AI for Social Inclusion: Opportunities, Risks, and Policy Priorities (HLPF Side Event)

Good afternoon. Good afternoon, ladies and gentlemen. We are about to start this discussion. Good afternoon, ladies and gentlemen. We are about to start this conversation. Let me, on behalf of the UN Department of Economic and Social Affairs, and the Permanent Mission of the Philippines welcome all of you to this discussion on AI for Social Inclusion: Opportunities, Risks, and Policy Priorities. I want to thank the Permanent Mission of Philippines joining hands with us in focusing on the very topical theme, but it's also extremely important for us because I hope all of you know DESA produces the annual World Social Report, and we want to use conversation today in shaping our report's chapter on AI. So I am giving you a preview of our report which will be released later this year, and the report comes out next year also. Both reports will look at the challenge of inclusion in the age of AI. I think nobody can question that AI offers tremendous opportunities. Are there any dissenters? Raise your hands if you think it doesn't. So we all agree. It has tremendous opportunities for human society, both making firms more productive more responsive public services, better access to health information and diagnostics, expanded access to higher quality education, and not to mention the tasks it's performing in terms of research and analysis for us. But these benefits are not going to happen on their own accord. These are not automatic. It's not a given. Without inclusive policies, AI can and will reinforce biases. It will widen inequalities and deepen existing digital divides. The moat will grow if we do not work in building bridges to address these gaps. This is why social inclusion must be at the center of our approach in shaping AI. In all societies, developing and developed. It's not a North-South issue. North is also facing challenges in terms of exclusion and growing inequalities. So AI policy and social inclusion must go hand in hand. Investments in digital infrastructure must be matched by investments in people, in education, lifelong learning, in digital and AI literacy, in social protection, and institutions that build trust and safeguard rights. But access alone is not enough. What matters is whether people have the skills and opportunities to shape AI, not simply shaped by it. So the task before us is to choose between innovation and inclusion. That's not to choose either/or. It is to shape innovation so that inclusion is built into the way it is developed, governed, and used. That is what we are going to focus on today. Inclusion should not be an afterthought, should not be an a sideshow. It should not be treated as something that you pursue once you have achieved the AI skills. It should be a central objective of our policies. So we look forward to hearing from our distinguished panelists. I will introduce them and I invite them to speak. And of course, from you, from all of you, your thoughts, your reflections that you can share with us. But before I do so, let me invite our co-organizer the Deputy Permanent Representative of the Philippines, Henrique, to share his reflections with us. Over to you.

Thank you. Thank you so much, ASG Hanif, and good afternoon, Excellencies, distinguished guests, colleagues, and of course, our panelists this afternoon. Thank you for taking the time to be with us. It's really a great honor and privilege for the Philippines to co-host this event. With UN DESA, our longstanding partner, and to bring this important conversation really to the center stage of discussions here at the HLPF. We meet at a pivotal juncture. AI is rewiring and rewriting the realities of our economies, our governance structures, and our societies today. The Philippines this year is chair of the Association of Southeast Asian Nations, or ASEAN, and our theme for the year is Navigating Our Future Together. In this context, we are focused on bringing and bridging and building a forward-looking, resilient, and above all, people-centered region. AI increasingly plays an important role in this regard. Experts estimate that AI adoption could contribute up to 10 to 18% of GDP across the ASEAN region by 2030. But harking back to the topic of our side event today, AI for Social Inclusion: Opportunities, Risks, and Policy Priorities, we must also ask the more difficult questions. And these are: Who stands to benefit from this growth? And perhaps more importantly, Who risks being left behind? As ASG Hanif mentioned, opportunities for AI for social inclusion are immense. AI can bridge gaps in public service delivery, modernize climate-resilient agriculture, and revolutionize healthcare, especially in remote island communities, for example. Yet the risks of inaction or unguided development are equally stark. If left unchecked, AI threatens to exacerbate existing digital divides, introduce algorithmic biases that reinforce historical inequities and biases, and expose our most vulnerable populations, particularly women and youth, to unprecedented digital harms, data misuse, and deepfakes. True social inclusion then would have to recognize that technological progress without an ethical anchor can be a dangerous engine for inequality. We here at the UN have been starting to institutionalize and develop guardrails against this. Recent documents and mechanisms such as the Global Digital Compact, the Global Dialogue on AI Global Governance, and the independent scientific panel. All these processes are very important, but in many ways they're only the tip of the iceberg. To make AI more inclusive and a force for social inclusion, much more really needs to be done. In the Philippines, just allow me to share some examples of how we're taking a proactive approach to shape a more inclusive digital future for our people. First, it's through strategic and centered frameworks. We have produced and rolled out our National AI Governance Framework, which is tied to our National AI Strategy Roadmap 2.0. And this framework establishes a human-centered, human rights-based approach to ensure AI adoption safeguards human dignity and data privacy. Second, through institutionalized and equitable research. We, we work to ensure that we are creators and not just consumers of the technology. So the Philippines has established a Center for AI Research, which is tasked with finding local and public good applications for AI, such as weather modeling for disaster response, and ensuring that the products of technology directly benefit rural communities. And third, through grassroots literacy and upskilling. We recognize that connectivity is more productive with enhancing capability of our people. We're training local government information officers and community leaders in the ethical use of AI, transforming AI from from being an elite concept to a practical tool for regional and rural development and empowerment. So distinguished colleagues, we think there should be two priorities: prioritizing digital literacy and upskilling, and also enhancing inter-regional and international cooperation. ASEAN has a guide on AI governance and ethics which provides a strong regional blueprint But we must also align regional frameworks and mechanisms with the global mechanisms and frameworks to ensure coherence and more support for developing and emerging economies. True innovation is not measured by the sophistication of algorithms, but by the breadth of their benefits. A metric for successful AI can be how effectively it uplifts the marginalized, reaches the underserved, and empowers the vulnerable. Thank you once again to UNDESA, especially SG Hanif, and to all our distinguished panelists and speakers, and all of you today in attendance. Thank you for your partnership, and we look forward to an enriching discussion. Thank you.

Thank you— Enrique, for your very thoughtful remarks, I must say. And you also shared how Philippines is grappling with these challenges. So I think this has given me a very nice segue to what the UN is doing. And I must say, UN was a bit ahead of the curve since 2017. In every statement that the SG has made to the GA in September, he highlighted the challenges that AI will pose to human existence. We are just focused on these uses. Let's not forget the uses and weaponization of AI. So these are very serious challenges we are confronted with, but UN has an independent panel which has recently issued a report. UN also organized a dialogue on AI governance in Geneva last week, so I want to share a video on that dialogue to share with you the messages coming out of the dialogue on AI governance. Please show the video.

No future builds itself, and so the choice before us is not between facing AI or fear of it.

It is between governing by design and drifting by default. The global dialogue is about civilian AI, but AI does not respect that line.

The same models and chips have moved into the battlefields. My main concern is with lethal autonomous weapon systems. Let us call them what they are: killer robots.

We refuse to let our platforms become battlefields. It means us as adults, as governments, as policymakers, as businesses to address the sinister uses of AI, such as deepfakes, which disproportionately target women and girls.

In every high-stakes decision, machines can inform, but humans must decide and answer.

So very powerful message from our leadership also, President of the General Assembly and the Secretary-General. The punchline is human beings should remain at the center of these developments. We should steer it to our advantage, not to the disadvantage of humanity. So as I mentioned, a distinguished panel that is going to share their reflections. I want to introduce them briefly. Ms. Shaolian Fu, Professor of Technology and International Development at the University of Oxford and a member of the UN High-Level Advisory Board on Economic and Social Affairs. I had the honor of working with her for some years and it has been a pleasure. Then we have Mr. Vijay Moody, Professor of Mechanical Engineering and of Earth and Environmental Engineering and Director of the Lab for Sustainable Energy Solutions at Columbia University. Thank you for joining us. And Mr. Naveen Gautam, Senior Legal Researcher at the Global Forum of Communities Discriminated on Work and Dissent, and he's also our youth representative in this panel. Thank you. So I will begin this discussion with a broad question, but each panelist getting 1 minute to respond. And then I have a specific question for each panelist. So the overarching question is for all of you, but let me ask, where do you see AI already helping people who have been underserved by existing systems? So it is being inclusive. And where do you see it, it is reinforcing exclusion? And the current disparities in societies. Let me begin with Shaolian. Over to you in a minute to respond.

Thank you. Thanks, Yoon Desu and also ASEAN Secretary for inviting me to this very important meeting. I think the topic is very important and timely. So, I go straight to the question, to answer the question, where do I see the opportunities and where is the kind of risks? First, I give two examples about inclusion. One is about AI in health. Now, AI in health accelerated discovery of drugs. Much, much faster than, you know, the traditional ways. And also, during the pandemic, AI has been used, you know, to screen all the CT scan images and identify patients who affected by COVID and severe diseases. And that helped, you know, accelerate the diagnosis, and also help people and regions who are underserved without sufficient medical professionals. So this is what we experienced during the pandemic. And another example is what I personally participated in an action research, which is using AI to evaluate the potential of ideas created by researchers, startups, by young people. And this is an area which has a significant barrier, and startups and researchers and youth cannot afford to pay high costs to do the valuation of their ideas for technology transfer and commercialization and set up startups. And this AI-enabled tool helped us to, you know, to do this in a much objective way. First is make it democratic rather than investors say it, decide it. Secondly, it's much faster from, you know, 2 to 3 months to 10 minutes, and the cost is less than one-tenth. So many— we have served more than 20,000 startups, and also now we are using it to help the university graduates. We know that youth unemployment is another pressing issue, and many youth have their ideas. Of course, some great ideas need to be, you know, to be recognized and being supported, and some have ideas maybe not that mature and they need some advice. To improve, or some, you know, should be advised not to jump into a not that mature idea. And this is an area underserved, and this AI tool can help the students, help the universities to provide career and employment advices to the students. A lot of students being able to, you know, given the advice and then choose the right direction, choose the right opportunity. So this is the opportunities. And of course, there are gaps. One of the gaps that I have seen is between countries and between regions, the students and the people who do not have access to broadband, even broadband to use AI. And also in our universities, Some richer students use more professional version of ChatGPT, and, you know, students from disadvantaged backgrounds can only use basic GPT. That creates inequalities too, the latter-known students in, you know, backward regions. So these are the examples about the opportunities and the risks. Of course, the implication is clear. I may come later, yeah, and then leave the floor to others.

Thank you. Excellent. Very precise and focused. Thank you so much, ma'am. Mr. Naveen Gautam, over to you, to this overall question where it is serving and where it is excluding. Over to you.

Yeah, is— am I audible? Yeah, thanks. I would kind of focus more on where these things have been kind of excluding altogether because we have been having lot of discussion around why AI is more of a positive aspect. But when you actually look from a lens of a person who has been facing similar forms of discrimination constantly, and how AI accessibility itself is a challenge for me, because once you create a digital divide which got created back during COVID now there is a major digital divide is further being created because of the AI as well, because technical technology actually depends on those who have access to technology. And most marginalized groups, especially work and descent-based discrimination, those who are facing— the indigenous communities as well, persons with disability, and also migrants and refugees— they have been kept away even for access to technology aspect as well. But all of us know that, you know, it all depends on digital literacy. Even you have access to the technology, then it will depend on digital literacy as well, whether you have access to digital literacy or not. So another level comes up when we talk about access to, you know, why AI is kind of— how it can be useful. One major thing which I have learned in the past is that these challenges are, you know, kind of because it's— AI often is not able to differentiate between what exactly marginalization is, what exactly vulnerability is, and what ends up as a discrimination which may be in the mind of one person but may not be a discrimination for other person.

Person.

So if you, if you kind of write about anything on the, uh, and ChatGPT, what, what my co-panelist has also referred to, and other apps, other applications as well, other AI apps as well, there is no specific mention of what caste-based discrimination exactly is. It's very vague. And they say the terminology is being used as maybe facing these forms of discrimination, but for us, we do face these forms of discrimination. That creates a biasness moderator in this. Especially in terms of free recruitment, education, credit, housing, and social protection, there are strong examples of biases which have been created by AI. And it's not a critical assessment, it's just like these are the things where we can always map out and see, you know, where we need to have strong improvement. Just to quote, International Labour Organization has also highlighted that labor market disruption driven by AI is likely to disproportionately disproportionately affect women and young people around Global South. Within that Global South, there is another South in the countries which are facing similar forms of discrimination— kept away from literacy, kept away from digital literacy, kept away from technology as well— which unfortunately is leading to what do we call as digital colonialism, uh, in hands of few people. And other people who are facing work and decent-based discrimination are always facing You know, they are kept behind. And again, I say I'm not kind of criticizing AI. I'm just saying these are the gaps where we need to fulfill, where AI can be very useful if we use it in a proper manner as well. Just to quickly give you some strong examples, I thought I'll just bring it to you. And these are real-life examples of workers. Basically, factory workers in India were actually asked to wear head-mounted cameras. While they are performing the work. And these are most of the unskilled workers so that AI can actually capture what they are doing. And later on, they are replaced by AI itself and they don't have any other option left because that's something— that's the only livelihood options they basically have. And you know what happens? Like many workers were not fully informed that the recordings could be actually used to train commercial AI aspects. So look at the way how we are looking at something which can be improved so powerfully through AI. But on the other hand, the way it has been used is quite exploitative in manner. So these are the few gaps I thought I'll just present so that we can see how do we move forward with this. Thank you.

Thank you. Coming from a young person, what you are saying is it is exacerbating the existing divides and challenges. But I will come back to you. The fact that it's a reality we have to deal with, we also need to look for solutions. I'm afraid we can't simply wish it away. Not that I'm advocating for it, but it's been coming our way faster than we expected. So please think about it when I come back to you. So Vidhi, over to you. This overarching question on exclusions and inclusions, underserved and those excluded, your thoughts.

No, thank you, thank you. So I'm more a field person, not a computer scientist, and work with governments. And I feel AI is just another layer of dependency on large corporations that I see. And it didn't start so much with hardware, but with software, with databases, with large systems. I saw that governments at a were overpromised and highly overcharged for services that initially they were told would be low cost. The recurrent cost and dependency were not disclosed. So I worry that AI will create yet another dependency on what is becoming a monopolistic enterprise. So I just want to be clear where I'm coming from. So now I'll give you some examples, okay? So, you know, lots of satellite data— I mean, engineers started to get collected at a very high resolution, 10-meter resolution, 8 bands, about, I would say, 10 years ago, which was public data. Open data, but these AI companies tried to use it and then sell products, or sell, or tell many emerging developing countries that, hey, we can do this for you, we can do this for you. And interestingly, they had never been to the field to collect any ground truth at all, so they started to sell and train models. As a result, two things happened. First of all, they got it wrong because they didn't even know how to evaluate whether they were right or wrong. They never involved the people in the country to check, right? And thirdly, they only ended up predicting right— I work quite a bit in rural areas, not that urban is not important, just— that's where I work. They ended up focusing on large farmers because that's all they could do, right? So they could do very well 10 hectares, 5 hectares. But the poor don't own 5 and 10 hectares. They own a tenth of a hectare. Government was very keen to help bring energy to where it was needed. Energy infrastructure is expensive, so estimating demand on the field correctly and well can save billions of dollars. So we were asked because of trust we had built in over years doing mundane things, really not AI. Just in this example was Uganda government. They said, you know, Vijay, we— what do you think? And I said, look, there's no ground data. For one-fifth the cost of what some international partners said they were bringing in, the government deployed 100 young people on motorcycles to collect field data across the whole country. 80,000 farmers were interviewed for less than one-fifth of what AI had promised. But now AI can deliver if you have this, right? So I want to highlight the importance that you can exclude vulnerable smallholders. It turned out we were not looking for this when we collected data. It turned out there were women. It turned out they were focusing on horticulture and nutrition. It turned out they were doing the right thing, but the government didn't know that to be able to then help amplify the innovations that the poor themselves were doing, right? So I think AI can be of great value. My My observation was that who owns the training data is actually going to be extremely important, and I think countries should watch out for this. The AI bit of it, their own students can do at the level that these big companies are promising. So I think we have to be at least for some of the work where I'm worried about field infrastructure, you know, how to leverage decentralized technologies. Detecting what people actually want, where livelihoods are at stake, where food and nutrition is at stake, we need cutting-edge tools. But we also need data and that should be owned and collected by the countries themselves.

So, over.

Thank you so much, Vijay. I think the fundamental issue is governance because monopolistic big corporations, top-down, will not serve societies and governance remains the key. Let me come to specific questions for each panelists, but thank you for sharing your initial thoughts, which have been very powerful, I must say, in capturing all dimensions. So, Shaolin, my question to you is, because you have done so much work, how can developing countries embrace AI in ways that support structural transformation and inclusion? Because middle-income countries are struggling for structural transformation. Could AI could help them, but how? How should they use it? Over to you.

Many thanks. I think AI brings opportunities and challenges and the risks. First is the developing countries, the middle-income countries should try their very best, like ASEAN Secretary and Ambassador Garcia has mentioned, try their very best to try to benefit from AI and harness the benefits for Sustainable Development Goals. That's number one. Also, minimize the risks and turn challenges, crisis into opportunities. I have 3 key messages here I want to make. First is, it is important and is totally right to emphasize the risks and also the governance of AI, and also emphasize that, you know, humans should make the decision and the AIs are tools. However, however, I worry that we have underemphasized the use of AI and how developing countries should embrace AI and benefit from it. And why I, I say so is the data shows that the diffusion and adoption rate in Global North and the South already is a big gap there. And in the developing countries, including China and India, the adoption rate according to a recent report is 15% in comparison to 26% in the developed countries. Of course, some countries as high as 60%. So this is including China and India. Exclude China's adoption rate, and then the other part of the developing countries, the low-income countries, is less than 5%. So we have not even used AI, not even benefited from it. So I think the first, you know, action is to really create the capabilities, the infrastructures, the needed regulatory frameworks, you know, and the policies to benefit the good side of AI. So that's one I want to emphasize. Secondly, is about AI. AI is a broad term. There are many different technologies and many different applications. Now, One of the reports suggests AI, you know, the most frontier AI is becoming, you know, learning how to make decisions themselves and they are more risky. However, that's the frontier of AI. But there are many matured AI which we know well. They are responsible, transparent, and inclusive AI like AI for health, AI for education, AI for, you know, climate change. There are not that advanced but appropriate AI, inclusive AI. We should focus on that and relocate resources to use those AI rather than, you know, the companies and the countries compete at that kind of risky frontier AI. So this is kind of differentiate different types of AI and use those inclusive and responsible AI, and then reallocate resources on this, and not in the competition of the risky frontier AI, which we don't know, which is a black box. So I think that's the second point I want to make. The third I want to say is really the developing countries need international, you know, collaboration and the international organizations and the regional organizations should play active role to help the developing countries to build up their skills capabilities, infrastructure, which is very, you know, capital intensive. They need funding and also, you know, governance, which is UN is leading, which is totally correct. And in this regard, ASEAN can play important role to coordinate the ASEAN countries and the African Union and also, you know, Latin America regional organizations can also help the countries and UN, you know, can coordinate global-wise the North-South collaboration. Some of the examples, in addition to the ambassador has introduced ASEAN's AI Center, like in China, there are kind of BRICS country AI center that helps the BRICS countries to build. And also there are some South-South collaboration AI capabilities center. These are the examples that the global community can work together helping the developing countries. Otherwise, because of the infrastructure constraint and also the skills gap, then the developing countries will lack even behind. So I think at this moment, we should, on the one hand, emphasize governance, at the other hand, really work together, help the developing countries to use AI and embrace the positive side, positive force brought by AI. Thanks.

Thank you so much, Shaoliang. So the message is prepare, policy environment, governance, regulatory measures. Then embrace— don't go for advanced AI, but embrace what is available and useful. And then third, cooperation, cooperation, cooperation— regional, global, national— to pursue this. Thank you so much. With you, over to you. You just mentioned you have been in the field. So AI depends on physical foundations, reliable electricity, connectivity, and computing power that many countries still lack. What will it take for developing countries to build these foundations so that AI narrows rather than widens the divide that they're facing right now?

Great question. I have somewhat of a contrarian view on this. See, historically, we really dependent on large interconnected infrastructure for electricity. So we thought we must do that before we do other things. However, The game is changing. So it is possible now to have far more decentralized infrastructure that actually is frequently coming out to be more cost-effective, but more importantly, faster than centralized infrastructure. So it is possible increasingly And I supporting Professor Fu's comment also, there is computing hardware also available at lower cost, right? At the scale that is needed to add that decentralized infrastructure. So, and I also think that in education, in training, To use some of these things, you don't need to build fancy laboratories with scanning tunneling microscopes and nano stuff, this or that. So actually, I see this. If governments take the lead, if we support them to get there, this can be done without as much dependence on the large physical infrastructure that currently, you know, hyperscalers and all these data centers, they talk about here. And I think that in the field, many—

you—

we also need to focus on— yesterday I was at another event where like close to 70, 80 countries spoke, and half of them use the word inclusive. And I think inclusive has so many shades of meaning, but I'll take one is I think that there are hundreds of millions of people where the households are making less than $100 a day— sorry, a month, $100 a month for a whole household, as opposed to $100,000 somewhere else. The questions they are asking are different from the questions that the 100,000-a-year people are asking. So I think inclusive means asking inclusive questions and addressing those inclusive questions. And I feel if we steer the young people towards some of these tools, lessons, they can be developed without huge university expenses and infrastructure that other physical sciences need. So I see that other side as an opportunity to unlock this without having to build all the infrastructure first. And so a little bit of what I'm talking about may be called leapfrogging. So over.

Thank you so much. And I couldn't agree more with you. Vijay, what about human capital?

I am talking about human capital. Sorry, I am talking about human capital. How do we develop human capital without dependence on the large centralized infrastructure? Sorry to interrupt, but I was actually saying that is the opportunity. But I'm saying that opportunity is closer at hand in many countries, because training people in these skills, I feel, will be easier than training in nanotechnology, because nanotechnology depends on all these other ancillary things. And I'm saying that there has been somewhat of that democratization, if I can use the word, using decentralized, smaller, but more pervasive technology. So I can basically provide a megawatt with 3-4 megawatt-hour of storage and a megawatt of solar for less than 10 cents a kilowatt-hour in rural Zambia. And if I do centralize, it's coming out to be 2 times that.

Thank you so much. So we are coming to Mr. Naveen Gautam. Naveen, the question is from your perspective of young people and communities facing discrimination. What would it mean for AI policy to be genuinely inclusive? You had made some opening remarks, but we want specific policy proposals where we make sure public sector has that design to make it inclusive. Over to you.

I think I've criticized AI a lot, but the reality is I have already done this research through AI itself, so that's another harsh reality. Just like computers and internet, when they came, there was so much of discussion happening. So it's going to stay forever, I believe. So we just need to see how do we look at the positive aspects, uh, sometimes as well. But, uh, especially as I have been kind of constantly trying to focus on why there is another level when it comes to those communities which are most marginalized, facing work and decent-based discrimination, as well as those from the indigenous groups as well. So digital literacy has one— has been one major thing which we should always ensure that it is not in hands of a few people. Even within the youth, there is another level where specific set of youth have access to these technology and digital literacy. But for us, for the youth who have been facing these forms of discriminations, they have always been often kept out of this thing because we don't have access to technology. So this is one part. Also, how do we recognize datasets on work and decent-based discrimination, indigenous group? Because I think, uh, co-panelist, uh, Mr. Vijay also very specifically mentioned, how do we go on the field and identify these realities instead of having this in the hands of few corporates? And this corporate also has people who belong to the dominant communities across the Global South. And this is not only about Global South, I repeat, this also happening in Global North. So recognition of datasets specifically on work and decent-based discrimination, indigeneity, migration, refugees— how many refugees, youth, young people who have access to this— is another major aspect. Access to training has already come up a number of times, so I strongly support that thing as well. But, uh, we need to see how the accessibility reaches to the youth, because for me who is from the same community facing similar forms of discrimination. It's also a technical, technological aspect, a challenge coming up, because I also sometimes face issues in terms of accessing AI. But if I look at other youth from the same, on the same group, but not from the same community, also from the dominant group, they also have a strong understanding of AI where they are able to use it very properly. So training is another aspect. Meaningful participation and fair representative datasets. How do we ensure that the AI is not specifically in the hands of a few people who are dominating the market? Instead, youth from the marginalized groups, historically marginalized groups, have access to that, and they're also in the decision-making process. It should not be like, you know, we have some people who are designing AI But the youth who are facing these forms of discrimination don't even have an understanding what's the basis of AI. So how do we bring those youth specifically and ensure that we have a strong understanding and they are into the decision-making process as well? And I think the final thing which would come up is most important thing in terms of the artificial intelligence which we are missing out right now is how do we respect local knowledge? How do we respect indigenous knowledge? How do we respect the knowledge which communities which are facing discrimination based on work and descent— how do we ensure that their knowledge is being embedded very strongly, which automatically doesn't create a biasness when anybody goes online and looks for the AI aspect? And I've already kind of very specifically told what's happening with the workers across India, and most of the workers are from the marginalized groups itself. So these are the few specific recommendations.

Would you like to also comment on the youth unemployment? Because we were to witness recent graduation ceremonies, most of the graduates were questioning AI because they are not finding jobs in the job market. What is the youth intake on that?

I think there are two different versions, uh, I would say, because there is a version which comes from the dominant communities where they are facing these forms of issues and they are not able to get employment. Obviously, sometimes it's AI. There is another version which says many youth are getting employment because with the help of AI as well. That's also another aspect. But from me, from my version, definitely from the communities which we are coming from, the employment is often taken away because of AI from our people, which is a reality. We are being asked to use AI. AI kind of trains itself itself through us. As I gave an example, workers are being asked to put a camera and then they kind of see how the workers are doing it. So unless and until we are part of the process in the decision-making as well as designing the AI, ensuring that we are not being kept away from the AI after AI comes into existence. I mean, that's why I'm saying there are 3 different versions on this thing as well. So that's my—

Go ahead, and we open the floor for questions.

Yeah. So I really want to echo what Rajiv and also Namveen has mentioned. First is, I think, you know, I echo there is opportunity for leapfrogging. And this AI business, although infrastructure is capital intensive, but most important is creativity. People's creativity, which is labor-intensive, and young people and the people in both developed and many developing countries too, their creativity will be the main force, and that creates the opportunity, windows of opportunity for catch-up and leapfrogging. I call it digital windows of opportunity for leapfrogging. Of course, we need basic infrastructure and the young people will be the main force. So where is the opportunity? Focus on the young people, the youth. They will be the main force, you know, in any countries which in the future, if we see leapfrogging, I believe will come from, you know, driven by the youth in their country, of course, supported by government and by by the society. So that is one. So therefore, I think universities, again, should play a very important role in training and skills, not only training the students in the universities, but during the summertime breaks, they could be training centers for people from the society. And our students could be ambassador. They can train many people, you know, in their communities who are not students. So it's 1 N, and they can help to, you know, build up the skills in their communities and in the society. So I think university should play a very important role. Of course, the cost should be supported by the government and the private sector, not all— you know, the universities themselves face constraints. Financial constraints, so— but they have human resources and they have lovely students who have a lot of agencies to do this. And another thing I want to say is, although there are opportunities, there is another under-debated area. We discuss skills, we discuss infrastructure, we discuss governance, However, a lot of inclusive AI innovations, because they are new, they are challenging the existing market structure. They are addressing the gaps. Therefore, they are challenging the existing— some of the existing players. They face a lot of structural, regulatory, institutional barriers. A young student creates some Inclusive AI. Actually, they are challenging the existing players in a given field. They are young, they are new, so actually they face barriers from the existing institutions, regulations, and existing big players who use the old way, who use the old way. So that's one area really for AI to benefit society, not, not only skills and infrastructure there, we need kind of societal regulatory system changes to enable it. So that's area under discussed.

Thank you. Every technology has incumbents and insurgents and there's always a public fight. Insurgents always question and challenge the incumbents. We saw in the '90s also. Naveen, you have a point, then I will open the floor. To get the audience to engage.

No, I think you have already touched upon— I just wanted to ensure and just bring this point to the audiences. You know, in India there are different levels even of education across South Asia as well. So university is one area where, you know, we are talking about how do we ensure that that knowledge also flows to the regional colleges and local college and also the primary schools, because that is the place where we really need to ensure that the communities are having access to that place. So, it is a very strong example how university can capacitate, but that also needs to create an umbrella. It flows down like a water and it actually impacts, the knowledge actually impacts the communities also who are facing the forms of discrimination, because they are the most marginalized and they do not have access to sometimes even the education which is given at the university level as well. So, they are also more dependent on regional colleges, which are pending. So just to add on to what my co-panelists said.

Thank you.

Thank you so much. So one message is clear from three speakers: AI does not stand for automatic inclusion. Inclusion will not happen unless we have policy design, governance, ethical framework, and targeted policy measures to include people in using, in creating, and in deploying it. So let me open the floor. Kindly introduce yourself before you ask this question because we don't have nameplates. I will have to point my fingers on you, so please don't be offended. But I already have 4 speakers, 5th one also. We'll begin with you, please. Please press the button so that he can give you the mic. Good. Yeah.

Hey, my name is Sachidananda Jonabidhananda.

You can call me Sachin. I'm a diasporic representative of communities discriminated based on work and descent.

And I sort of wanted to come in with this perspective and bring in this perspective where our communities are underrepresented in terms of data.

And as we all know, data is the building block of artificial intelligence.

And so we are systemically discriminated against and are not represented. And because AI is the future, basically we're being erased from the future.

And so especially like in the context of India, Mr. Vijay, and the lack of sort of data around Dalits and other marginalized communities, how would you address that?

Because that's like one of the most major issues. Thank you.

We'll come back to the panel after taking a few questions, but a very, very pertinent question. And you, I hope you won't be surprised, the panelists are from India and China, two southern AI countries. Over to you, lady.

Hello.

Yes.

Thank you very much. I'm Cecilia Molle. I'm here on behalf of the International Federation of Business and Professional Women, or BPW International, and I'm here specifically on behalf of our Canadian chapter. So within Canada, we recently launched an AI for All strategy. It's basically a high-level framework to focus on economic growth and enabling business sectors to use AI. However, it's fairly silent on other protection measures. And so you spoke earlier, all of you, about inclusion, and I'm really interested to hear more about how we can ensure that all voices are captured, in particular women and girls of all ages. We know that AI disparately impacts women and girls, and so I'd be curious to know I would love to know your perspectives on how, you know, civil society such as our organization can help influence our government to ensure that those protection measures are captured.

Thank you.

Thank you. Over to you, please. Hello.

Thank you for the amazing discussion today. My name is Henry Mido. I'm here representing the McGill Youth Advisory Delegation from Montreal, Canada. And my question is more focused on the stigma around AI. In my current context as a university student, there's still an overhanging and negative connotation with regards to educational spaces towards the use of AI, where students who do use it, even when it's for ethical and productive reasons or ways, are still being labeled as lazy or incapable, etc. And this can be— can have a lot of damaging effects on the relationship that youth build with AI and can even disincentivize them from using it productively at all. So through education, how can we reduce these negative stigma so that youth are more incentivized to use AI ethically and productively? Thank you.

Thank you so much. Over to you and then gentlemen, please go ahead.

Hello everyone, my name is Samar Thayer. I am a teen delegate with the Global CoLab Network, an organization of teens who meet weekly to help work on the SDGs. So my question to you is that we know that the data in our world is inherently biased because they reflect a biased world. It's all people's opinions, And even if datasets, there's someone recording it, it cannot ever represent the entire world. So if our data is fundamentally unequal, how can we train AI to be unbiased and not build off of current biases and expand them further? Thank you.

Thank you for a very insightful question. Over to you.

Hi everyone, my name is Brian Dong.

I'm from Johannesburg, South Africa, and I'm here on behalf of the Juventus Justice Foundation, which is incredibly proximate to AI Inclusion. We work with young people to— yeah, we work with young people to provide legal education and diversion support so that marginalized youth don't have to go into direct incarceration and are able to act in futures of of their own.

And so, in that vein, I wanted to ask the panel if they think that frontier AIs should play a role in legal systems, and if so, how do we balance the fear of exacerbating inequalities but also expanding legal access?

Thank you. Please go ahead, both of you.

Hi, everyone.

My name's Tavira. I'm representing Civicus. Civicus is a civil society organization based in the Global South. I'm from, originally from Indonesia. My question to you is that you raise a concern about indigenous knowledge and also language inclusion, which I think is often overlooked when we talk about the development of AI technologies. So how would you ensure that in the future when there is a development of AI technologies and when trained AI, we make sure that language inclusions are part of the discussion, including indigenous languages that contain a lot of Indigenous knowledge that will help us tackle climate change, help us with education system. So based on your experience, and also if there is a solution to that, what would it be to make sure that training AI includes language inclusion?

Thank you.

Hello. Sorry.

My name is Nassif Beach. I'm from YAPSF. I'm originally from Jamaica. And my question was, how are we going to aim to build inclusive communities for AI integration without leaving vulnerable communities, already vulnerable communities, more vulnerable to potential misinformation or AI hallucinations as those are running rampant in our current models today?

Thank you so much. Great. So could we begin from this— from here, please? You go ahead and then I'll come to all of you. Thank you.

Hello.

Hi, so my name is Elena and I'm also here on behalf of YSPF. My question is— it's not really a question, but I would love to hear more about your guys' opinions on youth, not youth, but specifically adolescents and younger children who use AI specifically when it comes to education, since there's been a lot of recent research in the past 2 years, specifically 2025 and the current year, 2026, on how younger generations have become— have had a declining rate of cognitive functions and essential skills like critical thinking and problem solving, and there is a direct connection between that with AI. So I'd love to hear just more opinions specifically when it comes to like elementary students and middle school students and how we can ensure that AI is used properly for them since they aren't properly developed either yet. Thank you.

Thank you so much. Please. I think we'll come— the 3 on this side, that's it, and then we'll go to that side. Please go ahead.

Hello, I am Deepak Tawar Jr.

I'm a youth representative and founder for Unity for Prosperity, a nonprofit that focuses on essentially publicizing youth authors in the Philippines and bringing light into stories that are less broadcasted.

Please get to your question, no marketing.

As stated before by some of the panelists, AI learning should target vulnerable populations, as you stated before, and these vulnerable populations are oftentimes not fully representative, especially in the existence of digital literacy. Could you please expand on some of the actions that are actually currently being taken to better this digital literacy, especially in youth groups that receive less attention?

Thank you. To you and then to the lady in the end.

Hello, my name is Advait. I am the founder of GovInform, which in which is— which I'm here for. We work basically to inform youth about ways to get involved in the local community. And in that role, I've gotten very involved with local governments, which leads to my question, which is the big theme for this panel is how we can get governments to respond quickly and effectively in making AI frameworks. But how can we ensure that when there is very often cases where governments cannot respond quickly to very local issues, which should be solved very easily?

Thank you so much. There was a hand there. No? Yes, please go ahead and then we will come around.

Yes. Good afternoon. My name is Nicole Weaver. I'm from the Ministry of Education of Curaçao. Based on some questions that the other people had about education, I wanted to go further about the workplace. AI is helping a lot of people in the workplace, our youth, they come with the background of basic AI in the workplace, that people are happy that they're using AI because it's cost, cost efficient and time efficient. But then the, the youth that goes into the, the workplace don't have the time to get the junior level knowledge that they need to get to senior level. And this is also something that could bring a gap for our youth when they go into the labor market. How do we work on that aspect?

Thank you. The entry-level jobs challenge. Yes, please go ahead.

Good afternoon, everyone. My colleague, Mari, and I, we come representing Incision, an organization formed by surgical professionals, and we have a couple of questions. How can governments ensure that AI reduces rather than reinforces existing inequalities in access to to healthcare and education. AI has enormous potential in healthcare, but many communities still lack access to essential surgical care. How do we balance investment in emerging technologies with investment in strengthening fundamental health systems? And also, today AI has become like extremely important, extremely useful for everyone. We can ask to AI anything, and AI can give us an answer. So, regarding the regulations in AI healthcare, how we can avoid this cannot— AI cannot become a risk for, let's say, low-income communities where people with no credentials, medical— that they say they are medical professionals but they don't have the credentials, using just AI provide health services.

Thank you so much. Last question and then we come to the panelists. Go ahead, please.

Hi everyone, my name is Dayuthi and I just had a question because I currently work on building AI models for women and endometriosis, which is a chronic gynecological condition, which is underrepresented currently. And something I've often faced is this idea or kind of backlash from other individuals regarding the fact that you're using these AI systems within marginalized communities, especially like women, despite the fact that they can reflect existing biases because women are marginalized communities within AI systems and data models. But at the same time, I think increasing the use of AI in women's healthcare can increase the representation of them for future clinical datasets and improve AI models and women's representation. So I'm curious to hear your perspective on this balance and how we can sort of address this without necessarily reflecting women in these existing biases, but also including the representation and increasing the representation within these clinical data sets.

Thank you so much. There are no more questions. Let me come back to the panelists. These have been very useful issues that you have raised. Are very germane to this discussion, so thank you so much. Let me begin with Vijay because—

First of all, if the young people around this room, the questions they are asking were also being asked by the young people in a rural school in Mali, we have actually achieved a lot, right? So I think, I think I just have to say this is phenomenal, the questions you all have. I am myself a university professor, so I know how we got things wrong. See, we We need to work— we allowed corporations to go to the student directly. This is the device. But what I am observing in the marginalized communities, in the low-income settings, in the— where, you know, we are missing the issues of data is coming up, literacy is coming up. Corporations tried going directly to the student. It did not succeed. It succeeded in the wrong way. Instead of targeting young, they were marketing to the young. I think my observation was that building the institutions, the schools, the clinics, the teachers, is the building up capability at that level, right? So I think that's critical because that is what the young can be— are learning from. They are getting literate from those people. So I think we need to rapidly reach that layer, the education, health system providers who can— today the good news is the decentralization that I talked about in technology can allow them to really, just like we did science experiments in school of some, you know, put a candle blows out if you cover the lid and all that. But there is so much that can be done today with digital tools that is school can teach their own young people. So I think, and you know, you raised the question of human capital. I think it's central and I think that we should— so we should focus on that level, otherwise the kind of question, you know, Elena asked about, you know, really addiction, really, so that, you know, we have to bring Young minds can be a tremendous positive energy, right, if the right tools are there. So I have to say that we have to fundamentally change our curriculums because it's such a massive force, just like physics and, you know, chemistry was a massive force of the last 100, 200 years. And I think empowering the teachers, empowering the, that layer of institutions. And lastly, I want to end by saying I saw so many technologies becoming white elephants, and I think if AI can be used to maintain, diagnose, operate, because we have thrown capital only at problems, and when I see capital Things fail and nobody knows. So I think we have to be careful, and I think AI itself can help us if we structure it right to diagnose what's not working and, and kind of dynamically fix it. We need to make that as part of the DNA. And final point is the digital literacy, very important, but I think we rapidly need to do that literacy for the decision makers in the governments, because they are being approached by corporations. If they who are representing the population cannot make the decisions in an informed way, we also have challenges.

Thank you. Thank you so much, Vijay. Remember in the late '90s, we underestimated the risks of social media, right? It was all upside, that's all I heard. And look where we are. So I think we cannot underestimate, but I'll come in when I come to summarize this discussion. Shulin, over to you.

Thank you. And first I want to thank all participants for the thoughtful questions, yeah, very good questions. I will focus on two 3 areas that I have firsthand experience. First is about, you know, the question comment about whether, you know, target to use more, you know, healthcare, AI-enhanced healthcare in the women in the society, and then increase its representation in the dataset. Definitely, it's a very good, you know, proposal. An idea. And that can help us to have more accurate, you see, more close to true, you know, populations analysis and algorithms. The same for the Dalit questions in India, need to, you know, provide some targeted programs to enhance their representation in the data. And then come to the worry about In the data, because of the structural system, certain communities, including women, are underrepresented or discriminated. Whether AI will definitely reinforce this discrimination? Let me tell a story. I don't tell you the result. I'm training the data looking at all the startups, including female founders and the male founders, in raising the funds. What the results show, the analysis results show that startups funded by women, they receive less funding given other factors the same. So women were discriminated in this capital market. And this is like I started OxValue and my colleague told me, "Oh, you are sailing against the wind in every aspect." You know, you are not— you are a woman, you come from ethnic minority background, and you are not young, you are not from STEM, etc. So we see this result. The coefficient for women is negative and significant. What we decided is we leave out this gender variable in our algorithm. So we don't ask the founder, their gender. So this becomes blind. So this negative, you know, coefficient— this variable factor was removed from our algorithm. So that is one way to reduce the systematic discrimination kind of embedded in the dataset. So I think that's one example. And also we can see, you know, companies funded by winning, they are more responsible. You see, the survival rate would be higher. And we keep this factor in the algorithm, which is a positive discrimination, you see. So that's— so AI, that also answers the question whether we can train AI to be inclusive. So this is something that, you know, in the analysis, when we see and our judgments can come in to train AI to be inclusive and remove those discriminated factors from the algorithm. So that's one example. So there are ways to, you know, to make the AI to be more inclusive and respectful. Responsible, at least to be equal, you know, equal to men and women. And another very interesting question is about stigma of using AI in learning and education. Very good question, interesting, because as a professor in the universities, we keep on discussing this. And of course, in Oxford is 900 years university, but it is the first university in the UK introduced a guidance on using AI in teaching and using AI in research, give clear guidance. So certain areas can be used and certain areas not allowed because we want to train our students key capabilities, key skills, which are the key skills For you, for young people, for your whole life, ability to learn. AI cannot replace our learning capability. We need to have this capability to learn no matter technology, how, where it goes, you see. And we have this learning capability. Secondly, critical thinking. AI doesn't have this critical thinking capability. So this is something, you know, for us to know Whatever answers created by AI or even by published papers, we know its strength, its limitations, when it works, when it doesn't work. This critical thinking is also very important. And so to, you know, we know distinguish, we know what is good. So these are the capabilities. That's why in teaching we need to reform. The teachers need to reform the way how we teach, what we teach, and even examinations, what kind of questions to answer, to ask, what we want to see from students' answer, what kind of capability skills we want to see from answers. So a lot of reforms the universities need to do, yeah. But I think the right approach, we know we need to equip our students with the skills to use AI, to use AI, to master AI. And finally, I think AI, there are different skills. So in the society, including the question about labor force skills, AI skills, I think, also distinguish, go deeper. There are skills to use it and the skills to create it. The skills to create AI and app may be more sophisticated, but the skills of to use AI, we don't need to be frightened by AI. So the success for AI, they need to think about the user interface to be friendly to users, to general users. If AI is very difficult to use, that AI app wouldn't be successful.. It needs to be very friendly. Like, I studied a platform, you know, which is a short video platform used a lot by farmers, migrants, you know, in the poor regions in China. Why it's different from TikTok? Because its user interface— the founder comes from the poorest region of China, so he designed the user interface very friendly to people who is even illiterate. So that user interface should be friendly and therefore that should reduce the barrier for people, you know, to use AI. So that's a problem. Digital literacy definitely is an area we need to address, but shouldn't be worried too much because it's too complicated. AI won't be successful, won't be popular. You see. So yeah, that's all from me. Yeah, thank you.

Thank you so much, Naveen. Over to you.

No, there were so many questions. I wish I had an AI thing which I could kind of summarize, which I generally do. That's where AI is kind of making you a slave, which is a reality but also very useful. So just to respond to your question about the biasness of being lazy because somebody uses an AI, which is true. I mean, which happens with me as well, because sometimes, you know, I also, you know, I'm not very comfortable in researching on something, so I kind of put it on ChatGPT or Claude or so many apps, and then it gives you some data. So that's a reality. And we— I mean, there will be so many stigmas attached to that aspect, and we cannot run away. People will keep calling you by names and all those things. But you have to consider AI as it's something which is a technology, not something which is impacting your personality. Because same thing happened when internet came, basically, right? Most of us didn't have information, but through internet we started getting information. But there were people who used to call them, oh, you got this information through internet anyway, so, you know, uh, so what about those people who don't have access to information? And all those things. Same is with what— when social media came, we never realized that, you know, there will be a revolution of social media which is happening now. So we have to live with the technology. Similar is with the AI thing. Second is we are not training AI. If you look at the broader aspect, AI is actually training us to train it, and which is a reality. And we are kind of more worried now that we have to put things into AI and that, that would kind of move away from biasness. No, no, you put a, put, put something about some one random biasness, AI will come back again with another biasness.. So technically what's happening, it's vice versa. AI is also training us indirectly. So that's one major fact which we need to take care. So we don't need to worry about people commenting and all those things. It's a technology, we have to be ready, and, uh, definitely it's going to stay forever. Uh, that's one thing. I don't think I need to respond so many questions, but I will respond to your part on indigenous. Same thing, what we— when we talk about movements or indigenous movement, caste movement, women movement, gender justice movement, it has to come through people who are from the community. Unless and until you have people from the community, you won't be able to have that perspective into any application. So if you need to ensure that AI has all the perspective, it needs to include everyone. And if you want to have that strong indigenous aspect and perspective within the AI, anything you're designing, you need to ensure that it is coming from the youth who are belonging to the Indigenous group as well as other groups which are facing similar forms of discrimination. But the strength, the power has to have— has to be in the hands of youth for decision-making and other aspects, because now AI often is being designed by the youth, it's impacting the life of the youth, and also it will keep impacting the life in the future of the world, and the youth will be at the center of AI as well. So— that has to be us only. That's all. Thank you very much.

Thank you so much. Let me— we should wind up because we are already 5 minutes beyond time. Ambassador, would you like to share any thoughts before I try to summarize quickly?

Thank you so much, ASG, and thank you to our dear panelists, and especially thank you to the youth representatives. I think hearing your voices makes this conversation all the more richer and tells us here at the UN just how much work needs to be done, but not in isolation, but with you, because the questions you asked, I think, embed also the answers, because eventually it'll be you who will need to find those answers or create better answers or help design those better answers. We don't have all the answers,, right? We have to work together. And this new world, this brave new world we live in, this digital world creates so many questions that we all have to— ironically, we can't answer the answers that AI is bringing to us, right? We have to also solve them in our own capacity as human beings. And this, I think, is perhaps one of the biggest challenges us as human beings face now. So I think just to thank you all, and it's been a wonderful and very enriching conversation.

Thank you so much. And I must commend, first of all, our very distinguished panelists and all of you. I must say, whenever I go into rooms talking about AI, I find it both stimulating and sobering discussion. It stimulates me with the opportunities, but it is very sobering too. And I want to flag 3 things for you to keep in mind. I don't— and I could be wrong— probably— I don't think human beings have ever invented a technology which takes its own decisions, and we are heading in that direction. Nuclear weapons were devastating, but they could not be fired by themselves. So this is a big distinction, and you might have heard they were trying to retire a model and model reacted. It refused to leave that database. So this is the technology which has its own mind, And the advance— Hawley— Shaulian mentioned there is a trajectory of AI which is about advanced AI. If we get there, that's very high risk. So this is unique about this technology. Sobering, stimulating. Second point is human beings can still take charge. And let's not forget, remember cloning? Cloning was a technology with a lot of promise., but human beings decided, "No, we will not use it. It's not good for us." So human beings can still take charge on regulation, policy formulation, human capability, labor market governance. All of these areas are in our hands and we should shape them, not the corporations. And that's where your role is critical to sensitize your governments, Please play the role that you're supposed to play as public officials to make societies inclusive, equitable, and of course progressive. They should progress also. Lastly, keyword I think we heard from all panelists, cooperation at all levels. There are no solutions that you can pursue alone. Please work with your communities, with your government, within your country and across countries through the United Nations, and we are here, UN is here as Ambassador mentioned, determined to make sure it serves humanity and it's not the other way around. So a round of applause for our panelists. One for you also, for our participants too. Thank you so much. My sincerest thanks to our DESA team which brought us together and the team from the Mission of the Philippines. Excellent work. Thank you so much.