Indias AI Leap Policy to Practice with AIP2

20 Feb 2026 14:00h - 15:00h

Session at a glance

Summary

This discussion focused on strategies for accelerating AI diffusion globally, particularly in developing economies and the Global South, with emphasis on moving from pilot projects to scalable implementation. The conversation took place during an AI summit in India and featured perspectives from international technology leaders, policymakers, and governance experts.

Doreen Bogdan-Martin from the ITU emphasized three key pillars for AI diffusion: solutions (building accessible infrastructure and connectivity), skills (developing digital literacy and capacity), and standards (ensuring interoperability and trust). She highlighted India’s success with the Bishini platform, which delivers government services in 22 languages, as an exemplary model of inclusive AI implementation. Dr. Panneerselvam Madanagopal discussed the critical role of startups as “AI natives” in bridging the gap between advanced technology and business needs, particularly for small and medium enterprises that may experience “technology overshoot.”

The panel addressed significant challenges in AI adoption, with Rachel Adams presenting concerning data from South Africa showing that two-thirds of citizens lack meaningful understanding of AI, creating a substantial democratic gap. This highlighted the urgent need for public awareness and digital literacy initiatives alongside technical deployment. Brando Benefi, co-reporter of the EU AI Act, stressed the importance of clear regulatory frameworks that provide precise boundaries rather than vague ethical guidelines, noting that trust requires explicit governance structures.

The discussion emphasized that successful AI diffusion requires balancing global standards with local adaptation, ensuring meaningful participation from Global South countries in standard-setting processes, and prioritizing education and institutional capacity building over purely technical solutions.

Keypoints

Major Discussion Points:

– AI Infrastructure and Accessibility: The need to build foundational infrastructure (connectivity, platforms) to make AI accessible globally, with emphasis on connecting the remaining third of humanity that is offline and ensuring AI doesn’t exacerbate the digital divide.

– Skills Development and Digital Literacy: Critical importance of AI literacy and skills training at all levels, from basic public understanding to technical capabilities, with particular focus on addressing the massive skills gap that exists globally, especially in developing economies.

– Standards and Governance Frameworks: The role of international standards in ensuring AI systems work together effectively while building trust, including specific discussions about the EU AI Act as a reference point and the need for clear, practical governance rather than vague ethical appeals.

– Trust, Ethics and Democratic Participation: Challenges around public awareness and understanding of AI (with examples like two-thirds of South Africans lacking meaningful AI knowledge), the risks of AI being used for authoritarian control, and the need for inclusive, participatory governance processes.

– Global South AI Diffusion and Local Adaptation: Strategies for moving from AI pilots to scaled deployment in developing economies, balancing global standards with local needs, and learning from India’s experience with large-scale technology diffusion while avoiding one-size-fits-all approaches.

Overall Purpose:

The discussion centered around launching the “Global South AI Diffusion Playbook” and exploring practical strategies for ensuring AI benefits reach developing economies inclusively and responsibly. The conversation aimed to move beyond theoretical “moonshots” to actionable implementation approaches.

Overall Tone:

The discussion maintained a constructive and collaborative tone throughout, with speakers building on each other’s points rather than debating. There was a sense of urgency mixed with cautious optimism – participants were energized by AI’s potential (described as an “AI earthquake”) while remaining realistic about challenges like governance gaps, skills shortages, and risks of exclusion. The tone was notably practical and solution-oriented, with speakers sharing concrete examples and specific recommendations rather than abstract concepts.

Speakers

Speakers from the provided list:

– Doreen Bogdan-Martin

– Role/Title: Not explicitly stated in transcript, but appears to be in a leadership position at ITU

– Area of expertise: Digital connectivity, AI accessibility, international telecommunications policy

– Rachel Adams

– Role/Title: Founder and CEO of the Global Center on AI Governance

– Area of expertise: AI governance, equity and human rights in AI development, advises governments, contributed to African Union Commission’s Continental AI Strategy

– Moderator

– Role/Title: Event moderator

– Area of expertise: Facilitating discussions on AI and technology policy

– Fred Werner

– Role/Title: Chief of Strategy and Operations for AI for Good and Chief of Strategic Engagement at ITU

– Area of expertise: AI standards, international AI collaboration, co-creator of AI for Good Global Summit

– Dr. Panneerselvam Madanagopal

– Role/Title: CEO of METI Startup Hub

– Area of expertise: Startup ecosystem development, deep tech startups, connecting government policy with entrepreneurial energy

– Brando Benefi

– Role/Title: Member of European Parliament, co-reporter of the EU AI Act, Italian MEP since 2014

– Area of expertise: European digital and AI policy, AI regulation and governance

Additional speakers:

None identified beyond the provided speakers names list.

Full session report

This panel discussion took place during a major AI summit in India, following the launch of the “Global South AI Diffusion Playbook” and exploring practical strategies for moving AI from pilot projects to scalable, inclusive implementation worldwide. The conversation brought together international technology leaders, policymakers, and governance experts to address how AI benefits can reach developing economies responsibly and equitably.

The Three Pillars Framework

Doreen Bogdan-Martin from the International Telecommunication Union (ITU) established a foundational framework with three pillars for successful AI diffusion: Solutions, Skills, and Standards. This framework emerged from observations of India’s success with platforms like Bishini, which delivers government services in 22 languages.

The Solutions pillar addresses the fundamental reality that one-third of humanity remains offline. As Bogdan-Martin emphasized, “without connectivity, there is no AI.” She highlighted initiatives like the GIGA programme with UNICEF to connect every school and the Partner to Connect Digital Coalition, which has secured commitments towards connecting hard-to-reach communities.

The Skills component focuses on developing digital literacy at all levels. Bogdan-Martin described skills as “that engine of agency,” pointing to India’s Future Skills Programme and the ITU’s skilling coalition comprising 70 partners offering over 180 learning resources in 13 languages.

The Standards pillar ensures AI systems work effectively together while embedding trust and interoperability. This includes developing standards to combat deepfakes and misinformation, which Prime Minister Modi had identified as threats capable of destabilizing societies.

Startups as AI Transmission Mechanisms

Dr. Panneerselvam Madanagopal from METI Startup Hub described this moment as an “AI earthquake” and “tectonic shift,” positioning startups as “AI natives” that serve as transmission mechanisms for AI diffusion. He outlined METI’s “three M’s” approach: Mentorship, Market access, and Money, emphasizing that “your customer is your best investor if you’re a startup.”

Dr. Madanagopal introduced the concept of “technology overshoot,” where AI capabilities have exceeded small and medium enterprises’ ability to understand and integrate the technology effectively. Startups can serve as bridges between advanced technology and business needs, helping traditional enterprises navigate AI adoption at their own pace.

With substantial funding available—METI Startup Hub manages almost 1,000 crores while the India AI mission has allocated about 8,000 crores—capital availability is not the primary constraint. Instead, the focus shifts to building capacity and enabling startups to drive broader economic transformation.

The Democratic Understanding Gap

Rachel Adams from the Global Center on AI Governance presented research findings that challenged assumptions about AI readiness. Her survey of over 3,000 South Africans, conducted in over 11 official languages, revealed that two-thirds of the population lack meaningful understanding of AI. Specifically, one-third have never heard of AI, while another third have heard of it but could not explain what it means.

This research exposes a critical “democratic gap” where AI adoption in public services proceeds without public understanding or meaningful participation in decision-making. Adams argued that citizens cannot contest or participate in decisions about technologies that directly affect their lives, creating fundamental challenges for democratic governance and accountability.

Governance Beyond Voluntary Ethics

Brando Benefi, co-reporter of the EU AI Act, argued that voluntary ethical frameworks alone are insufficient. “If you substitute these completely with mere voluntary ethical frameworks, I’m not sure we are getting anywhere,” he stated. Instead, he advocated for precise governance mechanisms with clear boundaries and accountability measures.

Benefi noted that some private sector actors deliberately delay standards implementation to avoid regulatory compliance, leading to proposals for time-limited mechanisms to ensure standards development keeps pace with regulatory requirements. He also acknowledged AI’s potential for mass surveillance and repression, particularly in institutionally fragile contexts.

The Dual Nature of AI Applications

Fred Werner from ITU’s AI for Good initiative illustrated AI’s complex dual-use nature through an Estonian startup example that detects blood sugar levels through voice analysis. While revolutionary for diabetes care, the same technology raises privacy questions about what else it might reveal—sleep patterns, diet, medication use, or attention levels.

This example highlighted the broader governance challenge: beneficial applications inherently create surveillance capabilities that could be misused. Werner emphasized that high-potential AI applications cannot be assumed to develop appropriately without deliberate attention to safety, security, ethics, human rights, and inclusive design.

Regional Adaptation and Global Standards

The discussion revealed tensions between global harmonization and local adaptation needs. Adams argued against one-size-fits-all approaches, noting that “interoperability can often mean the dominance of one particular region or worldview’s regulatory regime everywhere else.” She advocated for global consensus around core principles—accountability, transparency, safety, and human oversight—while allowing regional adaptation of implementation standards.

The challenge lies in ensuring meaningful participation from Global South countries in standards-setting processes, which have historically been dominated by those with greater resources. Adams called for deliberate funding and leadership opportunities for representatives from Africa, Latin America, and Asia.

Investment Priorities

When asked how they would spend a hypothetical billion dollars to accelerate AI diffusion, all three panelists emphasized education and skills development as foundational. Werner focused on digital literacy and questioned whether leapfrogging might be possible with AI infrastructure, similar to mobile payment innovations in East Africa.

Adams prioritized digital literacy to address the public awareness gap and building capacity in state institutions—competitions commissions, gender equality commissions, human rights commissions, and information regulators—as crucial for championing citizens’ rights.

Benefi supported literacy priorities while emphasizing building consciousness and capacity among civil society actors during rapid AI development periods.

Learning from India’s Approach

The discussion consistently referenced India’s experience as a potential model, particularly its inclusive approach to technology deployment. Adams praised the summit’s community participation approach, contrasting it with more elite-focused international AI conferences. The moderator noted India’s track record with large-scale technology diffusion, including digital identity and financial inclusion systems serving over a billion people.

However, participants acknowledged that replicating India’s model requires understanding both the specific conditions that enabled its success and necessary adaptations for different contexts.

Moving from Moonshots to Implementation

The overarching theme emphasized practical implementation challenges rather than technological capabilities alone. The discussion focused on building public understanding, creating governance frameworks, developing institutional capacity, and ensuring inclusive participation in AI development and deployment.

Success will be measured not by technological sophistication but by whether AI systems “work reliably, inclusively, and productively for many.” This human-centered approach, emphasizing equity and democratic participation alongside technical capability, suggests a more sustainable approach to AI development.

The conversation reinforced that AI diffusion is not merely a technical challenge but a complex socio-political process requiring careful balance between innovation and governance, global coordination and local adaptation, and technological capability and human agency. Success depends on continued collaboration across different perspectives, with particular attention to ensuring AI benefits reach those historically excluded from technological progress.

Session transcript

Doreen Bogdan-Martin: …as to how AI can actually benefit people in their lives, their homes, their communities, and their businesses. The second point that keeps coming up is that it’s not a one -size -fits -all model. I think we do need to be flexible. We need to be inclusive when we look at different AI approaches. I would say for all parts of the world, no matter where countries are in terms of their development journeys. India, as we see, is a leader, really showing how to get from AI ambitions to real results. And, of course, in doing so, keeping people, keeping… …that human -centered approach in focus, as we heard from the Prime Minister yesterday. The Bishini platform… that we’ve also heard about, delivers government services in 22 languages. I would say as well, similar AI -powered digital public infrastructure solutions in areas from health care to financial inclusion are really working to better serve all Indians, regardless of their economic status, their skill level, especially in rural communities. I would say inspired by these efforts, I wanted to quickly offer three observations, and you’ve actually already referred to them. Three observations about how we can move beyond moonshots from policy to actual practice here in the Asia -Pacific region and beyond. And they all begin with S, and you said them already, solutions, skills, and opportunities. Of course, standards. So Solutions is about building the infrastructure and the platforms that make artificial intelligence accessible because we cannot achieve AI for many. We can’t achieve AI for all if we still have a third of humanity that is offline. Without connectivity, there is no AI, and that’s why efforts like our school connectivity work with UNICEF called the GIGA initiative to connect every school is so important. Our work in terms of our partner to connect, Digital Coalition, which is about connecting the hardest to connect. We have a target of achieving 100 billion this year. So far, we’re at 80 billion in commitments and pledges to connect the hardest to connect. So we need to tackle that basic infrastructure component. The second element that we need to make sure that we diffuse AI globally in practice is skills. the fundamental importance of skills. Yesterday I was speaking to a young leader who actually likened connectivity to people feeling that they have digital agency. Skills are that engine of agency. Countries can learn directly from India’s experience of investing in people, namely through its Future Skills Program that’s providing upskilling to support thousands of students at all levels. ITU is also taking a similar approach, and my colleague Fred will be staying on for the panel today. We have a skilling coalition that’s very exciting with some, I think, 70 partners so far, bringing more than 180 different learning resources in 13 languages. And coming to my last S is that standards piece. Ensuring that AI systems work effectively together. Thank you. Standards complement solutions and skills not only for interoperability but also for embedding trust. As Prime Minister Modi mentioned yesterday, deep fakes and misinformation can destabilize entire societies. And people must be able to distinguish between real and AI -generated material. And that’s why the ITU, together with our partners from ISO and IEC, we created the AI Standards Exchange Database that has over 850 standards and technical publications, including multimedia authenticity standards, that prioritize traceability to combat deep fakes. ITU standards are voluntary. They are developed through an inclusive… multi -stakeholder process. So ladies and gentlemen, AI diffusion isn’t about everyone using the same technology. It’s about giving everyone the same bridge to opportunity and refusing to let the digital divide become an AI divide. So today’s playbook is going to help us really build that bridge, as will our continued cooperation and collaboration on AI solutions, skilling, and standards. In all of these areas, you can count on ITU as your trusted partner. Thank you.

Moderator

Thanks, Doreen. As you can see, Doreen has spent her career in ensuring. Every country, every community has access to or is part of digital economy. Could I just invite Doreen, Fred, Rachel, Brando, Dr. Bani -Selvan on the stage as we launch the Global South AI Diffusion Playbook. It’s a framework built around five interacting dimensions, infrastructure, data and trust, institutions for procurement, and skills and market shaping. It’s not designed as a strategy document, but more as an implementation guide, because the next phase of AI is not about moonshots, it’s about how do we ensure AI works reliably, inclusively, and productively for many. This is, I think, the photo op you guys were waiting for, so all yours.

Thank you. Doreen I know you have to leave thank you very much thanks for a great keynote as well thanks Doreenn if diffusion is about moving from capability to real economic impact then startups are obviously the transmission mechanism and a few people understand India startup ecosystem as deeply as Dr. Pani Selvam Madan Gopal CEO of Miti Startup Hub under his leadership Miti Startup Hub has become a key platform connecting government policy with entrepreneurial energy enabling innovations to move from lab to market and from pilot to scale over 6000 plus startups right so he brings over 2 decades of experience and at a moment when India is positioning itself not just as an AI industry but as an AI adopter but as an AI innovation diffusion hub his perspective on enabling startups to scale responsibly and globally particularly valuable.

Doctor, would I just have a few minutes for you.

Dr.Panneerselvam Madanagopal

Thank you, Access Partnership for having me this afternoon for this conversation. I think it’s an important element. How do you know there’s so much happening in the last four to five days in Delhi in Bharat Mandapam. So it’s important to get a grasp of what’s going on. And what each of us have to kind of take away from this and how each stakeholders in this ecosystem can help us. And startups become a very, very important player in this game. And essentially for two or three key reasons. One, they come in as AI natives. They come in with a significant understanding of the technology and the talent is kind of already there and then second they are here to the agility that they bring and the capability they bring to kind of transform businesses is becoming a very very important need for small and medium enterprises and even to large enterprises.

Just prior to this I was having a conversation with a large corporate and how they can actually use startups as a catalyst of change and transformation in their large corporate because the corporates are designed for systems and processes on scale and what need of the hour is actually agility adaptability and more importantly ability to change and bring innovation into a mainstream of any enterprise. So startups play a very very critical role so we at Métis Startup Hub are primarily driving the push to kind of ensure that startups have the wherewithal and the capability to drive and back this change that is required by the corporate ecosystem or the large enterprise ecosystem. So, briefly what do we do at METI Startup Hub?

We are the custodians of the deep tech startups in the country. This whole event has been put together by METI and of course Ministry of External Affairs has been phenomenal partners in this. So, our role in METI Startup Hub is essentially three M’s. Mentorship, market access and money. This is essentially what we provide for startups. We provide mentorship support to the entire journey from almost at an ideation stage to CDC, up to CDC level. And we provide them with market access. I’m a firm believer that your customer is your best investor if you’re a startup. And finding customers for startups is more important than finding investors, right? So, it’s important for me to find, give them the right market access support.

So, we work with large corporates across the board, across the country. when internationally we kind of drive market access support and last but not the least money we provide significant amount of, there is absolutely no death of capital in the Indian market, you know, through my agent, through my organization, MIT Startup Hub, we fund almost up to a thousand crores for startups and the India AI mission has another almost about 8000 crores to be funded for funding for startups. So there is absolutely no death of money in the market, government fund is available, you know, private capital is available, so that’s what we support. And our endeavor is to ensure that startups are at the heart of this renaissance of this change that is kind of happening in the ecosystem and how startups technology can power, help this small and medium enterprises to grow.

So that’s what we are trying to do. Thank you. Thank you. Thank you. journey. So that’s what we have been driving at and conversations like this help a lot and enabling them to drive this change. We have three things I mean there is obviously a lot of challenges. It’s not easier said than done. In some cases I was reading up in a way with medium enterprises we call a technology overshoot. The technology has actually overshot the need and now the ability of the medium enterprises to cope with this technology and say how do I understand what is my need? How do I integrate this into my business need and how do I ensure that this business my business is realigned with a new workflow, a new way of doing business with this current technology with AI or AI based supported technology to kind of drive.

So there is, while there are huge challenges, but every challenge is an opportunity. So, you know, and startups are very well placed to kind of bridge that opportunity because they understand technology and they understand business. So we are hoping to create this, what I call the AI bridge now, which is, you know, kind of bridge the technology and the business need. And it’s going to be a huge opportunity by itself to kind of drive, and startups are what we are hoping will build that bridge and drive the change. So at METI Startup Hub, our endeavor is to nurture, build, and enable tech and deep tech startups in the country. And we partner with all, we collaborate with all stakeholders, domestic and international, to ensure our startups get the right, opportunities and we solve.

problems and we enable capability through building capacity. So that’s essentially in a nutshell what we do and once again I thank the access partners for providing me this opportunity to briefly share my thoughts with you and we are in a cusp of somebody called this an AI earthquake happening in Bharat Mandapam. This is a tectonic shift and this is some laying foundation for something big and better coming our way. Of course with a lot of responsibility also because everything has two sides of its own so we need to be extremely responsible in what we are doing with the technology. Thank you once again. Thank you for the opportunity. Thank you.

Moderator

As Dr. said, this is really the earthquake of AI and we are at the epicenter. And as you can see, after five days, we are all very, very tired. We started late. We’ll end on time. That’s my promise to you guys. Where is the next chair? So let me introduce our panelists very quickly. Dr. Rachel Adams, she’s the founder and CEO of the Global Center on AI Governance, a leading research and policy institution focused on ensuring that AI development and deployment advance equity and human rights globally. She also advises governments and she was a key contributor to the African Union Commission’s Continental AI Strategy. I have Fred, Fred Werner. He is the Chief of Strategy and Operations for AI for Good and Chief of Strategic Engagement at ITU.

He’s based in Geneva, but as a co -creator of the AI for Good Global Summit. which is happening from 7 to 10 July in Geneva. He brings together a global hub for collaboration standards and actionable AI -driven impact. And I’m also pleased to welcome Brando Benefi, who is a member of European Parliament, and he was a co -reporter of the EU AI Act, which we all love so much, the world’s first comprehensive AI regulation. He is an Italian MEP since 2014, and he has played a key role in shaping European digital and AI policy. Welcome, all of you. Thank you. Quick one, yeah? I’ll really start with you, Brando, in this case. We talked about concrete gains that AI diffusion can unlock in Global South over the next three to five years.

How do we move from pilots to scale deployment? I want to understand a bit more from you. it’s been a while since we have had the EU AI Act. There have been some implementation, obviously, right? So how do you see the AI diffusion being unlocked and how do you see European partnerships with the Global South there?

Brando Benefi

Well, first of all, I apologize for my voice, but it’s the, I don’t know, work of these days. Maybe we are producing a lot, but this is also the impact, so I apologize for that. But to answer your question, I think that the EU AI Act can be an interesting reference point to reflect on what we can do to implement the idea of a global diffusion, especially looking at the Global South. Because, in fact, even the so -called global north or global minority, we can use different terms, is still struggling with the diffusion of AI among different actors. If you look at the data, for example, on the diffusion among small and medium -sized enterprises, most north of the world countries, they still have very low numbers because of lack of trust, because of lack of AI literacy, because of lack of systems that facilitate understanding on how the usage of AI can ameliorate the activity of a business, a public organization, a civil society reality, etc.

So, the AI Act is a legislation that doesn’t… doesn’t create a comprehensive framework that is vague. comprehensive but confusing maybe instead it chooses to identify a series of high risk areas of usage of AI and lets instead all the non -included use cases to not be regulated further than the existing legislation. Why I’m saying this? Because I think that to overcome one of the issues obviously when we look at the issue of diffusion there are many elements infrastructure, as I said, literacy but on the issue of trust and of risk management I think the UAI Act is an interesting reference point on having clear boundaries where we do not think we need more regulation where we let the systems be used freely, where we want checks and balances to be in place where we even choose to prohibit certain use cases and where we need transparency which is still a lacking element in many of our experiences with AI so I think that in the difference of the context these elements are quite relevant for even a context that is clearly different from the average European country but I think that to build trust we need to clarify where we want governance and limits to be in place and send a clear message to the population that even when we concentrate on EU use cases, on action, this is the topic of the summit we can also build in a smart way, in a clear way, not light, but clear, clarity, elements of protection, of guarantee that can create more trust in the adoption.

Moderator

Brando, I know why your voice is like that, because people want to hear more from you. That’s why you will have a busy day today as well. I’m sure people want to talk a lot to you. Rachel, coming to you, I think Brando talked about an important point about the trust and clarity, and you have worked extensively with global south countries, right? So how crucial do you think are trust and ethics for diffusion? How do you see that actually getting implemented in practice?

Rachel Adams

Yeah, I think it is going to take far more work than perhaps we feel it might. So, you know, Brando, I think you mentioned some very important points around public awareness and understanding. In South Africa, the center I lead, the global center on… AI Governance conducted a very comprehensive public perception survey in the country. We interviewed over 3 ,000 South Africans from all walks of life, all demographic groups. We interviewed them in their own language. We have over 11 official languages in South Africa. And two -thirds of South Africans do not have a meaningful grasp of AI. So one -third of South Africans have never heard of AI, and another third of South Africans have heard of it but could not begin to tell you what it meant at all.

So I think if we’re thinking about the relationship between the large -scale private investments we’re seeing in AI diffusion, the large -scale public plans we have around AI adoption in relation to where the public sits, what their kind of levels of understanding are, and awareness and literacy is, this is going to create, this is creating a very significant significant democratic gap, particularly where a lot of these adoption pathways are around the use of AI in the public service. People don’t know about these technologies. They don’t know about the risks. They don’t know about the opportunities. They’re not able to contest it. They’re not able to participate in decision -making. We have a real problem. So diffusion cannot be something that is only about putting in place the infrastructure that sees forward technical delivery and access.

It must be scaled with governance efforts.

Moderator

I think, Brando, we had that whole discussion separately where you talked about that getting technology in the hand of people doesn’t matter if you’re using it for a lot of autocratic rules, like for example, social scoring, right? So I think maybe going to you, Fred, on this point, looking at the positive side of the story, you talked about that day, AI for good or AI for good. So how do you, some of the use cases and standards that you think are really setting the stage for helping? drive the diffusion?

Fred Werner

Yes, I think there’s no shortage of high potential AI for good use cases, especially now in 2026. That maybe wasn’t the case in 2017 when we created AI for good, but we’ve really seen things go from the hype, the fear of a promise, mainly existing in fancy marketing slides, to the advent of Gen AI, the rise of AI agents, and now the physical manifestation of AI in the form of robotics, embodied AI, brain -computer interface technologies, and even space AI computing, right? And just to give you an example, we have an AI startup innovation factory that runs all year, and there was an Estonian startup that had a very interesting application that can basically tell how much sugar is in your blood based on the sound of your voice using a mobile phone and detecting voice patterns, right?

Now, this could be a game changer for diabetes. I mean, it’s a nasty, you know, global disease. Taking your blood sugar is expensive, inconvenient, sometimes painful. It’s a real pain. Right now it’s still a pilot, but you see the potential for scale. But on the other hand, if it can tell how much sugar is in your blood, what else can it tell about you? How late did you stay up last night? What did you have for dinner? Are you on medication? Did you have too much wine? Are you paying attention? Actually, are you paying attention? So you can see where it goes, right? So you can’t take it for granted that these applications will develop in the right way and will be mindful of a lot of things we were talking about here all week.

Are these solutions, are they safe? Are they secure? Do they have ethics baked in? Do they respect human rights? Are they designed with participation from the global south of the table? Are they sustainable when it comes to energy and all types of things? And one way to, I guess, bake that in could be with standards. It’s not the only solution. But when you look at these fast -emerging governance frameworks, popping up all around the world, of course, you have the EU AI Act, you have different frameworks from around the world. I think one of the tricks is you don’t have a one -size -fits -all, and AI is moving very, very fast. but there are many practical things that can start to be implemented so how do you take these ambitious words and texts and turn them from principles to implementation because the devil is in the details and standards have details so I think we’re at the point where these products, services, companies applications, you know even hardware, all these things need to start to interface and interact interoperably, sorry they need to interact internationally and sometimes internationally as well, you’re going to need standards to basically make these things work and that could be one of the way of baking in all of the common sense things into standards now I know the words lightning speed and standards development are not often used in the same sentence and that’s probably a fair statement but I think in the case of AI for example when the Global Digital Compact launched its call I believe two years ago in the fall It took ITU and its partners less than three weeks to respond to that call for international AI standards coordination by launching the International AI Standards Summit Series.

And actually, the very first one was held in this venue in 2024 as part of WTSA, our Treaty Setting Conference on Standards. And we also launched the International AI Standards Exchange Database, which Doreen mentioned a few minutes ago. But more importantly, when you’re looking at the standards gaps and what people should be working on, we’re working with our partners, ISO and IC, on multimedia content authenticity standards development. That’s a fancy way of saying deepfake detection standards. I’m not saying we’ve solved the puzzle, but there’s a lot of energy and work working with industry, C2PA, different bodies there. I think another major gap, which is not only standards related, is, of course, the skills gap. So when we had our governance day in Geneva last year with ministers from over 100 countries, there’s a lot of things they couldn’t agree on.

But one thing they all agreed on is how to address the AI skills gap and democratize access to skills. globally and that didn’t matter if you were a developing or devolved country and then of course the other was how do you handle the epidemic of deepfakes so I think I’ll pause there thank you but hopefully that gives a kind of picture of how you can go from AI use cases to high potential looking at the dual nature of AI and how standards can be one of the tools to help address those issues. Thanks.

Moderator

Thanks Fred. I mean if that app looks at me right now I think it’s going to tell me that I’m very caffeinated and sleep deprived right but on that point I think standards are obviously the physical manifestation of governance I think we did talk about that that’s very important and Rachel maybe I come back to you I think we do talk about policy tools are important financing mechanisms are important governance approaches because there are many different approaches to AI governance throughout the world how do you see that the participatory whether the governance is actually participatory today some of the frameworks from global north do you think that’s getting imposed on south or south is coming up global south is coming up with their own frameworks how do you see the situation on the ground

Rachel Adams

How do we use it to help advance developmental outcomes or public value? So I think we can see from those kind of three regulatory or governance approaches from EU, China and the US, there’s this very kind of pragmatic adoption of different elements of that within different global south regimes. I know with the African Union’s continental charter on AI, they’re very, very deliberate to include the word regulation. And there was a huge emphasis on human rights and on gender issues and on children’s rights. So I think that what we want is to have maybe less of a focus on global consensus than I think we’re often talking about, partly because interoperability can often mean the dominance of one particular region or worldview’s regulatory regime everywhere else.

And we’ve seen with the GDPR framework. For example, that that has had a limiting effect on the African continent. So I think we rather want to be seeing kind of. a global consensus around a set of principles, accountability, transparency, safety and human oversight and of course a set of standards but noting that different regions are going to need to adapt those standards in different ways. Sometimes those standards might be a kind of gold standard and sometimes they might need to be a minimum standard and we want to be thinking more about the capacity building approaches to try and meet that standard. One of the things we are worried about from a global south and an African perspective is that standard setting processes in the past have always been dominated by those with the time and the resources to really participate in them.

As you said they’re slow and they’re deliberately slow because there’s a lot of expertise we need to bring to the table and once they’re concretized and finalized they become binding. In their own way particularly on the technical side. really want to ensure that as we’re building out these standards, particularly for generative AI and agentic AI, which is still in formation and that is a socio -technical technology and it evolves as it is used in context, but we have representation from Africa, from Latin America, from Asia that is meaningfully included in these standards processes through deliberate funding, through leadership on committees, through co -authorship of these standards. So I think that’s very important to stress.

Moderator

I think that’s an interesting point of view because I’m based in Singapore, so we have 11 countries in the Southeast Asia region and everybody runs at their own pace. And everything we talk about is how do we go from starting point, a lot of it is about where do you start and then talk about where do you end and what is the process along the way. I think that’s what you are. But Brando, I’ll maybe let you respond to some of the points she raised about… the regulatory experience that you have had you have talked to people here obviously you would have talked to other people there is always tension between local adaptation versus harmonization should we have a single set of rules throughout the world what are some of the aspects or highlights that you want to maybe highlight in that sense

Brando Benefi

well first of all on the standards I think it’s a fact that we need to accelerate on that and that we have seen some voluntary delaying I have to be very frank because I look at the implementation of the AI Act where we didn’t need standards so when we decided that some use cases you mentioned social scoring but I can tell you predictive policing emotional recognition in workplaces and study places these are use cases that including if I may also mention manipulative subliminal techniques that are prohibited and they didn’t need standards guidelines on application of these prohibitions were sufficient and we are already implementing that why? other parts of the law for example adequacy of data for training or levels of cyber security that are deemed sufficient these are elements parameters for the high risk use case applications where you need standards otherwise you can’t apply these rules and the standards are being in my view based on the elements I got from those in the standardization process sometimes deliberately delayed because there are some private sector actors that don’t want these standards to be there and so on we need to build mechanisms, I will not delve into that for time reasons, but mechanisms that we are building also in the European context to make it sure that there is a time limit for the standards to be in place because otherwise certain aspects of the governance will not be possible to be implemented.

I want to pick up briefly on also what you said on the risk of AI being used for in fact non -democratic developments, in fact to restrict participation spaces, freedoms. I think this is especially important when we look at fragile, institutionally fragile contexts, which are often countries of the global majority, global south, how you want to call it. we need to be aware that AI can be used for mass surveillance easily for repression of freedoms and to put people under pervasive control even without them fully understanding it I think that we should know that and at the same time I fully share the spirit of the summit concentrate on what we can do for good to mention the summit because a lot of things the example that was just made but yesterday I was meeting with a company from my own country from Italy that is here that deals with systems to anticipate physical status of drivers and to prevent accidents due to physical fatigue, to make it easier to identify earlier this kind of situations that would lead to actually a car accident.

So even in very specific areas, we can find in myriad ways how we can use AI for good. But my point is that enthusiasm for diffusion should not be in substitution for building frameworks that, I insist on my previous point, are precise and not generical ethical appeals, which, to be frank, are not very useful if they are not… pointing to clear deliverables. I want to conclude on this point to be clear that I think an ethical approach is needed. Without ethical approaches, any rule will not be able to function. But if you substitute regulation, governance of all kinds, it can be more binding or more, I would say, co -legislation, co -decision processes. But if you substitute these completely with mere voluntary ethical frameworks, I’m not sure we are getting anywhere.

Especially, I insist, in contexts that might…

Moderator

I think AI for good always starts with AI not for bad. That’s always the starting point and that’s an important consideration. I did promise you guys I’ll leave you on time. So I’ll just have to do very quick two questions. I just need 30 seconds to 60 seconds responses. Fred, I’ll start with you. if you had a billion dollars to accelerate AI diffusion across developing economies where will you start

Fred Werner

I think education skills I think that’s really the starting point actually I was in Johannesburg South Africa for AI for good impact Africa and I had a there’s a lot of conversations about you know using the whole mobile payment revolution of East Africa leapfrogging decades of infrastructure could the same thing be done with AI in Africa I haven’t made up my mind on it yet depending on who you talk to you might be convinced or not I think the opportunities there but also you can’t take it for granted that even if that did happen it would go in the right direction and I think that sort of basic understanding whether it’s for children for diplomats from grade school to grad school that skills gap is massive and I think that would probably be the best spend of money to start there

Moderator

Brandon what will you do with a billion dollars

Brando Benefi

I would say I subscribe to priority because I think that literacy, understanding, build consciousness, building capacity also among civil society actors is extremely important when we see a big acceleration of development of AI as it’s happening around us. Thank you.

Rachel Adams

I completely agree on the digital literacy because I think one of the biggest risks we face, which we haven’t spoken much about, is labour displacement, which I think is going to become significantly more serious. The other thing I would do is invest in building the capacity of our state institutions, of our independent institutions of democracy, our competitions commissions, our gender equality commissions, our human rights commissions, our information regulators. Those are the bodies that will be able to champion the rights of citizens in the face of big tech monopolies.

Moderator

I would have personally bought the shares of all the company CEOs who were here yesterday. but thank you for that. Quick question Rachel while I have you you have spent this week in India, you have seen the entire thing, you have seen the energies around this what is one lesson you learned from India which you think we should deploy globally?

Rachel Adams

I think India has made it very very clear that AI isn’t for everyone I think compared to any of the other summits I’ve been to I think it’s wonderful that there are children from schools here, that we have so many people that are local that have come to the summit and feel included, I think feeling like I am in India at the Indian summit has been the biggest kind of heartening and exciting thing for me.

Moderator

Yeah thanks you have been super inspired to hear the story of how India was able to through a billion plus people create digital ID, financial inclusion, digital payments so there’s a track record of let’s say technology diffusion at scale but in a way that’s beneficial for everyone So that could be a good model for AI diffusion. I know there’s still a long road to go, but if you can do it in India for a billion plus people, I think it should work in smaller places as well. Rando, with whatever is left in your voice now.

Brando Benefi

Well, I think we can learn a lot from what we are seeing here in these days, and I’m convinced that we need to be determined in building more global cooperation. I don’t think we can get the best out of AI diffusion if we abandon the path of building more common understanding and learning from each other. I think this summit can be a moment of this process, but this is something that must happen during all the year.

Moderator

I think, thanks to all of you. My lesson was obviously shake hands with your enemies, even if you are. that’s the only way to do diffusion across the world I would like to thank all the panelists thank you very much and Brando especially with your voice giving away I hope you have a good stay and thanks for enjoying and joining the panel, thank you very much thank you

Doreen Bogdan-Martin

Speech speed

119 words per minute

Speech length

Lessons from India for Global South Diffusion

Topics

Social and economic development | Closing all digital divides | Capacity development

Dr.Panneerselvam Madanagopal

Speech speed

138 words per minute

Speech length

International Standards and Coordination

Topics

Artificial intelligence | Internet governance

Brando Benefi

Speech speed

109 words per minute

Speech length

Lessons from India for Global South Diffusion

Topics

Social and economic development | Closing all digital divides

Rachel Adams

Speech speed

146 words per minute

Speech length

International Standards and Coordination

Topics

Internet governance | Artificial intelligence

Fred Werner

Speech speed

179 words per minute

Speech length

Trust, Ethics, and Governance

Topics

Building confidence and security in the use of ICTs | Artificial intelligence

Moderator

Speech speed

166 words per minute

Speech length

Efficient knowledge sharing

Topics

Capacity development | The enabling environment for digital development

Agreements

Agreement points

Skills development and digital literacy are fundamental prerequisites for successful AI diffusion

Speakers

– Doreen Bogdan-Martin
– Rachel Adams
– Fred Werner
– Brando Benefi

The enabling environment for digital development | Artificial intelligence | The digital economy

Unexpected consensus

Priority of skills development over pure technology deployment

Speakers

– Doreen Bogdan-Martin
– Rachel Adams
– Fred Werner
– Brando Benefi

Arguments

Skills are the fundamental engine of digital agency and must accompany AI diffusion efforts

Two-thirds of South Africans do not have meaningful grasp of AI, creating a significant democratic gap

Investment in education and skills should be the priority for accelerating AI diffusion in developing economies

Digital literacy and capacity building among civil society actors is extremely important during AI acceleration

Explanation

Despite coming from different backgrounds (international organization leader, academic researcher, technical strategist, and policymaker), all speakers converged on the view that human capacity building is more critical than technological infrastructure alone. This consensus is unexpected given their different roles and could indicate a shift from technology-first to human-centered approaches in AI diffusion.

Topics

Capacity development | Artificial intelligence | Human rights and the ethical dimensions of the information society

Need for concrete governance mechanisms rather than voluntary ethical frameworks

Speakers

– Rachel Adams
– Brando Benefi

Arguments

Standards development must include meaningful representation from Africa, Latin America, and Asia through deliberate funding and leadership

Voluntary ethical frameworks alone are insufficient without clear governance and regulatory deliverables

Explanation

The consensus between a Global South researcher and an EU policymaker on the inadequacy of voluntary ethical approaches is unexpected, as it bridges the typical divide between developed and developing country perspectives on AI governance. Both agree that binding mechanisms are necessary.

Topics

Artificial intelligence | The enabling environment for digital development | Human rights and the ethical dimensions of the information society

Overall assessment

Summary

The speakers demonstrated strong consensus on fundamental prerequisites for AI diffusion: skills development, clear governance frameworks, infrastructure needs, and recognition of AI’s dual nature. There was notable agreement across different stakeholder perspectives on prioritizing human capacity over pure technology deployment.

Consensus level

High level of consensus on foundational principles with implications for shifting AI diffusion strategies from technology-first to human-centered approaches. The agreement suggests a maturing understanding of AI implementation challenges that transcends traditional North-South divides and emphasizes the critical importance of inclusive, well-governed AI development.

Differences

Capacity development | The enabling environment for digital development | Human rights and the ethical dimensions of the information society

Unexpected differences

Role of startups vs institutional capacity in AI diffusion

Speakers

– Dr.Panneerselvam Madanagopal
– Rachel Adams

Arguments

Startups serve as AI natives with agility to transform businesses and act as catalysts for change in large enterprises

Investment in state institutions and independent democratic bodies is crucial to champion citizens’ rights against big tech monopolies

Explanation

This disagreement is unexpected because both speakers are addressing AI diffusion in developing economies, but they have fundamentally different views on the primary drivers of change. Dr. Madanagopal sees startups as the key transformation agents, while Rachel emphasizes strengthening state institutions to counter big tech influence. This represents a market-driven vs. regulatory-driven approach to AI governance.

Topics

The digital economy | The enabling environment for digital development | Human rights and the ethical dimensions of the information society

Urgency vs. caution in AI deployment

Speakers

– Fred Werner
– Dr.Panneerselvam Madanagopal

Arguments

AI applications like voice-based blood sugar detection show potential for healthcare but raise privacy concerns

Startups serve as AI natives with agility to transform businesses and act as catalysts for change in large enterprises

Explanation

This unexpected disagreement emerges around the pace of AI deployment. Fred emphasizes the dual nature of AI and the need for careful consideration of risks alongside benefits, while Dr. Madanagopal focuses on the agility and transformation potential of startups without extensively addressing risk mitigation. This reflects different perspectives on balancing innovation speed with safety considerations.

Topics

Artificial intelligence | Building confidence and security in the use of ICTs | The digital economy

Overall assessment

Summary

The main areas of disagreement center around governance approaches (regulatory vs. flexible frameworks), investment priorities (skills vs. institutions), and the role of different actors (startups vs. state institutions) in AI diffusion. While there is broad consensus on the importance of skills development and standards, speakers differ significantly on implementation approaches and priorities.

Disagreement level

Moderate level of disagreement with significant implications for AI diffusion strategy. The disagreements reflect fundamental differences in philosophy between market-driven and regulation-driven approaches, as well as different priorities for addressing AI risks vs. promoting innovation. These disagreements could impact the effectiveness of international cooperation on AI diffusion if not reconciled through dialogue and compromise.

Partial agreements

All speakers agree that skills development and digital literacy are crucial for AI diffusion, but they disagree on the specific approach and priority. Doreen focuses on technical skills and infrastructure, Rachel emphasizes public awareness for democratic participation, Fred prioritizes basic education from grade school to grad school, and Brando specifically highlights civil society capacity building.

Speakers

– Doreen Bogdan-Martin
– Rachel Adams
– Fred Werner
– Brando Benefi

Arguments

Skills are the fundamental engine of digital agency and must accompany AI diffusion efforts

Two-thirds of South Africans do not have meaningful grasp of AI, creating a significant democratic gap

Investment in education and skills should be the priority for accelerating AI diffusion in developing economies

Digital literacy and capacity building among civil society actors is extremely important during AI acceleration

Topics

Capacity development | Artificial intelligence | Human rights and the ethical dimensions of the information society

Both agree on the importance of standards for AI systems, but disagree on the approach. Doreen focuses on technical interoperability and trust-building through existing international frameworks, while Rachel emphasizes the need for inclusive participation from Global South countries in the standards development process itself.

Speakers

– Doreen Bogdan-Martin
– Rachel Adams

Arguments

Standards are essential for AI systems to work effectively together and embed trust

Standards development must include meaningful representation from Africa, Latin America, and Asia through deliberate funding and leadership

Topics

Artificial intelligence | The enabling environment for digital development | Capacity development

The enabling environment for digital development | Artificial intelligence | The digital economy

Takeaways

Key takeaways

AI diffusion requires a three-pillar approach: Solutions (infrastructure and connectivity), Skills (digital literacy and capacity building), and Standards (governance frameworks and interoperability)

One-third of humanity remains offline, creating a fundamental barrier to AI accessibility that must be addressed through basic connectivity infrastructure

Trust and governance frameworks are essential for AI adoption, with clear boundaries needed between regulated high-risk uses and flexible applications

Startups serve as critical transmission mechanisms for AI diffusion, acting as ‘AI natives’ with the agility to bridge technology and business needs

Digital literacy gaps are massive globally – two-thirds of South Africans lack meaningful AI understanding, creating democratic participation challenges

Regional adaptation of AI governance is preferable to one-size-fits-all approaches, allowing flexibility while maintaining common principles

Investment in state institutions and democratic oversight bodies is crucial to protect citizens’ rights in the face of AI deployment

India’s model of technology diffusion at scale (digital ID, financial inclusion, payments) provides a potential blueprint for AI diffusion globally

Resolutions and action items

Launch of the Global South AI Diffusion Playbook as an implementation guide with five dimensions: infrastructure, data and trust, institutions for procurement, and skills and market shaping

ITU’s commitment to continue as a trusted partner in AI solutions, skilling, and standards development

Continued development of the AI Standards Exchange Database with over 850 standards and technical publications

Acceleration of multimedia content authenticity standards development for deepfake detection

Building mechanisms with time limits for standards implementation to prevent deliberate delays

Ongoing work through ITU’s skilling coalition with 70 partners providing 180+ learning resources in 13 languages

Unresolved issues

How to effectively scale AI applications from pilots to full deployment across diverse global contexts

Addressing the fundamental tension between rapid AI development and the slower pace of standards development

Resolving the ‘technology overshoot’ problem where AI capabilities exceed small and medium enterprises’ ability to integrate them

Ensuring meaningful participation of Global South countries in international standards-setting processes

Managing the risk of labor displacement as AI adoption accelerates

Preventing AI from being used for mass surveillance and repression in institutionally fragile contexts

Bridging the massive digital literacy gap while AI development continues at lightning speed

Suggested compromises

Flexible governance approaches that identify high-risk AI areas for regulation while allowing non-regulated use cases to operate under existing legislation

Global consensus on principles (accountability, transparency, safety, human oversight) while allowing regional adaptation of implementation standards

Balancing gold standards with minimum standards based on different regions’ capacity and development levels

Combining voluntary ethical frameworks with binding governance mechanisms rather than relying solely on either approach

The concept of ‘technology overshoot’ provides a framework for understanding why AI adoption faces resistance beyond infrastructure or skills gaps. It suggests the technology has developed faster than organizational capacity to integrate it meaningfully.

Impact

This introduced a new analytical framework that explained implementation challenges in terms of pace mismatch rather than capability deficits. It supported the later discussion about the need for intermediary institutions and ‘AI bridges’ between technology and business needs.

Overall assessment

Explanation

This explores the dual nature of AI applications that have beneficial uses but raise significant privacy questions about what else the technology might reveal about users

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.

Indias AI Leap Policy to Practice with AIP2

Summary

Keypoints

Major Discussion Points:

Overall Purpose:

Overall Tone:

Speakers

The Three Pillars Framework

Startups as AI Transmission Mechanisms

The Democratic Understanding Gap

Governance Beyond Voluntary Ethics

The Dual Nature of AI Applications

Regional Adaptation and Global Standards

Investment Priorities

Learning from India’s Approach

Moving from Moonshots to Implementation

Solutions‑focused connectivity initiatives

Explanation

Evidence

Major discussion point

Topics

Standards complement solutions and embed trust

Explanation

Evidence

Major discussion point

Topics

ITU AI Standards Exchange Database with 850 standards

Explanation

Evidence

Major discussion point

Topics

80 billion commitments to connect hardest‑to‑connect communities

Explanation

Evidence

Major discussion point

Topics

Skills are the engine of digital agency

Explanation

Evidence

Major discussion point

Topics

India demonstrates human‑centered AI and multilingual public services

Explanation

Evidence

Major discussion point

Topics

Startups as platforms for market‑ready AI solutions

Explanation

Evidence

Major discussion point

Topics

Mentorship, market access and funding to build AI talent

Explanation

Evidence

Major discussion point

Topics

India AI mission earmarks ~8000 crore for startup funding

Explanation

Evidence

Major discussion point

Topics

METI Startup Hub links policy to deep‑tech startups

Explanation

Evidence

Major discussion point

Topics

Startups benefit from clear standards for market entry and scaling

Explanation

Evidence

Major discussion point

Topics

EU AI Act provides regulatory clarity for infrastructure rollout

Explanation

Evidence

Major discussion point

Topics

Risks of AI‑driven surveillance and need precise governance

Explanation

Evidence

Major discussion point

80 billion commitments to connect hardest‑to‑connect communities

India AI mission earmarks ~8000 crore for startup funding

Invest $1 billion in digital literacy and civil‑society capacity building

Invest $1 billion to strengthen state institutions, competition, gender and human‑rights bodies