered principles for AI governance.

Keypoints

Major Discussion Points:

– Historical Context and Continuity of Global Cooperation: The discussion emphasized that globalization and international cooperation are not new phenomena, drawing parallels between ancient wisdom (like the Sanskrit concept “Vasudeva Kutumbakam” – the world is one family) and modern AI governance challenges. Speakers argued that many current regulatory principles have historical precedents and that looking backward helps us move forward.

– Global South Representation and Inclusive AI Governance: A significant focus was placed on the underrepresentation of Global South voices in AI policy discussions, despite these regions having the most to gain from AI applications. The conversation highlighted cultural differences in concepts like privacy, the need for diverse perspectives in policy-making, and the importance of including traditionally underrepresented countries in tech governance processes.

– Practical Barriers to AI Implementation: Several concrete obstacles were identified that prevent AI for good use cases from scaling globally, including data access and sharing challenges, connectivity issues (with one-third of the world still unconnected), skills gaps, and the need for content in local languages and dialects.

– Sovereignty vs. Openness in AI Development: The concept of “open sovereignty” emerged as a key theme – the idea that countries can maintain control over their AI development and deployment while remaining open to international cooperation. This was presented as a “third way” between complete dependence on foreign technology and isolationist approaches.

– Role of International Organizations and Frameworks: Discussion covered the evolving roles of various international bodies (UN, ITU, Council of Europe) in AI governance, with emphasis on Geneva as a neutral hub for digital governance and the recent launch of UN mechanisms for AI governance including a scientific panel and global dialogue platform.

Overall Purpose:

The discussion aimed to explore pathways for effective global AI policy coordination, drawing lessons from history and diplomacy while addressing current challenges in creating inclusive, human-centered AI governance frameworks that serve all nations and populations.

Overall Tone:

The discussion maintained a constructive and optimistic tone throughout, despite acknowledging significant challenges. Speakers were collaborative and solution-oriented, with occasional moments of friendly debate (particularly between the Indian speakers about Global South/North classifications). The conversation was academic yet practical, balancing philosophical concepts with concrete policy recommendations. The tone remained respectful and forward-looking, emphasizing cooperation over competition in AI governance.

Speakers

Speakers from the provided list:

– Avinash Dadhich – Panel moderator/host, lawyer (mentions being anti-economist and a lawyer)

– Jovan Kurbalija – Historian, author of a book on history of law and technology and diplomacy

– Frederic Werner – Representative from ITU (International Telecommunication Union), works on AI for Good initiatives

– Alexandra Baumann – Swiss diplomat, represents Swiss perspective on global AI policy coordination

– Kalika Bali – Research scientist at Microsoft (clarifies she’s not representing Microsoft as employer), works on low-resource communities and languages, responsible AI and safety, advisor to Indian government’s safe and trusted AI pillar within India AI mission

– Hu Ling – Professor, Chinese academic focusing on IP issues and AI policy

– Lucia Velasco – Works at United Nations Office for Digital and Emerging Technologies, leads AI policy at UN, describes herself as “recovering economist”

– Alexandre Mirlesse – Career diplomat, former advisor to French president for the organization of 2025 AI Action Summit in Paris

Additional speakers:

None – all speakers who participated in the discussion were included in the provided speakers names list.

Full session report

Global AI Policy Framework: International Cooperation and Historical Perspectives

Executive Summary

This panel discussion brought together international experts to explore pathways for developing effective global AI policy frameworks through multilateral cooperation and historical analysis. Moderated by Avinash Dadhich, the conversation featured perspectives from diplomatic, academic, and international organisation representatives, examining how historical precedents can inform contemporary AI governance challenges whilst addressing critical issues of representation, sovereignty, and practical implementation barriers.

The discussion revealed areas of agreement on fundamental principles including the need for inclusive governance and building upon existing institutions, while also highlighting different approaches to regulation, representation, and the balance between sovereignty and cooperation.

Opening Context: Historical Foundations of Global Cooperation

The discussion commenced with Avinash Dadhich establishing a historical perspective, emphasising that globalisation and interconnectedness are not modern phenomena. Drawing from ancient Sanskrit wisdom, particularly the concept “Vasudeva Kutumbakam” (the world is one family), Dadhich argued that current AI governance challenges should be viewed through the lens of millennia-old patterns of human cooperation through trade, art, and cultural exchange.

Jovan Kurbalija reinforced this historical framing with concrete examples, highlighting what he called the “original sin” of 1996 when technology companies were exempted from content responsibility, creating many current governance challenges. His observation that “we didn’t need horse law when horses were introduced” challenged the assumption that AI requires entirely new regulatory frameworks, arguing instead that existing legal structures can address most AI issues with only targeted interventions needed for cybercrime and data protection. Kurbalija also made a lighthearted reference to Indians inventing zero and the digital world, connecting historical innovation to contemporary challenges.

Practical Implementation Challenges

Frederic Werner from the International Telecommunication Union identified concrete obstacles preventing AI for good use cases from scaling globally. He described the current situation using the analogy of a “V8 engine firing on five of eight cylinders” regarding global connectivity, noting that whilst organisations possess valuable data, privacy concerns create reluctance to share information necessary for beneficial AI applications.

Werner highlighted that connectivity challenges extend beyond infrastructure availability – many regions have technical coverage but lack incentives to connect due to the absence of content in local languages and dialects. He also identified universal challenges including the AI skills gap and combating misinformation and deepfakes.

Werner pointed to concrete initiatives addressing these challenges, including the AI4Good summit, Innovation Factory, machine learning challenges, Robotics4Good, the launch of an AI skills coalition, and collaboration on AI watermarking for multimedia content authenticity.

Switzerland’s Multilateral Approach

Alexandra Baumann presented Switzerland’s strategy for AI governance, outlining three key points. First, inclusive multilateral cooperation involving all stakeholders across regions, with particular attention to underrepresented countries. Second, bridging global principles with local realities, supporting regulatory interoperability whilst respecting national contexts. Third, strengthening Geneva as a neutral hub for digital governance, building upon existing institutions and the AI for Good initiative.

Baumann referenced Nina Frey and ICAIN (International Computation and AI Network) as part of these efforts, and mentioned the World Summit on Information Society process for Global Digital Compact implementation. She explicitly disagreed with “the colleague from the States” (referencing Dean Ball, not quoted in the transcript) who had questioned the value of standards and norms, arguing strongly for their necessity in creating coherent international AI governance frameworks.

Global South Representation and Power Dynamics

Kalika Bali, speaking in her personal capacity while noting her role as advisor to India AI mission’s “safe and trusted AI pillar,” presented arguments about Global South representation in AI policy discussions. She highlighted that these regions face unique challenges, including being primarily oral rather than written digital cultures, which challenges Western-centric assumptions about data and digital engagement.

Bali’s observation that “most of the Global South is not a written digital culture” exposed potential blind spots in current AI governance frameworks. She argued that privacy concepts are culturally specific, providing an example about marriage and bedroom privacy to illustrate how cultural contexts shape privacy understanding, and that global regulations must understand these differences rather than imposing uniform standards.

Bali challenged traditional terminology, stating: “Let’s not call it Global South and Global North, let’s call it Powerful and Not-so-Powerful,” reframing the discussion around power dynamics rather than geographical classifications. She also suggested that India has potential to bring Global South countries together due to similar problems and successful policy methodologies.

Avinash Dadhich questioned this framing, arguing that the traditional Global North/South divide is losing relevance and advocating for a more unified human-centric approach, highlighting different perspectives between the two Indian speakers on how to address representation issues.

Intellectual Property and Innovation Challenges

The discussion addressed complex intellectual property issues arising from AI development. Dadhich noted that AI development creates IP conflicts as algorithms use global data, potentially violating copyrights whilst requiring extensive data for innovation.

Hu Ling provided perspective on this challenge, referencing the “illegal rise of the internet” as a historical precedent where initially “illegal” content sharing eventually proved valuable to society. Hu Ling suggested that AI represents a potential new mode of social production, but emphasised that AI companies must prove they add genuine value rather than simply redistributing existing value to justify their use of copyrighted materials.

UN AI Governance Initiatives

Lucia Velasco provided updates on United Nations initiatives for AI governance, explaining the UN’s role as a neutral universal convener. She outlined the timeline: a 2023 advisory body, the 2024 Global Digital Compact, and a September 2025 launch of two mechanisms – a scientific panel seeking 40 experts and a global dialogue platform, both focusing on non-military AI domains.

The scientific panel will provide evidence-based policy assessments, whilst the global dialogue will enable multilateral discussions. Applications for the scientific panel were open until October 31st, with emphasis on achieving geographic and disciplinary diversity. The first full UN global dialogue meeting is scheduled for July 2026 in Geneva, coinciding with the ITU-AI for Good Summit.

Velasco emphasised that the UN’s value lies in providing legitimacy, inclusivity, and integration, with a key priority being building capacity so developing countries can meaningfully participate in AI governance discussions.

Open Sovereignty: A Third Way Forward

Alexandre Mirlesse introduced the concept of “open sovereignty” as an alternative to binary choices in AI governance. This approach combines national independence with international openness, offering countries a path to deploy AI beneficially whilst maintaining decision-making autonomy.

Mirlesse argued that AI governance should move beyond traditional divisions toward this third way that acknowledges both sovereignty needs and cooperation benefits. He provided concrete examples, including the French unemployment agency experimenting with AI and mentioned Mistral as a French AI player developing alternatives to dominant platforms.

The concept includes practical elements such as cloud and data standards that guarantee interoperability and reversibility, preventing technological lock-in. Mirlesse advocated for mixed AI ecosystems combining open and commercial models, large and small systems, and multilingual and specialised applications.

His concluding statement framed the fundamental choice: “Either we want AI to be a marketplace of dependence, you choose your boss and you go with that technology, or it can be an infrastructure of freedom.”

Key Themes and Different Approaches

The discussion revealed both shared concerns and different approaches to addressing them. Speakers generally agreed on the importance of inclusive representation and building upon existing institutions, though they differed on specific implementation strategies.

There were notable differences in regulatory philosophy, with Kurbalija advocating for using existing legal frameworks with minimal new regulation, while Baumann emphasised the need for new multilateral approaches specifically designed for AI governance. The exchange between Dadhich and Bali highlighted different perspectives on how to frame and address power imbalances in global AI governance.

Data sharing challenges, connectivity gaps, and the need for cultural sensitivity in AI governance emerged as common concerns across speakers, though they approached solutions from different institutional and philosophical perspectives.

Conclusion

The discussion demonstrated both the complexity of global AI governance challenges and potential pathways for international cooperation. The combination of historical perspective, practical experience, and diverse viewpoints created a foundation for understanding both opportunities and obstacles in developing effective AI governance frameworks.

The concept of open sovereignty, emphasis on building upon existing institutions while ensuring inclusive representation, and recognition of cultural differences in approaching AI governance provide concrete directions for continued work. The ongoing UN initiatives and other international efforts offer mechanisms for advancing these discussions toward practical implementation.

Session transcript

Avinash Dadhich

Thank you very much. I will start with a small poem. The poem says like this.

Oh man, he is my family, he is my relative, the others are my strangers. Oh man, don’t think like that, that he is your family, the others are strangers. The entire world is your family.

So in Sanskrit, it’s called Vasudeva Kutumbakam. So this line or this poem was written 3000 years back in one of the Vedas called Atharveda. And it’s not Indian or ancient wisdom.

I think this wisdom was in all civilizations. Wherever you go, you will find this thing. Historically, we were all connected.

I think now for some people it looks like the globalization is a new thing, but it’s not new. I think through the trade, art, music, literature, Chinese, Japanese, Indian, Middle East, European, I don’t know what the American civilization by that time, but even the South American civilizations, they were all interconnected.

They learned from each other. So definitely globalization is not a new thing. This is happening again and again in different waves.

So I will start now this panel. I know this is quite a heavy topic, so we’ll start with very historically. I think history is sitting with me.

He just wrote a book, very nice book on the history of law and technology and the diplomacy. So diplomacy is again a very ancient art and technology is again very ancient thing and I believe that humans are victim of repetitiveness. We always repeat things, but then our memories are very short, so we forget and then we reinvent, but most of the answers always lies in history.

So I would like to ask you that what do you think about, what we can learn about this global AI policy framework from the history of diplomacy and technology?

Jovan Kurbalija

Thank you, Avinash, and for a great quote from the Sanskrit and I will, I always tease my Indian friends that the world was so simple before you invented Shunya, number zero, and now I invite all of my Indian friends to fix the world because we had Roman numbers and it was so easy.

And then you have zero and one, the digital world came and now we have all of these things, NIA and these things. What, there is a famous quote by Churchill, the further forward you would like to… See the further backward you need to look.

And there is a paradoxical insight here, that things are not as new as we think. Obviously, we live now and we have a bit of chrono-Narcissism, we are excited, everything is happening now, and these things. But when you step back, you see, for example, that the simple law, rule of Hammurabi’s law, that if builder build a house, and if that house collapse, the builder is responsible.

And that principle was basically carried through Roman law till 1996, where we were told that tech companies are not responsible for the content and what they’re doing. And I think this is original sin that we are currently facing. We are discussing law, regulation, governance, and these principles are rather old and simple, and I would say some sort of AI common sense that we should use more and more in order to use the power of technology, which is enormous.

But the first step is to step back and say, hey, is it that new? And I will conclude with one point. In 1996, again, this critical year, one professor at the Chicago University asked a question when he was asked, do we need internet law?

He asked a simple question, did we need horse law? We didn’t need, horses were introduced, and if you really think about what’s happened over the last three decades, you have only three intensive regulatory interventions on cybercrime and on data. Everything else, you basically can apply existing law, which I, by the way, congratulate Thomas for slowness in reacting, and I think there is sometimes to be slow can pay back, especially with AI, what we also heard from a colleague from the States.

Avinash Dadhich

No, I think it’s a beautiful and especially this, I think Thomas, I really love your point that weight was slow and I think in India we are also doing it, but I think Europe is proactive and Europe is not proactive first time.

Europe is proactive since maybe 16, 17 or maybe before that, who promoted all these liberal democratic ideas, equality, fraternity, liberty, democracy. I know democracy is not a new idea, but I think after thousands of years, this idea was reinvented in Europe. When in 1833, when the UK government decided to abolish slavery, I know it looks very funny now that, okay, is it possible to exist with slavery?

But during that time, it was a huge business. So the UK decided that, okay, we don’t want to continue with slavery and they made a law and not only they made the law, but they ensured that this trade should not happen in other parts of the world. They try to convince France and other countries that you should also have some rules, regulations.

France made a law after 50 years and slowly and slowly you can see within 100 years, most of the countries accepted that slavery is a bad thing. So it’s not a new thing that Europe did something crazy in terms of maybe other countries. Europe is crazy, but we need craziness to bring some innovative human-centric ideas.

The ideas which serves the human civilization, lot of economic rationality which is brought by the economist. I’m sorry, I’m a lawyer, so I’m very anti-economist, but in economics, I’m sorry, I’m making few enemies now, but you are a professor of law and economics, so you are 50-50. So in economics, you can justify anything.

But when we talk about life and society, there are a few things which we cannot compromise and definitely a lot of good things are happening. So coming back to the good things, my question is to my friend from ITU, Fred, a lot of good things are happening, okay? We are so pessimist now, we think that nothing is happening.

This global collaboration is not happening, coordination is not happening. But if you observe, that is happening right now in ETH, like your conference, OECD, UNESCO, Council of Europe, across the globe, a lot of good things are happening. Why these good things are not becoming the discussion point in general public, or why they are not becoming more global?

Frederic Werner

Yeah, thank you, that’s a very good question. Probably one of the most interesting and fun parts of my job, I mean, it is called AI for good after all, is all of the AI for good use cases that come across our desk every day. And this has been happening since 2017, when most of the use cases existed in PowerPoint, and they didn’t really exist anymore.

But we’re seeing a lot of those use cases materialize before our eyes now, with the rise of generative AI, now agentic AI, some of the executives at our summit this summer saying AGI is just around the corner.

And how do you know all of these high potential applications, for example, in healthcare, work equally well on men or women, or on children or the elderly, persons with different skin colors, persons with disabilities, or in low resource settings where things we take for granted like 5G, Wi-Fi, access to data, electricity, cloud, these are things we take for granted, but they’re real concerns in many parts of the world.

So these are not natural reflexes for the fast moving tech and startup industry. I think the track record so far has been let’s build it and we’ll fix it later or put a bandaid on later. But these are things that we think about deeply at AI4Good.

And I think that’s been foundational to the summit and all the activities that’s been happening. And so I think there’s no shortage of use cases. I’ve seen hundreds of them, if not thousands of them.

They come through our different programs like the Innovation Factory, our machine learning challenges, Robotics4Good for example. But there are quite a few bottlenecks that are preventing these from scaling. I mean, one is definitely data.

Access to data. I can’t tell you how many times I’ve been in a room and I’ll ask for a show of hands, like who has really interesting data? All the hands go up.

And then I say, who’s willing to share the data? People look at the shoes. This is not happening.

So until we figure out how to share data in a way that’s useful, but still respects privacy, and there are techniques for that, that’s a major bottleneck. Another bottleneck, and of course, being the ITU is connectivity. Right now, a third of the world still remains unconnected.

And for me, that’s the equivalent of a V8 engine that’s only firing on five of the eight cylinders. And we’re not basically benefiting from the collective problem-solving creativity of all these people, of their culture, of their storytelling. And what’s interesting is many parts of the world actually have coverage, but they’re not able or not willing, or there’s no incentive for them to connect because there’s no content in their local language and their dialect.

Some people are illiterate or they have disabilities. And I like to think of the potential of AI to break down those barriers. Now you have speech to text, text to voice, where a lot of these barriers in using AI for interpretation and translation, what that hopefully could reduce these barriers.

And then you don’t have connectivity for the sake of connectivity, but actually connectivity to have all those cylinders firing on all pistons. And perhaps the last piece of the puzzle that’s definitely a bottleneck would be more on the governance of a policy. part- you know we’ve seen in the past- two governance days that we did at the summit- you know we make great efforts to bring all the major regulatory and governance processes from around the world to the summit.

From from U. S. from EU from UK from Japan from Korea from China.

And what we found was that of course there are philosophical differences among these processes. But the goal was also to see what we have more in common than we have different. And two major pain points I think that everyone agreed on without question.

What were basically the- how do we address the AI skills gap- so because of that we launched the AI skills coalition. And the other part of the picture was how do we address the epidemic of misinformation and deep fakes. And because of that we launched a collaboration on AI watermarking and multimedia content authenticity.

That’s a fancy way for saying deep fake detection techniques with standards. And as a result of that we actually created a toolkit. Which could help member states if you’re thinking about how do I address misinformation or how do I even build a strategy around addressing that.

There’s actually a checklist all on how you could go back so there’s a technical part to it and also a policy part of that. So I’m not sure if I answered your question but again this does the optimistic side of all the use cases of their- but unless if we address all those issues versus gonna remain use cases and Shenzhen or in Silicon Valley or in Brussels or in some nice university.

Like here- so to make those use cases really go through the global south. I think we need to solve many pieces of puzzle the data piece connectivity piece. I think the governance slash standardization piece and there’s actually a long list of other problems but I don’t want to be too much of a downer so.

Those are the focus of myself and ITU for now thank you.

Avinash Dadhich

No I think very good job. And I think we should not think like that it’s not no use you know. It has some value.

A lot of good things are happening. We might not seeing any impact right now maybe due to some temporary political leadership and some situations you know. But all good things, they are creating energy and this energy is invisible, but it’s connected.

Okay, like whatever good, like in this room a lot of good things are happening, a lot of positive vibes are moving around. So this energy makes a sense, but in Indian astrology we say that some lands are very, you know, privileged and loved by the God or by the supernatural power, whatever you say, that they have the capabilities or an opportunity to create a lot of peace.

And I think Switzerland got that opportunity in the last, I think, 100 years. So this is the place where the United Nations and all the good people are living here, last 60-70 years. And maybe a few years back we lost our faith, not completely, but especially in the Global South we started talking about the UN is not doing anything.

Because when some organization is doing so many things, especially good things, then you start losing the relevance. I think an economist can explain you the marginal values goes down. But when nothing good is happening at the top level, then again you start saying, oh, okay, UN, can you do something?

So from, you’re a career diplomat and I want to ask you that a lot of good things are happening. We are talking about coordination. Can you share some of your thoughts about the cooperation?

How can we cooperate with each other and this entire energy ecosystem can convert into some action?

Alexandra Baumann

Thank you very much, Avinash. So from a Swiss perspective, there’s three key points that we believe that you mentioned at global coordination and that they are crucial for that. First it’s an inclusive and human-centered approach to put the human at the center of AI, which is multilateral and multistakeholder.

So global coordination will always require an inclusive participation from all stakeholders across all regions. especially from those from countries traditionally underrepresented when it comes to tech governance processes. We therefore support initiatives such as the International Computation and AI Network, ICAIN.

I’m very happy I see my colleague Nina Frey, who used to be, or is still a diplomat, but we sent her to ETH to actually promote ICAIN for the next two years and to put strong in it. Which aims to promote equitable distribution of technological benefits, because we still believe that that’s crucial if we want to work together. We also strongly support Switzerland, the UN’s role as a leader and facilitator in this, and we support the political goals expressed in the Global Digital Compact.

And for the implementation, I think we should rely on existing structures, because there’s lots of talk of creating these new mechanisms, but we have the World Summit on Information Society process. We had it for the last 20 years, so let’s just use it and implement GDC within what we have. So AI policy, in our view, must be inclusive, development-oriented and people-centered, upholding human rights, democratic values and international law.

And these are principles that were established with VCs 20 years ago. And for us, these are non-negotiable foundations for a trustworthy AI. Second, we need these global norms to actually, we need to bridge them with the local realities.

So global principles are very important, but implementation must account for national contexts and capacities, as you were mentioning. There are still big gaps there. So we advocate for mechanisms that support regulatory interoperability, where countries can align on shared norms and…

You know, contrary to the previous speaker from the United States, we strongly believe in standard and norms. So even if local legal systems are not the same, so they need to adapt for this. And for this, we see value in regional and cross-sector dialogues, such as a dialogue between Europe, the Global South and Asia, to actually also build trust and avoid the fragmentation that it could lead to.

And we very much welcome, and Thomas was mentioning it before, the Council of Europe’s Convention on AI and its work that it includes on human rights, democracy and the rule of law. Because these instruments, they establish globally shared principles while leaving the necessary flexibility to implement them according to national context. And third of all, and that’s very dear to my heart, is strengthening Geneva’s role as a hub for digital governance.

And therefore, I’m very happy to sit next to you because we very strongly believe that we want to highlight the AI for Good initiative, which brings together over 40 institutions from Geneva and the world in a multidisciplinary approach.

And I think this is the way to go ahead. And so we are actively supporting the International Geneva as a neutral and inclusive platform because we have a lot of, you know, everybody’s there. So let’s use it and see what we can build on the ecosystem that’s already existing.

So, you know, in conclusion, we see global AI policy coordination, not as a competition, but as a shared responsibility of all stakeholders to actually ensure that artificial intelligence serves human dignity, global stability and sustainable development.

So these are the three principles that I would highlight.

Avinash Dadhich

No, beautiful principles and I think these principles are already pre-loaded in our human consciousness. I believe that our brain or our mind or whatever you say it, is pre-loaded with all good and bad ideas. So whenever time comes, as per the situation, those ideas, they start uploading, unloading, downloading.

So it’s not so difficult for the human brain to be brainwashed because those ideas are already there. But I appreciate the presence of this global Geneva concept because I think this is very much required in the coming years. Maybe a few years back we were not very comfortable with this idea, especially in the Global South, that why things are happening only in Geneva.

But now I think even the Global North, they are also saying, you know, I think Geneva should be there. We need someone, somewhere at least people can go and talk. And the local people should not have any huge vested interest because they have good patience.

So now I go to Calica. See, I believe that problem and solutions, they live together. You know, it’s all that timing issue or the perspective issue, whether you see a problem or solution.

So this problem of AI is coming mostly from the big tech. And I’m so good to see here, at least, you know, you are here and you are facing the audience and you are involving in the process of solving this problem. OK, this is like an allegation that all big tech companies are doing all bad shit, but companies like Microsoft, they are sending their representatives.

I think this is a good initiative that, you know, going out and facing people. So what do you think about this idea of how can we create a framework? Because you are coming from private sector, but I must tell you one thing.

She told me in private that she’s not involved anything with Microsoft is doing. in terms of product development. She’s a scientist, and she’s doing some good things with Microsoft.

I think that’s why maybe they sent you here. They want to send good people, okay. Good.

So, the rainbows, they keep it somewhere else. They send all good people. That’s good.

I think we need good people in this room. Yeah. So, as a Microsoft, or as a scientist, you know, what do you think that, what should be the roadmap to have a framework for the techno-legal and creating a balance between innovation and our democratic ideas?

Kalika Bali

So, first of all, I’m not here as a representative of my employer. I am a research scientist who spent like over 20 years working with low-resource communities, low-resource languages, and have recently started looking at responsible AI and safety and trust. And I’m actually spending half my working life with the Indian government as an advisor to the safe and trusted AI pillar within the India AI mission.

So, if anything, I’m a representative of the global South, if you can say that. And from that perspective, I really appreciate what, actually, I’m going to be a judge in one of the innovation things just next week. Next week, I will be there to judge one of these.

But I really appreciate you bringing that thing, and I would go towards that a little bit more. I think it’s imperative that in all these kinds of discussions that we have, we have a larger representation from the global South or global majority, whatever you want to call it, because their voice is completely missing from the table.

Whenever we, I’ve attended a bunch of summits, workshops, all these things earlier, and I find that there’s very little voice of the global South on the policy side of things. Use cases, tons of them, right? People just think that, oh, it’s a data problem.

It is a data problem. Most of the Global South is not a written digital culture, right? Of course, it’s a data problem.

If we are talking about oral culture, it wouldn’t be a data problem because it has primarily historically been an oral culture and not a written culture, right? There are other issues also. Most of the Global South has been historically resource poor.

So the infrastructure is missing, right? Now, if you’re talking about policies related to compute, you’re talking about policies related to data, right? So if you don’t bring them and take their perspective, you’re actually really only talking about Europe and US, not about anybody else.

Because if you think about it, you have all these laws, all these regulations, all these liabilities related to data. And you’re talking to a country or a region, set of countries, where they actually don’t have data, right? So now, the option is, like, here are all these.

How do you balance the fact that here are all these regulations related to data, right? And so do we not collect any data? Do we collect data?

What do we do about data? So if we are already short of data, so then what do we do? We stay behind in building technology?

Does this law kind of restrict us in some way? You have to hear those things. You have to have those voices on the table, right?

And there are so many such issues. You talk about privacy, right? The notion of privacy is so dependent on the culture of the place you are, right?

I got married. And my mother had no problem, you know, newly married people walking into my room to pick up something from the room while we were sleeping on the bed. Okay.

So privacy is a very, very culturally specific notion. What is private is very culturally specific. Have those discussions.

I’m not saying that we should just like open up everything and say, okay, here it is. But at least I try to understand those differences and how we can reconciliate them. Like, you know, how can we bring the whole world together as Vasudeva, Kutumbakam.

But so there’s that notion as well. And then I really think that, you know, I’m an Indian. I’m not saying this because I’m an Indian, but I really do think that India has a potential to kind of bring the global south together in some way.

Because our problems are similar, we have done things in a way that actually can address some of those problems. We’ve come up with certain policies, we’ve come up with certain methodologies of doing things. And therefore, I really think like a network of the global south, AI network for the global south is one of the things that would, I would really wish to see happen.

Avinash Dadhich

So, which I think, I’m so sorry because we are both Indian. So we are okay with this debate and discussion and disagreements because Shastrath is like a debate is historical Indian concept where two scholars used to argue with each other for hours. And until and unless the other scholar says, okay, now end it, it will not ever end.

But I think this is a global forum and I would like to talk about this classical debate which started in 1960s and 70s about. Global South, Global North. I think it’s losing its meaning now.

What is Global North? Because Global North, before I think last one hour, we’ve seen the very beautiful unity of Global North. You know, Europe and US, very classical unity and very beautiful understanding of AI regulation.

So it’s not anymore Global North. When we say Global South, I think China, Japan and East Asia, lot of interesting countries are emerging like Singapore, India, Africa is very different. So I think this classical debate or the classification between Global South and North South is losing little bit meaning now in today’s world.

I think that’s my understanding. I think now we need to see as a human civilization. Obviously, cultures are very different.

Even within the European Union, when you visit Europe, you will find huge differences. But I think we need to have more human-centric approach. When I say human-centric, means that we are all humans.

Unfortunately or fortunately, I don’t know, but I read a very interesting fake news a few days back that some ship is coming from somewhere and they are carrying some aliens and they are going to attack humans.

I read that fake news many times. And to some extent, I started believing in it. And I said, wow, what if this news is right?

And I was happy. And you know why I was happy? And I said, yeah, this is the time now every country will start thinking as a human.

This will be the first time they start thinking, oh no, they are not going to attack on Global South or North South. They are going to attack on Earth. And as one civilization, we have to react.

The way we reacted in Corona. So, I think this AI is a huge opportunity for Global South and North South, whatever you say. But I believe that this is an opportunity for us to think like a human.

All this divide, because within the Global South you will find a lot of Global and Global and North and Global South.

Kalika Bali

I agree with you, Avinash, that we should be human, but where are the people or where is the representation?

Avinash Dadhich

No, representation takes time. I think we are here and more and more people will come. See, representation always comes with opportunities, power, power dynamics.

Kalika Bali

Exactly. So, let’s not call it Global South and Global North, let’s call it Powerful and Not-so-Powerful.

Avinash Dadhich

Even the North South you don’t get so many representation, even Canada is not here. See, I believe that, this is, we are okay with that, okay, after that you will, okay. So, but I believe that it’s a time to move forward from this Global South thing.

Coming to my neighbor, my Chinese friend, Professor Ling, you want to talk about something very interesting, the value of information and knowledge, because data is the oil, data is the key, and all these big algorithms are using data globally, and definitely they are violating copyrights.

So in that scenario, how to make a balance in the innovation, because they need data for the innovation, but at the same time they need to respect intellectual property created by some other people, they need to compensate those people.

So this debate is happening across the globe, I think a lot of courts are already filled with a lot of applications and cases. So what do you think, how to solve this problem at the global level, because this is not a national problem anymore, okay, when they are creating global algorithms, it’s an international problem. How to solve this problem?

Hu Ling

Thank you very much. I totally agree with you that IP issues is just not a national or just a domestic problem, it’s a universal problem for all the countries, but maybe it depends on the degree of social production. I think that the AI, as well as the internet, it just raised the issue of social production, that everybody can, with some smart tools that can contribute to it.

to commons, right? No matter whether those commons are regulated by a public sector or the private sectors. But I think the IP conflict issues arise just because, maybe think about 20 years ago when the internet arise and there is nothing on the internet, so that people don’t know how to use that and people still wish to go back to TV or newspapers or traditional books.

But nowadays we see there are tons of different contents and programs. So how does this happen? Maybe I used to do some research based on Chinese experience that called this era called the illegal rise of the internet.

Illegal, maybe you can judge it from a moral or some legal perspective. But I think this is from the development of society that I think this is a necessity for this stage to occur. Because if there is no lower cost of information online, maybe people wouldn’t love to go to the internet.

So that’s what we see 20 years before. And nowadays we see the AI, right? The AI, when we train AI, pre-trained AI with tons of data, but where those data come from?

Maybe there’s in China, in the United States, in the EU, there are so many lawsuits concerning the legal battles between the AI companies and the platforms and the traditional authors, singers, artists.

So I think that those conflicts continue to occur. But my opinion is that maybe when we talk about the law of the horse, right? Maybe how can we identify what kind of law AI companies or platforms or those companies prefer?

I’m not just standing on behalf of those businesses. Because if you claim that AI could be a kind of a new mode of production, you have to prove it, right? If you just at a lower cost, maybe with some robots to grab all those digital materials and print that.

But how can you prove it that this represents a kind of new mode of production? Otherwise, it’s just stealing or just move some values from here to there. But it won’t add more values.

So I think that today’s problem, as well as the IP policies, that how to facilitate those creation based on the IP materials with a lower cost. On the other hand, it has to move fast to prove it as maybe like the internet nowadays, right? The internet has proved it as a new mode of production.

But how about AI? I’m not sure about that. But maybe we can continue to observe.

Avinash Dadhich

Thank you, Professor Ling. So now I directly go to Lucia. Lucia, I think she’s joining online.

Given your role in leading AI policy at United Nations Office for Digital and Emerging Technologies, what are the AI priorities for the UN this year? And what is the role of UN place in the global governance?

Lucia Velasco

Yes, hi, everyone. I’m a recovering economist. I’m sorry, I’m not your favorite person.

But I’m happy to help with the seating arrangements and not having to seat with Fred since I see that the stage is really full of people already. It’s difficult to say something after Dean Ball and then the Vedas, but I’ll try to keep up with that level and share perhaps some news about what has been happening in New York this past week. The high-level week of the 80th UN General Assembly really marked an important milestone for AI governance.

And it built on two years of negotiations and processes. And we have a feeling at the United Nations of a certain sense of culmination of part of a journey. As many of you might remember in 2023, the Secretary General created his high level advisory body on AI who produced the report governing AI for humanity where they recommended new governance mechanisms among them, the scientific panel and the global dialogue on AI governance.

In 2024, the member states of the United Nations, the 193 member states adopted the Global Digital Compact at the summit of the future, providing the political mandate. And in September this year at UNGA 80, the same member states formally launched this mechanisms within the United Nations and limited to non-military domain of AI. So the scientific panel will bring together 40 experts serving in the personal capacity to produce annual evidence-based policy relevant but not policy prescriptive assessment on the opportunities, risks and impacts of AI.

The global dialogue on the other hand responds to some of the comments or the asks that have been made already. It’s a multilateral and multi-stakeholder compliment, a universal platform for open, transparent and inclusive discussions on AI governance, covering issues from safe and trustworthy AI systems. to bridging digital divides and ensuring interoperability of governance approaches, because I really liked the comment from Karika about the different understandings that we have of things in different parts of the world, and the privacy example was very telling.

The first full meeting will take place in July 2026 in Geneva alongside the ITU-AI for Good Global Summit, so Fred will be very busy I imagine. At the launch of the dialogue, which happened last week on September 25th, the Secretary General also opened the goal for a candidate to the scientific panel, which runs until October 31st, and the goal seeks scientists of outstanding expertise across core AI research, applied AI, as well as disciplines addressing impacts, ethics, rights and governance.

So please allow me to emphasize and to encourage applications from across all regions and to help us spread the word, because we need to ensure that the panel reflects true global balance in perspectives and expertise, and together these two mechanisms are designed to work together.

The panel provides that independent scientific evidence base, and the dialogue would enable political and multi-stakeholder deliberation. So those would be together, you know, obviously setting up the panel effectively, making the global dialogue a truly inclusive forum, building on the fantastic work that the ITU has been doing. Together with building capacity and reducing the existing divide so developing countries can also meaningful participate in AI governance and not be left behind would be our priorities for the upcoming year.

Obviously the UN’s role and I’m going to say this because I imagine I will not have time to speak later but I want to make this point really really clear the role of the UN in global governance is not to regulate AI and there’s a lot of confusion about governance of AI and regulation of AI.

I think the UN’s role right now is becoming that neutral universal convener that universal platform that provides legitimacy inclusivity and integration making sure that AI governance approaches are not fragmented or driven only by a few actors but actually serves the interests of all.

Thank you I’ll stop here.

Avinash Dadhich

Thank you Lucia. Thanks for your updates. I think it’s an exciting time for United Nations.

So our last panelist is Alexandre. He’s a career diplomat and former advisor to the French president for the organization of 2025 AI Action Summit in Paris. So why I’m telling you because my question is related to this.

What are the most important learning from the Paris AI Action Summit 2025 to genuinely engage to diverse global voices and ensure dialogue leads to tangible outcomes to impact which is the theme of New Delhi AI Impact Summit 2026.

Alexandre Mirlesse

Thank you very much Avinash and to honor the tradition of Swiss punctuality and so much great stuff having already been said I limit myself to one lesson from the summit. And this lesson is the following, I think too much, except on this panel, which I really like, the AI global policy debate is framed in terms of dual and divide. It’s the American model versus the Chinese model, it’s the West against the rest, it’s the global South versus what remains of the global North, etc.

I think what we did, and I say we because the summit happened in Paris, but it was co-chaired by France and India, and now India will host the next summit, so it’s a close cooperation between a country from the so-called global North and the so-called global South.

The key lesson is that this story of an AI Cold War divide is not only dangerous and kind of a self-fulfilling prophecy, it’s wrong. When you look at what happened in the room where we negotiated a declaration on AI that was signed by 56 countries from all continents, what came clear was that our populations, our people, our country, what they want is basically to put AI to good use, to echo what Frederic said, to modernize health care, education, agriculture, public services, and to do so in a way where they decide for themselves.

The key word was sovereignty, but it doesn’t mean autarky, it doesn’t mean being isolated. Sovereignty is compatible with openness, but not with vulnerability. So if you have to remember one key word from the Paris AI Action Summit is this notion of open sovereignty, which I think is the way, the third way, to develop a global platform for cooperation on AI.

And I think this way, you asked me to be concrete, it’s not abstract, it’s not only a political concept, it’s practical. This third way is based on, you know, deploying AI at scale in our countries, in schools, in hospitals, in public services. I was yesterday speaking in front of the French unemployment agency that is experimenting with AI to connect people with job offers in their region.

So you revert the narrative that AI destroys jobs, but by actually creating more opportunities. Second, it’s also reducing dependency on just a handful of very concentrated infrastructures that can be used as geopolitical levers of coercion. And third, I would say this open sovereignty, it’s not putting all your eggs in one basket.

It’s diversifying your tech partnership. It’s called risk management or hedging or diversification. You could have part of your stack that comes from one country and part from another.

I speak on behalf of a country, France, that’s heavily dependent on foreign technology, but still with players like Mistral or others aims to have a degree of national stack as well. And for that, we have assets to build on. We have our local data, which is extremely rich in countries like India and the Global South with languages, with demographics, with young talent.

And we also have a political instinct, I guess, of non-alignment to quote a historical term that is very alive and I think could have a new youth in this age of AI. To conclude, maybe with one word on the French and European perspective, I think France and Europe in general, along with India, with the UN and all other partners who want to architect this coalition, we have a role to play to try and bridge those gaps between those geographies, because we always stood for sovereignty.

You mentioned, Avinash, that those ideas of democracy, et cetera, they are born in Europe. I think it’s not true. They have several sources.

But there is part of this legacy that is still very valid and that sovereignty should be based on the people’s wish and also compatible with openness. Practically, it means three things that we could put in this global coalition for an open sovereignty in AI. First, and I will quote my Swiss colleague, it’s neutral rails.

We should have cloud and data standards that guarantee interoperability and reversibility, not putting you in the hands of just one actor. Second, it’s diverse models. Our countries don’t need one full stack, but they need a mix of open and commercial LLMs of large and small, of multilingual and specialized ones.

And third, it’s partnership, North-South partnerships that are rooted, that build on our strengths and put AI to good use. So the real choice is very simple. Either we want AI to be a marketplace of dependence, you choose your boss and you go with that technology, or it can be an infrastructure of freedom.

I’m in favor of the second approach, and to end on a quote by an economist, sorry Avinash, I think it’s time to give some reality to what Paul Romer called in his Nobel lecture, the possibility of progress, because that’s what people are asking for.

Thank you very much.

Avinash Dadhich

Thank you very much.

Avinash Dadhich

Speech speed

148 words per minute

Speech length

2256 words

Speech time

909 seconds

Globalization and interconnectedness are not new phenomena, having existed for thousands of years through trade, art, and cultural exchange

Explanation

Dadhich argues that globalization is not a modern concept but has existed throughout history as civilizations were interconnected through various means. He emphasizes that humans have always been connected and learned from each other across different regions and cultures.

Evidence

Disagreed with

– Kalika Bali

Disagreed on

Global South vs Global North classification relevance

Jovan Kurbalija

Speech speed

159 words per minute

Speech length

383 words

Speech time

143 seconds

Historical principles like Hammurabi’s law of builder responsibility were abandoned in 1996 when tech companies were exempted from content responsibility, creating current governance challenges

Explanation

Kurbalija argues that a fundamental shift occurred in 1996 when the principle of creator responsibility, which had existed since ancient times, was abandoned for tech companies. He views this as the ‘original sin’ that created many current AI governance problems.

Evidence

References Hammurabi’s law stating that if a builder builds a house and it collapses, the builder is responsible – a principle carried through Roman law until 1996 when tech companies were told they are not responsible for content

Major discussion point

Historical Context and Lessons for AI Governance

Topics

Legal and regulatory | Liability of intermediaries

Agreed with

– Avinash Dadhich

Agreed on

Historical precedents provide valuable lessons for AI governance

Existing legal frameworks can address most AI issues, with only intensive regulatory interventions needed for cybercrime and data protection

Explanation

Kurbalija suggests that most AI governance issues can be handled by existing laws rather than creating entirely new regulatory frameworks. He advocates for a measured approach, drawing parallels to how horses were integrated into society without needing ‘horse law.’

Evidence

Cites a 1996 Chicago University professor’s question about whether we needed ‘horse law’ when horses were introduced, and notes that over three decades, only cybercrime and data required intensive regulatory interventions

Major discussion point

Historical Context and Lessons for AI Governance

Topics

Legal and regulatory | Cybercrime | Privacy and data protection

Agreed with

– Alexandra Baumann
– Lucia Velasco

Agreed on

Building on existing institutions rather than creating entirely new frameworks

Disagreed with

– Alexandra Baumann

Disagreed on

Approach to AI regulation – existing vs new frameworks

Alexandra Baumann

Speech speed

136 words per minute

Speech length

618 words

Speech time

270 seconds

AI policy must be inclusive, human-centered, and multilateral, involving all stakeholders across regions, especially underrepresented countries

Explanation

Baumann emphasizes that effective global AI coordination requires broad participation from all stakeholders and regions, with particular attention to traditionally underrepresented countries in tech governance. She advocates for putting humans at the center of AI development through multilateral and multistakeholder approaches.

Evidence

Frederic Werner

Speech speed

182 words per minute

Speech length

Barriers to AI Implementation and Solutions

Topics

Privacy and data protection | Data governance

Agreed with

– Kalika Bali
– Hu Ling

Agreed on

Data sharing and access challenges are major barriers to AI development

Kalika Bali

Speech speed

149 words per minute

Speech length

742 words

Speech time

298 seconds

Global South voices are largely missing from AI policy discussions, despite having different cultural contexts, resource constraints, and data challenges

Explanation

Bali argues that AI policy discussions lack adequate representation from the Global South, whose perspectives are crucial because they face unique challenges different from Europe and the US. She emphasizes that without these voices, policies may not address the real needs of the majority of the world’s population.

Evidence

Notes attendance at various summits and workshops with little Global South representation in policy discussions, and explains that most Global South regions are oral rather than written digital cultures, creating different data challenges

Major discussion point

Global South Representation and Perspectives

Topics

Development | Cultural diversity | Capacity development

Agreed with

– Alexandra Baumann
– Lucia Velasco

Agreed on

Need for inclusive global representation in AI governance

Privacy concepts are culturally specific, and Global South countries face unique challenges like being primarily oral rather than written digital cultures

Explanation

Bali contends that fundamental concepts like privacy vary significantly across cultures, and that Global South countries have different technological and cultural contexts that must be considered in AI governance. She argues that policies developed without understanding these differences may be inappropriate or ineffective.

Evidence

Disagreed with

– Avinash Dadhich

Disagreed on

Global South vs Global North classification relevance

Hu Ling

Speech speed

136 words per minute

Speech length

498 words

Speech time

218 seconds

AI represents a potential new mode of social production, similar to how the internet developed through initially ‘illegal’ content sharing that eventually proved valuable

Explanation

Hu Ling draws parallels between current AI development and the early internet, suggesting that what appears to be copyright violation may be a necessary stage in developing new productive capabilities. He argues that this ‘illegal rise’ pattern may be essential for technological advancement.

Evidence

References the internet’s development 20 years ago when there was initially no content, leading to what he calls the ‘illegal rise of the internet’ through lower-cost information sharing that eventually made the internet valuable and widely adopted

Major discussion point

Intellectual Property and Data Rights

Topics

Intellectual property rights | Digital business models

Agreed with

– Frederic Werner
– Kalika Bali

Agreed on

Data sharing and access challenges are major barriers to AI development

AI companies must prove they add value rather than simply redistributing existing value to justify their use of copyrighted materials

Explanation

Hu Ling argues that for AI to be considered a legitimate new mode of production, companies must demonstrate that they create additional value beyond simply moving existing value from one place to another. Without this proof, AI development could be considered mere theft of intellectual property.

Evidence

References ongoing legal battles in China, the United States, and EU between AI companies and traditional content creators (authors, singers, artists), and poses the question of whether AI represents genuine value creation or just value redistribution

Major discussion point

Intellectual Property and Data Rights

Topics

Intellectual property rights | Legal and regulatory

Lucia Velasco

Speech speed

115 words per minute

Speech length

665 words

Speech time

346 seconds

The UN has launched new AI governance mechanisms including a 40-expert scientific panel and a global dialogue platform, both focusing on non-military AI domains

Explanation

Velasco announces the formal launch of new UN AI governance structures that emerged from two years of negotiations and processes. These mechanisms are designed to provide both scientific expertise and inclusive dialogue platforms for global AI governance.

Evidence

Agreed with

– Alexandra Baumann
– Kalika Bali

Agreed on

Need for inclusive global representation in AI governance

Alexandre Mirlesse

Speech speed

172 words per minute

Speech length

Open Sovereignty and Third Way Approach

Topics

Digital standards | Telecommunications infrastructure | Development

Agreements

Agreement points

Historical precedents provide valuable lessons for AI governance

Speakers

– Avinash Dadhich
– Jovan Kurbalija

Data sharing remains problematic as organizations have interesting data but are unwilling to share it due to privacy concerns

Privacy concepts are culturally specific, and Global South countries face unique challenges like being primarily oral rather than written digital cultures

AI represents a potential new mode of social production, similar to how the internet developed through initially ‘illegal’ content sharing that eventually proved valuable

Summary

All three speakers acknowledge that data access and sharing present fundamental challenges for AI development, though they approach the issue from different perspectives

Topics

Privacy and data protection | Data governance | Intellectual property rights

Digital access | Legal and regulatory | Telecommunications infrastructure

Unexpected consensus

Measured approach to AI regulation rather than rushing to create new laws

Speakers

– Jovan Kurbalija
– Hu Ling

Arguments

Existing legal frameworks can address most AI issues, with only intensive regulatory interventions needed for cybercrime and data protection

AI companies must prove they add value rather than simply redistributing existing value to justify their use of copyrighted materials

Explanation

Despite coming from different backgrounds (diplomatic/legal vs academic), both speakers advocate for patience and careful evaluation before implementing new regulations, suggesting that existing frameworks may be sufficient

Topics

Legal and regulatory | Intellectual property rights

Recognition that AI governance requires balancing innovation with human rights

Speakers

– Alexandre Mirlesse
– Lucia Velasco
– Alexandra Baumann

Arguments

Countries should develop AI infrastructure based on neutral rails, diverse models, and North-South partnerships rather than choosing dependence on single actors

The scientific panel will provide evidence-based policy assessments, while the global dialogue will enable multilateral and multi-stakeholder discussions

AI policy must be inclusive, human-centered, and multilateral, involving all stakeholders across regions, especially underrepresented countries

Explanation

Representatives from different institutional backgrounds (French diplomacy, UN, Swiss diplomacy) converge on the need for balanced, inclusive approaches that avoid technological dependence while promoting human-centered development

Topics

Human rights principles | Legal and regulatory | Development

Overall assessment

Summary

The speakers demonstrate significant consensus on key principles including the need for inclusive governance, building on existing institutions, addressing data and connectivity challenges, and moving beyond traditional geopolitical divisions. There is broad agreement on human-centered approaches and the importance of balancing innovation with rights protection.

Consensus level

High level of consensus on fundamental principles and approaches, with differences mainly in emphasis and specific implementation strategies. This suggests a strong foundation for collaborative AI governance efforts, though practical implementation may still face challenges in reconciling different national interests and capabilities.

Differences

Legal and regulatory | Digital standards

Unexpected differences

Digital access | Legal and regulatory | Telecommunications infrastructure

Takeaways

Key takeaways

AI governance challenges are not entirely new – historical precedents exist, and existing legal frameworks can address most AI issues with targeted interventions only needed for specific areas like cybercrime and data protection

Global AI coordination requires inclusive, human-centered approaches that bridge global principles with local realities while respecting cultural differences and national contexts

Major barriers to AI scaling include data access issues, connectivity gaps (one-third of world unconnected), and governance fragmentation

Global South voices are significantly underrepresented in AI policy discussions despite facing unique challenges related to oral vs. written cultures, resource constraints, and different privacy concepts

The traditional Global North/South divide is becoming less relevant; the real distinction is between powerful and less powerful actors in AI governance

AI development creates universal intellectual property conflicts that require balancing innovation needs with creator compensation

The UN is establishing new AI governance mechanisms (scientific panel and global dialogue) to serve as neutral convener rather than direct regulator

An ‘open sovereignty’ approach offers a third way beyond US-China binary divisions, combining national independence with international openness through diverse partnerships and neutral infrastructure

Resolutions and action items

UN launched scientific panel with 40 experts and global dialogue platform for AI governance in non-military domains

Applications for UN scientific panel open until October 31st, with emphasis on global regional balance

First full UN global dialogue meeting scheduled for July 2026 in Geneva alongside ITU-AI for Good Summit

ITU launched AI skills coalition to address skills gap and collaboration on AI watermarking for deepfake detection

Created toolkit to help member states address misinformation and build strategies around it

Support for International Computation and AI Network (ICAIN) to promote equitable distribution of technological benefits

Advocacy for using existing World Summit on Information Society process to implement Global Digital Compact rather than creating new mechanisms

Unresolved issues

How to effectively share data while respecting privacy concerns – organizations have valuable data but are unwilling to share it

How to ensure meaningful Global South participation in AI governance beyond just use cases

How to balance AI companies’ need for training data with intellectual property rights and creator compensation

How to prove that AI represents a genuine new mode of production rather than just value redistribution

How to address cultural differences in privacy concepts and other values in global AI governance frameworks

How to connect the remaining one-third of the world’s population to benefit from AI developments

How to create content in local languages and dialects to incentivize connectivity in underserved regions

How to move from AI governance discussions and frameworks to tangible implementation and impact

Suggested compromises

Open sovereignty approach that combines national independence with international openness rather than choosing binary alignment

Diversified technology partnerships where countries use mixed stacks from multiple sources rather than single-vendor dependence

Regulatory interoperability that allows countries to align on shared norms while adapting to local legal systems and contexts

Building on existing institutions and processes (like Geneva’s international organizations) rather than creating entirely new governance structures

Neutral rails approach with cloud and data standards that guarantee interoperability and reversibility

This clarification addresses a fundamental misunderstanding about international governance, distinguishing between convening/coordinating roles versus direct regulatory authority, which is crucial for realistic expectations about global AI governance.

Impact

This comment provided important clarity about what global coordination can realistically achieve, tempering expectations while highlighting the UN’s actual value proposition. It helped ground the discussion in practical realities of international governance rather than idealistic but unrealistic expectations.

Overall assessment

Explanation

This supports the concept of open sovereignty by avoiding dependence on any single AI approach or provider

Disclaimer: This is not an official session record. DiploAI generates these resources from audiovisual recordings, and they are presented as-is, including potential errors. Due to logistical challenges, such as discrepancies in audio/video or transcripts, names may be misspelled. We strive for accuracy to the best of our ability.