Artificial intelligence and the Future of Counter-Terrorism: Reshaping Institutions and Operational Realities (Counter-Terrorism Week Side Event)

Good morning, everybody, and welcome. Thanks for bearing with us while we were waiting for everybody to arrive. My name is Oren McCarthy. I am the liaison officer for the United Nations Interregional Crime and Justice Research Institute here in New York, and I have the pleasure of co-moderating today's session on artificial intelligence and the future of counterterrorism, reshaping institutions and operational realities. I'll be co-moderating together with my colleague Akvila Oginotiene. From UNOCT. Just a quick couple of opening, kind of to set the scene and housekeeping points. First of all, we are web streaming on UN Web TV, so to all following online, thank you for joining us. We just note that this event is the follow-up of an event that we did in December, which was co-organized by UNICRI, UNOCT, the Permanent Mission of India, and the Permanent Mission of UAE. I think it's a testament to the issue here we're here to talk about, to see that the member states on board in this event are growing. We now have the permanent mission of India, permanent mission of UAE, permanent mission of Japan, and the European Union. And so, as I said, it's a testament, I think, to the importance of the conversation that's been started around this issue. And so with that, I will now hand over to our distinguished opening speakers. And so I have the pleasure to introduce His Excellency Mr. Harish Prava Taneni, who is the Ambassador and Permanent Representative of India to the United Nations. Ambassador, over to you.

Thank you, dear Moderator, Excellencies, dear colleagues, all protocols observed. It gives me great pleasure to join you today. It was only in December that we came together here at the UN in this format to discuss building operational capacities for the use of AI in counter-terrorism. 6 months is a relatively short span of time in traditional domain. However, in the AI space, 6 months is a lifetime. In this period, we have expanded our partnerships and collaboration to include Japan, the EU, and other partners, while retaining UAE and UNICRI, who we work closely with, along with the UNOCT. So this is for us a a positive development of the growing importance of this subject in the UN. While I can use the time allotted here to talk about our co-organizers who are firmly committed to counterterrorism as much as we are, especially in the AI domain, um, I want to do something different in my opening remarks. I just asked AI what happened in the AI and CT space in the last 6 months. And the answer was of particular note, and I'm going to read it out verbatim. It says, AI has not yet revolutionized counter-terrorism in the last 6 months, but it has sharply changed the operating assumptions. CT agencies are now treating AI both as a threat multiplier for terrorists and as a capability multiplier for states. It further listed some specific use cases, and they are as follows: One, the threat picture has shifted from online propaganda to AI-enabled operational ecosystems. Generative AI can help extremists produce propaganda at scale, tailor recruitment material, generate misinformation and deepfakes, translate content across languages, and accelerate radicalization online. Second, counter-terrorism is becoming more predictive and data fusion-driven AI is changing CT from a largely human-led intelligence cycle into a machine-assisted pattern detection cycle. Third, online CT is moving towards AI-assisted moderation, but with more caution. AI moderation is still bad at context. Fourth, AI has pushed CT into urban security border management, and critical infrastructure space. AI now is embedded in conversations on facial recognition, biometrics, border screening, public space surveillance, predictive policing, major event security, and critical infrastructure protection. So friends, it is safe to probably say that AI is on its way to transform CT from a contest over networks and narratives into a contest over data, speed, trust, and governance. I'm certain that the implications are only going to increase, and this is going to grow difficult for the counterterror agencies as terrorists adopt, morph, and disrupt beyond propaganda and recruitment. Colleagues, it is unfortunate that the 9th review of the GCTS ongoing, going through probably the last-minute round of amendments, proposals, and voting right now as we speak, That explains why my expert isn't here. He's sitting there looking at the amendments and the voting. Uh, this 9th review of the GCTS could not agree on language on this topic, and the strategy will have to be most probably a technical rollover. So for this topic impacts us all equally, and the GCTS was a perfect place to highlight a joint resolve via the Delhi Declaration on countering the use of new and emerging technologies for terrorist purposes The CT Committee of the UN Security Council recognized that innovations in technology may offer significant CT opportunities. In this regard, it is pleasing that its pillars have been taken forward by our friends from UAE and Algeria, and India will try to operationalize the third pillar, hopefully, when we enter the Security Council next. However, given the pace of changes in this technology, we will cross that bridge when it comes up. In closing, let me once again reiterate my Prime Minister's word of caution that in today's interconnected world, security cannot be an afterthought. It has to be at the core of our decision-making. I wish you all fruitful deliberations. Thank you very much.

Thank you very much, Ambassador. Very appreciative of the support to UNIKO and UNOCT on this issue, and I can only fully agree in terms of the technology really not being an emerging technology anymore, being an operational reality. So thank you once again. I have the pleasure now of introducing His Excellency Mr. Minami Hiroyuki, Ambassador in Charge of International Cooperation for Countering Terrorism and International Organized Crime. Ambassador, over to you.

Thank you, Chair. First of all, I'd like to express my sincere appreciation to the governments of India, the United Arab Emirates, the European Union, as well as the United Nations Office of on Counterterrorism and by United Nations Interregional Crime and Justice Research Institute for organizing this timely and important side event. It is a great pleasure for Japan to co-host today's discussion today together with all of you. 3 years ago, during the 2023 United Nations Counterterrorism Week, Japan co-hosted a side event on generative AI together with Australia and the European Union. At that time, we warned that the malicious use of generative AI by terrorists for propaganda, recruitment, and facilitation was no longer a distant possibility. At that time, it was emerging a reality. Today, rather than repeating that warning, I would like to share some of the practical initiatives that Japan have been advancing to address this challenge. The Hiroshima AI process launched under Japan's G7 presidency in 2023 was a pioneering framework for establishing international guiding principles for trustworthy AI. It demonstrated the importance of promoting innovation and managing risks at the same time. More recently, at a G7 Asean Summit, Prime Minister Takeuchi reaffirmed the same principle, the same logic applied to counterterrorism. We must harness the enormous potential of AI to strengthen public safety. At the same time, we must prevent terrorists and violent extremists from exploiting exploiting this technology.

Advancing—

achieving both objectives together is one of the most important policy challenges facing the international community. A resilient international foundation begins with sound domestic governance in each country. Earlier this year, Japan welcomed a visit by the CTIG members to Japan. During its assessment, CETA highly appreciated Japan's centralized policy coordination through the Headquarters for AI Strategy, as well as our advanced national AI governance framework based on ethical principles and responsible government— governance, sorry. CETA also encouraged Japan to continue Sharing its experience and good practices in the use of AI for counterterrorism with the international community, Japan stands ready to do so. Building on these efforts, Japan has been promoting a practical project to address terrorist threats involving the misuse of AI and other new technologies. First, Japan supports UNOCT's project on gaming and preventing violent extremism. This project brings together young game moderators from Southeast Asia for live simulation exercises. Through these activities, our participants strengthen their ability to detect signs of extremist activities and hate speech in online gaming space. They have also developed a youth-related online safety handbook, helping to build digital resilience from the ground up. Second, Japan supports UNOCT's project to strengthen OSINT capabilities for investigation into terrorist financing involving cryptoassets. Through hands-on training using blockchain analytical tools, investigators from Southeast Asia enhanced their practical ability to trace and seize terrorist funds transferred through highly anonymous digital channels. Japan also contributed to Project Thins WAVE+ by Interpol, which addresses emerging threats such as deepfakes. In addition, Japan supports capacity-building workshops implemented through the International Institute for Justice and the Rule of Law , helping criminal justice practitioners strengthen their responses to the malicious use of AI by terrorists and violent extremism. Through these efforts, Japan is contributing to a stronger and more practical governance against emerging threats in the age of AI. In the field of AI, governments alone do not have all the answers. Later in today's program, Japanese AI expert will speak about the latest trends in generative AI and as well as pioneering efforts toward trustworthy implementation in society. Building on the knowledge gained through the Hiroshima AI process, Japan will further strengthen cooperation with industry, researchers, international organizations, and other partners. Such a public-private and multi-stakeholder cooperation is essential if we are to keep pace with rapid technological change. This thing suggests AI is a technology that can be used by terrorists also. It can also be used by law enforcement agencies also to protect people from terrorism. The decisive question is not the technology itself. The decisive question is who uses it and for what purpose. Guided by principles of trustworthy AI, Japan will continue to work closely with the United Nations and all member states. Our goal is clear: AI must be— must not become a weapon for terrorists. It should become a force that protects people and strengthens public safety. To this end, Japan will steadily advance law enforcement capabilities, advancing public-private partnerships, sharing practical knowledge and experience, and deepening international cooperation. In doing so, we believe that innovation and security must always go hand in hand with respect for international law and human rights. Thank you for your attention.

Thank you very much, Ambassador, um, and thank you for the support and partnership. As you mentioned, going back to the 2023, um, CT Week, it's great to be partnering once again with Japan on another AI and counterterrorism-related topic. And indeed, also thank you for the, the support to, I guess, the work that UNOCT and UNICRI started in the space of AI and counterterrorism several years ago. Contributions from Japan was, was one of the kickstarters in many ways for, for our work in the space. So with that, I will now hand the floor to His Excellency Mr. Maqsoud Khruz, who is the Envoy of the Minister of Foreign Affairs for Countering Extremism and Terrorism of the United Arab Emirates. Excellency, over to you.

Excellencies, distinguished delegates, I would like to express my utmost honor to welcome you to this discussion on artificial intelligence and the future of counter-terrorism. I would like to extend our highest appreciation and sincerest gratitude to India, Japan, the European Union, UNOCT, and UNICRI for their distinguished partnership and excellent collaboration and their partnership in convening this event. AI is no longer a future consideration for counter-terrorism institutions. It is already reshaping how they operate. AI can strengthen our ability to process large volumes of data, identify patterns, support decision-making, and allocate resources more effectively. At the same time, extremist and terrorist actors are increasingly exploring widely available AI tools for propaganda, recruitment, and operational support. This raises serious concerns about the potential future impact of AI on extremist activities and terrorist capabilities. What makes AI particularly significant is the speed at which these capabilities are becoming accessible scalable, and globally available. Recent events in our region have underscored the urgency of this discussion. The UAE has also faced unlawful attacks by armed groups operating across borders. These incidences demonstrate how terrorist actors can exploit rapidly advancing technologies to project violence across borders, threaten civilians and critical infrastructure, and undermine regional security and stability. They also underline the need to prepare for the ways in which artificial intelligence could further amplify existing risks. As AI becomes more accessible, we must ensure that it is not exploited to evade detection and support attacks. This requires us to disrupt the financing, supply, and technical networks that enable extremist and terrorist groups to acquire and deploy such capabilities. The central question is whether our institutions can evolve quickly enough to harness its benefits while managing its risks. We must ensure that innovation strengthen security, prevention, and resilience rather than creating new vulnerabilities. Yet technology alone is not enough. Technological capability must be matched by institutional readiness. New technologies can spread globally in months, while institutions and operational systems often take years to adjust. For many member states, the challenge is translating awareness into operational capability. The next phase must be implementation through clear legal and regulatory frameworks, defined institutional responsibilities, detection and response capabilities, information sharing, and practical measures to prevent extremist and terrorist groups from acquiring dual-use components, expertise, and financial support. The UAE has sought to address this challenge through investments in artificial intelligence, cybersecurity, and digital resilience because it has been established that technological innovation and security preparedness must advance together. The position of the UAE is that the benefits of AI must be broadly shared. In this spirit, the UAE recently announced a $1 billion AI for Development initiative to support AI infrastructure, skills development, and AI-enabled services across Africa. This demonstrates that strengthening national capacities and expanding access to emerging technologies are essential to ensuring that no country is left behind. We must move from discussions to implementation. No country can address these challenges on its own. Effective responses require cooperation across governments, the UN system, industry, academia, think tanks, and civil society. The UN has an important role. To play in supporting member states, strengthening coordination, and advancing capacity building, while ensuring that the use of AI in counterterrorism remains grounded in international law. The UAE remains committed to working with all partners to ensure that artificial intelligence strengthens prevention, resilience, and security rather than being exploited to deep existing threats to civilians, critical infrastructure, regional stability, and international peace and security.

Thank you very much, Excellency. We appreciate the support and partnering on this event once again and continuing the partnership we started in December. So, and we're of course looking forward to hearing more from the insights from the UAE shortly. But before that, we have one more speaker. So we— I will now pass the floor to Ms. Anna Stravska, Special Envoy for Counterterrorism, the European External Action Services for the European Union. Ms. Stravska, over to you.

Thank you. Excellencies, distinguished colleagues, it's a pleasure to be here. And on behalf of the European Union, I would like to welcome you all as one of the hosts of this meeting. My thanks go to to my fellow co-hosts, His Excellency Ambassador Parvat Khethani from India, His Excellency Ambassador Minami from Japan, and His Excellency Special Envoy Cruz from UEA, for the partnership that has brought us together, and also to our colleagues from UNICRI and UNOCT, whose work underpins so much of what we will discuss today. My fellow co-hosts have already spoken to what is at stake. Let me offer the EU perspective. In counter-terrorism, artificial intelligence is a double-edged sword. The same tools that help our law enforcement and security authorities to make sense of large amounts of data and act earlier are being turned against us. Terrorists and violent extremist actors are exploiting widely available technologies to produce propaganda, recruit, and to plan and conceal their operations. Europol is the EU agency— warned us early in 2024 that supporters of Islamic State have started using generative artificial intelligence to create propaganda. As one example, a poster depicting armed fighters and featuring the slogan "Join the Islamic State" was disseminated online. Another AI-generated image showed a kitten wearing an explosive belt accompanied by a message glorifying martyrdom. This shows how, thanks to generative AI, a single individual with basically no background in technical skills can generate in minutes what once required an entire propaganda cell. This also shows how extremist propaganda is adapting and using technology to produce absurd or emotionally arresting imagery to attract attention online. And this is the reality that our services are confronted with today. The EU's response to this development rests on a principle we come back to again and again, that innovation must be balanced with responsibility. These technologies are too powerful to set aside— and too consequential to deploy without rules. Strong safeguards are what allow, allow us to use these tools lawfully and to keep the trust of the public we serve. This is the reasoning behind the European Union's AI Act. It established the world's first comprehensive framework for artificial intelligence. It takes a risk-based approach grounded in fundamental rights and is now being brought into implementation step by step. At the international level, we have paired it with the Council of Europe's Framework Convention on Artificial Intelligence, the first legally binding treaty in this field. The point is not to regulate for its own sake, but to set a human-centric standard that others can build on. However, frameworks alone are not enough. The harder task is to turn this principle into real capability, to ensure that agencies in Europe and also in other regions have the tools and safeguards to use these technologies well. This is where the European Union has chosen to invest, both inside the EU but also externally with trusted partners. Through our programs, City Tech+ and AE-POL initiative, which we found and deliver together with UNICRI, UNOCT, and Interpol, we are helping law enforcement and counter-terrorism professionals to build precisely that capacity, with respect for fundamental rights built in from the outset. And we do this because security cannot belong only to those who happen to have the most advanced tools. The digital divide is real, and a capability gap in one part of the world is a vulnerability for all of us. So we will keep working through partnerships with the private sector and civil society and through the exchange of good practices across borders so that more countries can harness these technologies and defend against their misuse. And I would like to add another example within the framework of the Global Counterterrorism Forum, which is a forum which EU co-chairs together with Egypt. We will now propose launching a dedicated grant on new technologies. And a lot has been already done by the GCTF-inspired institutions that are around this table, starting with HEDAI, also IIJ. So thank you for that. And you can count on us that you will remain a committed partner to handle these challenges. Thank you.

Thank you very much, Ms. Drevska. And of course, thank you to the European Union for the support to both the AI-POL initiative and the CTTF. Initiative. Um, as you mentioned, a lot of the content that will follow will be built on the backbones of that. Um, so we're delighted to be here to present, um, and share insights from both of these initiatives momentarily. Um, so with that, uh, we wrap up the opening remarks. So thank you, uh, on behalf of everybody to the distinguished representatives and speakers. Um, and we will now move into the next session, which is focused on keeping pace with innovation, the state of AI and counter-terrorism. And I have the honor and pleasure of handing the floor to His Excellency Dr. Mohammed Al-Kuwaiti, who is the head of the United Arab Emirates Cybersecurity Council, and we're delighted to have you back again with us once more. Thank you.

Thank you very much for this.

Again, honor to be with you all, and thank you for those great messages and points that you already been pointing out. I have a few slides, if you allow me, in order to actually go over them with everybody here, and actually it articulates more into the essence of using AI into that terrorism aspects. Allow me first to start by giving an overall look of our digital cities that we live in UAE, and based on many of the goals and aspects that we see. I don't know if I have clicker or someone can help me with that.

Okay.

The goal of our leadership is to definitely build that tolerance, build that, again, collaborative efforts with everybody. That's what we saw. If we move to the next slide that focuses on our smart cities. UAE have placed globally into so many of those competitiveness rankings of using digitizations and using artificial intelligence, as well as using many of the cyber aspects, and we saw that based on many of, be it the ITU Global Cyber Index or many of the aspects that we, as a matter of fact, experience. We have turned so many cities into smart cities. Those smart cities depend on critical infrastructures that actually depends on technology, be it in the healthcare, be it in the energy, be it in the renewable energies, be it in many of the educational aspects, aviation, you name it, any of those Today's depends on those technologies, and this is where the fragile aspects comes, and this is what, as a matter of fact, we've seen in many of the threats that we face from those technologies. And those threats, we categorize it into those three aspects. One is cybercrime, second is cyberterrorism, and the third is cyberwarfare. And yes, we face them all. As a matter of fact, lately, We faced a combination of many of those threats, as a matter of fact, in that perspective. But one of the most and most critical things is the cyberterrorism. It's not only the misinformation and disinformation that we see, it's not only targeting critical infrastructures that actually really represent the essence of our life. It's not only, as a matter of fact, changing many of the phishing or deepfake that we've seen, It's even more than that. It's actually manipulating the mindset of so many of the aspects that we've seen. It's going into the people's minds and as well as soul in some of those aspects. And this is the root behind that is definitely technology is not perfect. You will always find a zero days. You will always found outdated system. You will always found misconfiguration. And this is where the people comes in this. Handy. They are the weakest link, and we need to make them the strongest link, as a matter of fact. And this is where comes the awareness, this comes the partnership. That changing nature of terrorism really represented by many of those aspects. AI have facilitated that. We saw many of those predictive analysis intelligence, we saw algorithm-driven targeting, we So warm GPTs, as well as cyber-enabled terrorism, deepfake, many of those aspects, as a matter of fact, in the hand of those, again, terrorist groups and make their job easier, not only in variety, on volume, but even in velocity. And this were the key changes that we actually experience in many of those aspects. We heard yesterday about using some of the secure communication for any of those radicalized groups to actually communicate among themselves. We saw many of the privacy or ethics that actually touches upon our community, operational capabilities that we need to make sure that we close those gaps. And this was actually presented in the next slide. If— this is where we need actually a force multiplier to counter many of those terrorism. Acts using technology. Definitely, we use AI to counter many of those AIs. Definitely, we need the people who are actually really enabled and really use that AI in the best way possible, be it intelligent analysis, early threat detection, operational support, and many of those tools that are available out there, but it all depends on the people and how they use it. We saw it in the offensive perspective in the cyberattacks. We saw it in the defensive perspective. Yes, we use it positively in order to defend. In order to actually predict the behavioral analysis of many of those networks in order to self-healing and self-learning algorithm that we use it to early detect any of those attacks. Even in the deepfake, there is so many tools that allow you to do that. But the accuracy and the complexity of that as we actually evolve in technology, it will become more harder and harder in that aspects. And this is where AI, the powered attack and the use or the misuse of this AIs in order to build such automation, personalization, phishing, deepfakes, or even targeting personalities in that aspect. So definitely the technology will continue evolving with us. Today we speak about AI, tomorrow by quantum, day after tomorrow autonomous systems or any of those technologies in that perspective. Yet the people, they are still the weakest link in that perspective. And this is what we need to strengthen by partnership, by collaborating, by working together. And really putting the right governings in that perspective with us all. Our model is actually based on various aspects of this. We have a national cybersecurity strategy and the national AI strategy, as a matter of fact, that focuses on 5 main pillars: governings, buildings, capacity buildings, as well as focusing on, as a matter of fact, the technology itself and centric on the partnership, and this is where many of those partners are really helping us to come handy. No one nation can basically counter many of those aspects. We need that partnership. We need to work together with the private sector, with the civil liberties entities, with the NGOs, and that's what actually have really resulted in many of the aspects that we've shown where We have developed under each one of those pillars so many initiatives that allows us to actually partner and work together with everyone in that perspective. Technology is really, as mentioned by my colleagues here, it's a dual edge in that perspective, but using it in the right way, putting the right governings and really leveraging that, protecting, defending, and innovating for startups, for private sectors, for many of the things that we really needed in order to counter those terrorism is the way forward. We cannot do this ourselves. We need, as shown by the last pillar, to partner, partner with everybody, and that's what we call it, those 4 Ps: public, private, people, partnership. And that's what makes us all, as a matter of fact, to emerge stronger and work together in order to defend many of those attacks against our nations. Thank you very much, and thank you for the time.

Yes, thank you.

Um, thank you, Your Excellency. I'm very much appreciative of the insights from the UAE, and I think I can only underscore the importance of, as you said, the human element and the importance of the human. That's something we tried to bring out as well, I think, in particular through the AI-Pol initiative, the importance of the human element. And in particular for us, has been mentioned in the opening session as well, the importance of trust in the systems and the trust in the agencies. So thank you very much once again. I now have the pleasure of inviting Mr. Kentaro Matsumura, who is a counter-terrorism expert at the Permanent Mission of Japan, to share some insights in terms of innovation to application in counter-terrorism. Ken, over to you.

All right.

Good morning, distinguished colleagues. Japan has long been committed to the responsible use of artificial intelligence particularly at the intersection of AI and security. Through both research and advocacy, we continue to work towards ensuring that emerging technologies are developed and deployed in a manner that is safe, accountable, and aligned with our shared values. Today, we are pleased to bring the latest thinking from the private sector and academia into this conversation. It is my great pleasure to welcome Dr. Yuki Mitsufuji, visiting research professor at New York University. Dr. Mitsufuji will talk us through the current state of cutting-edge research and offer his perspective on how these advances can contribute to the safe and responsible use of AI, including in the context of counterterrorism. Dr. Mitsufuji, you have the floor.

Bimmi, um, could you show the slides, please? So, it's my great pleasure and also honor to be here to be able to talk about, you know, recent trends in protective AI. So, today I won't go into any, you know, deep dive into each technology, but I will briefly skim through the latest technology accepted at the major conferences, AI conferences. So the title is like AI for creators, but today I'd like to talk about in the context of counterterrorism. Next, please. So as you may know, Japan is increasing the amount of contents, export contents, and especially anime and J-pop are the main target, main contents. So this is the first context. And next, please. And also, people in the world, different from last decades, more familiar, become more familiar with anime and music, J-pop contents. And this picture is from Demon Slayer. I believe that more children are familiar with this type of anime content. So that means they will be more influenced with this content. If— and next, please. So after we see that, we see many powerful video or image models, AI models, we also see some IP characters in the output of gen AI models, which is not, you know, lawful. So we see three key problems in recent video generative models. The first one is copyright infringement. So we see a lot of characters, famous characters, in the output. And the second one is not directly related to copyright because the style is not currently protected by any— by law. But still, style is preferred to be selected in the output of GenAI models. So, style imitation is another problem. And the third one is memorization, especially verbatim memorization, the direct copy of a training dataset. So the first one, actually, if— it's actually easy for people to generate contents based on some famous characters and embed some political messages or some inappropriate messages for children. So to protect this, Japan-based lab developed some technologies to actually not only detect but also, uh, a way to, uh, find a way to remunerate, uh, to creators. So this one on the left-hand side was presented at ICLR, one of the biggest AI conferences held, uh, in Brazil, uh, this spring. Uh, the technology could easily, uh, identify the characters used in the output and properly calculate the contribution to the content. So this is the first one. And the second one is style imitation. So here, you know, since these rights are not protected by law, so the only way the Japan lab came up is to actually steer the output away from the undesired output. So it's called negative prompting. So we could actually make the model steer away from certain contents by giving specific targets to the model. So in this case, a Ghibli effect is a target. So if you show the Ghibli image to the model, And at the inference time, we could avoid generating Ghibli style. And the third one is memorization. So memorization is actually a big problem of GenAI models. Currently, there's no theoretical way to— theory-based way to avoid this problem. So let's say whatever we see on the on the website, GenAI, all the GenAI models have to some extent this problem. So, to avoid this, there's a way to actually avoid it in the training stage, but also at the inference stage, we could avoid this problem by properly generate contents by inserting 2 or 3 sentences. And this was presented at the Computer Vision Conference called CVPR. So next, please. And I would like to also talk about the implication to music market and industry. So the chart on the left-hand side shows the ratio between human-made music and AI-generated music in a platform called, you know, Deezer, the company called Deezer, a French-based digital service provider company. And the ratio as of 2025 January was like this. So only 15% of the whole market was AI-generated. But now, as of— not now, but, you know, back in January 2026, it increased up to 40%. So we could easily imagine that in 2030, uh, the figure like this, uh, will be the ratio. So the market would be filled with, uh, AI-generated contents. Uh, could you go next, please? So this actually is a very, uh, bad situation for creators because the, the, the current ecosystem can remunerate correctly to creators, but the gen AI output doesn't, you know, feedback anything to creators. So this technology is called training data attribution that could actually connect the output and also training dataset and calculate the contribution, how much each training dataset training sample contributed to generate the output. So here, the example shows like the red signal contributed 50% and blue 20%, yellow 30%. Yeah, this is the, you know, rough idea of attribute training data attribution. And next, please. So I would like to briefly give you an idea how we could usually, you know, calculate this kind of type of contribution. So, the top figure shows the full model. So, you have a model in the middle of the page and the text is fed to the model and you get AI song. And the signals surrounded by magenta are the training dataset. So, how you calculate the contribution is very straightforward. So, if you look down, there's another model and this one was trained on the dataset, almost the same dataset, but the only exception is to remove Yellow Song from the dataset. So, we call it leave-one-out model. And if we remove this red— Yellow yellow signal and feed the same text prompt and see the output. If the difference is very small, it means that yellow signal was not important to generate that song. And if you further look down to the model B where we, you know, where the red signal is removed, we see a big difference. So, it means that red each signal was very important. But as you can easily imagine, this is almost computationally infeasible, so there's a, you know, workaround. So could you go to the next page, please? So the easy way to do this is to forget certain, you know, signal rather than, you know, removing certain song from the dataset and retrain. So this is how it's usually done in the training data attribution. But it's still very difficult to achieve this because forgetting is not equal to, you know, retraining with the dataset, removing one signal. So there are a lot of technical challenges and Japan-based lab is trying to solve this problem. So this is the end of my presentation and thanks for listening.

Thank you very—

oh, thank you very much to the distinguished representatives from Sony. And of course, thank you to, um, Kentaro and the Permanent Mission of Japan, um, for sharing the insights, I think, from a Japan perspective in terms of, um, the generative capabilities and I think some of the challenges we face in terms of advancements in terms of generative AI, um, which naturally, of course, has widespread ramification ramifications in the counter-terrorism context. So with that, we will now move to the next side, next part of the session, which will focus on UN system support to member states, in particular the AI-POL and CT Tech initiatives. And I have the pleasure to hand over to my co-moderator, Ms. Gino Tene. Akvila.

Thank you very much, Oren, and good morning, Excellencies, dear participants of the side event. I will speak about how we see at the UN system artificial intelligence reshaping or adjusting counter-terrorism responses of member states, what kind of support is available to member states through UNOCT Global Programme on Cybersecurity and New Technologies, and what are emerging and growing needs when it comes to AI and counter-terrorism. And as it was emphasized by many speakers today, AI presents both threats, opportunities, and risks. And it was recently very well acknowledged in, in the report of the Special Rapporteur on Counterterrorism and Human Rights on AI and CT. So let us start with opportunities. First and foremost, artificial intelligence fundamentally augments security agencies' capacities to collect, process, and analyze vast quantities of data and identify patterns, anomalies, and threats more rapidly, efficiently, and effectively. AI systems are analyzing today many millions of data points at the same time, like communications metadata, financial transactions, travel records, and social media activities, and identify correlations and behavioral patterns by providing real-time or near real-time threat assessments. Next, artificial intelligence is increasingly used alongside biometrics such as facial recognition by public authorities and law enforcement agencies to verify people's identities in public spaces, regulate access to major events, identify individuals upon arrest, or while monitoring activities in public spaces. AI is also used for enhanced mass surveillance, predictive policing, and force deployment. It helps law enforcement and border security agencies to identify terrorist leads, analyze mobility patterns, interactions, and associations between individuals and groups. It also helps to faster identify terrorist content online by automatically scanning, detecting, and classifying massive datasets, which helps to disrupt terrorist content and distribution channels. And also on top of that, AI is increasingly used for cybersecurity, for protection of our critical infrastructures against terrorist cyberattacks and urban spaces. But at the same time, the use of artificial intelligence tools for countering terrorism, if not governed properly, may result in interference in the right to privacy and liberty and might not meet the requirements of necessity, proportionality, as required under international human rights law. It can have damaging effects to civic spaces, political participation and democracy, which themselves build resilience against terrorist violence. And on the other hand, also, like, if artificial intelligence systems are not secure enough, they can be breached by malicious actors, which in turn can cause significant damage to our counterterrorist efforts. Next, please. UNOCT has established Cybersecurity and New Technologies Program to help member states to to navigate these opportunities and risks. And the program is a direct response to the 7th review of the United Nations Global Counter-Terrorist Strategy, where member states requested United Nations Office of Counter-Terrorism to support innovative measures and approaches to build capacities of member states to respond for the challenges and opportunities that new technologies provide in countering terrorism, including artificial intelligence. And the program delivers this assistance through 3 major streams of work. First, we help member states to improve their policy approaches to new technologies by helping them to conduct threat assessments, identify and prioritize risks, and develop policies to address those risks. Second, the program supports member states in improving their law enforcement capacities to use new technologies for counterterrorism investigations. And last, we also support member states in improving their capacity to protect their critical infrastructures against terrorist cyberattacks. And as you can see from the slide, artificial intelligence is an integral part of all these efforts. We work on integration of artificial intelligence in counterterrorist strategies, in cybersecurity of critical infrastructures, and in law enforcement capabilities. Next, please. So this slide provides a little bit of an overview of evolution of artificial artificial intelligence in the programme portfolio. And all the journey, this started in 2021, where we, together with UNICRI and with the support of the Government of Japan and the Kingdom of Saudi Arabia, started to look what AI means for counter-terrorism, how terrorists can abuse artificial intelligence for their purposes, but also how member states can use artificial intelligence to counter-terrorism. And it resulted in two reports which were very well received by international community, and I think they are still quoted today by academia and researchers. In 2022, we started to look at artificial intelligence and from cybersecurity perspective. And together with the Republic of Korea Counter-Terrorist Preparedness Network, United Arab Emirates, and Austrian Institute of Technology. We organized a number of tabletop exercise to test and develop better responses to AI-generated malware, both at national level and at city level. In 2023, this collaboration continued to also together with International Telecommunication Union and Organization of American Member States, where we also integrated responses to deepfakes as part of a cyber incident response, because we see in cyber domain that cyber incidents are also followed by generation of a lot of deepfakes, and how Member States really need to approach that. And we were very happy also to contribute to the major and the biggest global cyber drill that was held in Dubai on, on the dark web and cybercrime as a service. Also, this year in 2023, it marked the launch of one of the major initiatives on new technologies in general in the UN system called CTech, which we launched together with Interpol and with the European Union. And the initiative started to look how to approach new technologies in general when it comes to counter-terrorism. Initially, we focused a lot on development of the knowledge products, what are the good practices. We invited member states, academia, tech sector to contribute, and it produced, I think, 9 knowledge products to guide capacity-building activities of UN system, but also of other interested organizations. And I'm proud to say that we are still using those knowledge products today. In 2024, we collaborated with the Counter-Terrorist Preparedness Network again to look at the city preparedness to the use of artificial intelligence by terrorists. What needs to be done, how cities can employ AI to improve cities' operation and security of the cities and urban spaces. 2025 marked the launch of CTech+ initiative with the European Union and Interpol, and it had quite strong elements already on artificial intelligence. Through the initiative, we delivered training to investigators on the use of artificial intelligence in counter-terrorism investigations. 9 partner states have benefited from that training, and I think we trained more than 300 people when it comes to that. Also, in 2025, together with another program at UNOCT, Parliamentary Engagement Program, we joined forces and we developed a training on regulating, legislating, and overseeing the use of artificial intelligence for counter-terrorism. The training program is delivered to the parliamentarians, and we launched the pilot training in Jordan in November with the parliamentary engagement, Parliamentary Assembly of Mediterranean, and there'll be another side event at 1 o'clock to discuss parliamentary approaches to AI. This year we are continuing our journey with UNICRI and with our partners. So we are working on updating a handbook that we maintain with Interpol on, uh, on open-source investigations. So there'll be a big chunk of updates related how artificial intelligence can enhance open-source investigation. It will be made available to all member states' law enforcement agencies to use it. Also, we are continuing the training for parliamentarians, and in February we delivered the same training to ASEAN Parliamentary Assembly In September, in Italy, together with UNICRI and together with the AAPOLE initiative, we will be launching a training on artificial intelligence and digital transformation for security sector agencies aiming at the leadership of law enforcement, for them also to enable them to adopt artificial intelligence responsibly, both for counter-terrorism but for other law enforcement purposes. And this year, we also launched a number of public-private partnership exercises under CTech+ initiative, together with Interpol, but also with the GFCT and Tech Against Terrorists. They were very well received, and they also address AI-generated terrorist content online and fostering public-private partnership to address that. Next slide. Please. Of course, there are like a lot of efforts, but from the engagement from member states, we see that they increasingly require tailored capacity-building assistance to help them to adopt artificial intelligence in counterterrorism. Their interest is quite high. The fast-paced and dynamic nature of technology poses significant challenges and opportunities for member states on understanding the implications, developing appropriate policies, and building right capabilities. AI landscape is dynamic and evolving faster than the authorities can keep up, while terrorists at the same time, they are adept and effective in leveraging new technologies in artificial intelligence. While counterterrorist agencies, we also see they're quite slow of responding because of the lack of awareness, on how to use new technologies. The use of artificial intelligence is not sufficiently legislated or requires a lot of resources and investments. But despite of that, all member states that we engaged, they want to better understand key technological trends and implications of AI in countering terrorists and develop appropriate approaches at policy and operational level and to build necessary AI and data infrastructure to use AI in counterterrorism. So to finish, I would like to thank all our program partners, the European Union for this forward-looking City Tech and City Tech+ initiative, Japan, and the United Arab Emirates and India, as well as Republic of Korea, Organization of American States, Kingdom of Saudi Arabia, for enabling us and the UN system deliver the much-needed support to member states. And we are looking forward to continue that cooperation. And with this, I will pass the floor to my colleague, Oren McCarthy. He's liaison officer and program officer at UNICREE, and he will provide an overview of the support that UNICREE is providing to member states when it comes to AI.

Thank you, Akfile. And it's just from UNICREE side. I mean, uh, from that timeline you presented, it's a pleasure to have been part of that journey along the way, um, from the start going back to that contribution from Japan to the initial, um, studies into AI and counterterrorism, and in various steps along the way. Um, but for the few minutes I have with you, um, I'd like to just focus on, um, one particular initiative, um, which again, if you think of that timeline, ACVILA had for us really kind of kicked off in 2023, and this is the EU-funded work that was mentioned in the opening remarks, specifically the AI-POL initiative. Next slide, please. But before talking about AI-POL, I want to just cast the timeline a little bit back further to a piece of work that we did, which is really the backbone of everything that we do in the AI space, both within the counter-terrorism context and more broadly in law enforcement as a whole. And that is the Toolkit for Responsible AI Innovation and Law Enforcement, which which is funded by the European Union and developed by UNICRI in partnership with Interpol. And if you scan the QR code, it'll take you to the, um, the website where you can act— all the access, all the resources. And the toolkit itself is, is fundamentally what's been referred to by, um, Interpol Secretary General as the blueprint for responsible AI and law enforcement. It is effectively, um, a practical guide for law enforcement agencies to better understand AI, understand how to apply it in the law enforcement context, how to build foundations and to do so responsibly. Next slide, please. I think one of the important questions that we try to address in our work is this: what is responsible AI? Very, very briefly, for us, you know, we talk often about effective AI. Effective AI is responsible AI, and responsible AI is effective AI. So for us, it's very much intertwined. But digging in a little bit more deeply into what is responsible, Obviously, when we talk about conversations about AI, we talk very often about ethics and the legal foundations of it. And for us, responsible AI is something that sits at the intersection of law and ethics. But as well, I think importantly, it also sits at the intersection of the principles of policing itself. As you may be aware, policing as a function, as a service, is very much a principled function. And agencies all around the world will have a statute, a code of conduct, a code of ethic, a constitution, some sort of constituent document that will set out principles. And many of the principles we talk about, what are responsible AI, are very much reflected in those foundational documents in many ways. So for us, responsible AI in law enforcement is really at that intersection of law, ethics, and the pre-existing principles of AI. Sorry, can we just jump back one? Backwards, please. Yeah, but fundamentally, if we're to describe what it is, we would say according to us and the toolkit that we've developed, responsible AI is, is something that respects the law and human rights, minimizes harm to the— to any people, respects human autonomy, promotes fairness, and is supported by good governance. And this is for us the foundation of responsible AI. Next slide, please. So what is the toolkit itself? As I said, it's a practical guide. It's a series of 7 documents, theoretical documents and practical documents. And the theoretical documents cover things like an introduction to responsible AI. What is that slide I just presented? What does it all mean in a law enforcement context? It's got a technical reference book, so explaining the technical concepts of AI for the law enforcement community in non-technical language. It's got those principles that I mentioned, and what are the principles of responsible AI innovation and law enforcement? And importantly for us,. And for the next part of our work, it's got an organizational roadmap. How do we get our agencies into a state of readiness to implement AI effectively and responsibly? Um, at the same time, as I said, it's got practical tools. Uh, so we have an organizational readiness self-assessment that we encourage agencies to use to, to see how prepared they are to, to innovate and to do so responsibly. It's got a risk assessment tool to help agencies understand, uh, and measure the level of risk, uh, with respect to specific tools. And it's got a workbook, a responsible AI workbook. And the workbook essentially is a companion guide that helps, encourages agencies to think about the various responsible AI considerations that may arise over the entire lifecycle of an AI system. Next slide, please. But moving up to more present times, the AI-PoL initiative. The AI-PoL initiative was launched about a year and a half ago, as I mentioned, funded by the European Union and implemented UNIQRI in partnership with Interpol. Effectively, the AI-Pol initiative seeks to take that framework that we built out in the— with the toolkit for responsible AI innovation and to pilot that in specific agencies to see can we advance and support responsible AI in practice. We are, for the— this initiative, starting our cooperation with the Nigerian Police Force, the Ministry of Internal Affairs of Kazakhstan, the Brazilian Federal Police, the Royal Oman Police, and the Central Bureau of Investigation investigation of India, and we're delighted to have representatives from Nigeria and India speaking to us a little bit more about their experiences in the next stage. Next slide, please. Um, and you can just click on— there's an animation as well. Yeah, so, uh, at its core, the AI-Pol initiative has a couple of main work streams. Um, first is leadership engagement. Um, over the previous phase of work, we engaged a lot with working-level people around the issue of responsible AI in law enforcement.. But the feedback we kept receiving was, um, if we want to make these change, we need to talk to leadership and we need to build knowledge and understanding within, at a leadership level. So the AI-PoL initiative seeks to target executive leadership in law enforcement, build their knowledge and understanding, and support them to, to carry out this internal self-assessment and build things like strategies and policies around responsible AI innovation. The second is community engagement. It's been mentioned in the opening remarks, um, the importance of the human side of things and in particular public trust. And this is critical for, for us in our work, building and maintaining public trust. And so we have an entire work stream about trying to bring the public closer to AI and law enforcement, build understanding, understand the concerns the public may have, and try to navigate those concerns and meaningfully engage the public in this, in this space. The third is knowledge development. Again, this is a new space for all of us. It's a evolving space has been mentioned. And so we're trying to keep up with this, continue to build knowledge within this community, and continuously add new documents into that toolkit. And I'll talk about that again in a moment. And then finally, it's not an official work stream of the program, but essentially we are trying to strive to build a community around responsible AI innovation and law enforcement, raise visibility about the efforts that are being made in the space, and, and just to connect and build bridges between agencies who are all navigating these same shared challenges.

Next slide, please.

So, as I said, at its core, the AI-POL initiative, if we're to boil it all down, seeks to address one main question: are you ready to use AI? And for us, one of the key instruments in the toolkit that we seek to use is this organizational roadmap and the organizational readiness assessment. And we are working at the moment now with each of those 5 countries that I mentioned to to conduct organizational self-assessments to measure, you know, do we have the right culture in place in our agencies? Do we have the right people expertise that we need to innovate and to do so responsibly? And do we have the right mechanisms, for instance, oversight committees, you know, peer review kind of mechanisms and so on, to be able to support responsible AI? Next slide, please. So as I mentioned, it's, it's an ongoing initiative, but we have had several good practices, and I believe we're going to hear a few moments a little bit more about this, but I'm pleased to say that, um, we have conducted organizational readiness assessments for 4 of the 5 participating agencies, and we hope to have the 5th one finalized soon. Um, but just a flag, I think Nigeria in particular, the Nigerian police force, um, has advanced quite significantly, having completed the organizational self-assessment and already conducted an in-person country workshop only a couple of weeks ago where where we sat and we presented to the Inspector General of the Nigerian Police, um, in terms of the, the key findings and takeaways, um, from that self-assessment. And while we're not in a position to disclose the findings, um, I think the key takeaway from that was, um, that, you know, the innovation in the agency itself was at a foundational level, um, but transitional in terms of its knowledge and understanding about responsible AI innovation. So in terms of next steps, we'll continue to support the Nigerian police force as well as the other 4 participating agencies in things like building out strategies, building AI literacy, and continue to engage with the, the public and facilitating innovation. Next slide, please. And then finally, Akfile mentioned it already, but we will be conducting a training program on AI and digital transformations in partnership with the Ministry of Foreign Affairs of Italy, and that'll be taking place in September in Rome. And as has been mentioned, we'll be involving the, the 5 participating AI-Pol countries as well as the 9 CT Tech countries, and essentially talking to leadership about the importance of AI in driving a digital transformation, um, and, and how do we leverage, um, AI in this regard. And spinning out of that, we will be, um, developing another guidance document, um, which will essentially, um, build on those— that's the foundations of, of the training program. Is leveraging AI and driving digital transformations in a counter-terrorism context. And so we'll be building out that guidance document and releasing later this year, early next year. So with that, thank you. Thank you once again, I think, to OCT for partnership and of course European Union for funding this initiative. And looking forward to hearing your insights.

Our next segment is dedicated to hearing participating member states' perspective. And we have representatives from Nigeria and from India to share their experience working with CTech+ and AI-POL initiative. And I would like to invite Mr. Adam Muhammad Alkali from National Counterterrorist Center at the Office of National Security Adviser in Nigeria. I should note that NCTC of Nigeria was an excellent partner of CTech+ initiative and did a lot of efforts in adopting new technologies for counter-terrorism, including AI, responsibly. Over to you, sir.

Thank you, Excellencies, distinguished delegates. It is a distinct honor to participate in this important side event. As earlier stated by earlier speakers, artificial intelligence is no longer a technology of the future. It is already shaping the counterterrorism landscape. Terrorist groups are increasingly exploiting artificial intelligence, encrypted communications, cryptocurrencies, and other emerging technologies to recruit, finance, and coordinate operations across borders. In the Lake Chad Basin, Nigeria has witnessed how rapidly terrorist groups adapt to these technologies, including the use of armed drones by the Islamic State West Africa Province. At the same time, these technologies represent unprecedented opportunities for governments to prevent, detect, and disrupt terrorist activities, provided they are deployed responsibly in full respect of human rights and the rule of law. Nigeria's response is anchored on a strong legal and strategic framework comprising the Terrorism Prevention and Prohibition Act of 2022, the National Security Strategy, the National Counterterrorism Strategy, the National Policy Framework and Action Plan for Preventing and Countering Violent Extremism, and the National Counterterrorism Center Strategic Plan 2025 to 2030. Together, these frameworks are driving the integration of artificial intelligence, digital technologies, and data-driven decision-making across Nigeria's counter-terrorism architecture. Excellencies, one of the most important enablers of this transformation has been the CTech+ initiative, a joint initiative of the UN OCT, UNCCCT, Interpol, and the European Union. Nigeria is proud to be one of the 9 participating countries, and we will remain appreciative of that. CTech+ has so far delivered tangible outcomes. In 2025, CTech+, through the National Counterterrorism Center, undertook a comprehensive assessment of our law enforcement capabilities to counter terrorism by the use of emerging technologies. This resulted in the development of Law Enforcement Capabilities Framework, whose recommendations have been incorporated into the National Counterterrorism Center Strategic Plan 2025 to 2030. I have few copies which I've made available to the member states. The initiatives also supported the integration of emerging technologies, threat assessment into ongoing review of the National Counterterrorism Strategy, ensuring that our national policy keeps pace with evolving threats. Equally important has been the investment in human capacity 24 investigators drawn from 9 security and law enforcement agencies have received training in open source intelligence, darknet investigations, cryptocurrency tracing, and the application of artificial intelligence in counterterrorism investigations. Overall, more than 20 Nigerian institutions have participated in CTech activities, which has significantly strengthened our interagency coordination and national preparedness. In April 2026, Nigeria hosted the CT Tech+ private-public partnership tabletop exercise that brought together government institutions, regulators, and private sector actors to test coordinated response to technology-enabled terrorist threats. The lessons learned are already informing our operational procedures. We also welcome the ongoing live operation component of CT Tech+ and look forward to the forthcoming Artificial Intelligence and Law Enforcement Leadership Program in Italy later this year, which will further strengthen strategic leadership in this critical domain. Excellencies, Nigeria wishes to underscore two priorities. First, international capacity-building initiatives such as the CT Tech+ should be sustained and expanded particularly for countries confronting increasingly sophisticated terrorist threats. Second, investment in technology must be matched by strong governance frameworks that ensure artificial intelligence is deployed responsibly, ethically, and in accordance with international law. To this end, Nigeria stands ready to share its experience, contribute to international norm settings, and support regional capacity building through the National Counterterrorism Center as it transitions into a regional center of excellence for counterterrorism in West Africa and the Sahel. In closing, Nigeria's experience demonstrates that meaningful partnership produces measurable results. On behalf of the Federal Republic of Nigeria, I thank the core convenants, the Permanent Mission of India, Japan, and the United Arab Emirates, together with the European Union, the UNOCT, UNCCD, and UNICRI for convening this timely and important event. Thank you.

Thank you very much, Mr. Alkali. Next, we have a representative of India who is joining us online, Mr. Amit Vassava. He's a police superintendent from the Central Bureau of Investigations of India, and I would invite him to deliver his remarks.

I'll just share my— good morning everyone, this is Amit Wasawa. I work with the Central Bureau of Investigation India. CBI is the premier investigating agency of the country. As discussed before, India is one of the participating countries for AI-POL initiative. Apart from the gains for my agency, I am personally indebted to AI-POL program. I, despite being proficient in cybercrime investigation, the AI scene had surreptitiously crept far ahead and I had clearly missed the bus. Since then, uh, Since the AI Poll Conference, I have been hooked to learning about AI, and I have championed the adoption of AI projects at CBI. One of the roles of CBI is vigilance, requiring searching massive databases of investigations, complaints, witness statements, and inquiries. Sharanj, an AI tool recently launched, enables us to retrieve data based on context and meaning rather than simple keywords. We face a unique challenge from cybercrime gangs who send fake legal notices to victims and threaten them in name of CBI. We have recently launched an AI-based tools on our website to authenticate genuine legal notices. At CBI, we also use NetGrid, which is an AI-based search and facial recognition tool with massive national-level databases for finance, financial transaction, travel, immigration, telecom, etc. This has led to solving of otherwise difficult cases and catching of many absconders. We also use several AI-based tools for financial transaction analysis and social media data analysis. Prahaar is India's first ever comprehensive code— codified national counter-terrorism policy and strategy. It formalizes a strict no-tolerance approach towards terrorism. The policy focuses on proactive intelligence fusion, institutional coordination, and tackling of modern hybrid threats such as cyber radicalization, cryptocurrency terror financing, and drone-based logistics. I would also take this opportunity to point out an important issue of algorithmic bias. And why implementation of AI system in India face a unique challenge of diversity, scale, and language. Anyone who visits India for the first time is overwhelmed by the crowds, the noises, the smells, the traffic, and of course the cows. It is a chaos indeed. We have talked about AI-based tools for analyzing financial transactions. India is so digitally advanced that even beggars will come knock on your car window with a QR code for online payment. The traditional excuse of not having change doesn't work in India anymore. With more than 700 million free-of-cost transactions each day, India's UPI has set a spectacular global gold standard. Can anyone guess what would be the rank of Delhi in terms of population if it were a country?? It would be 60th in rank, which means more than of 140 or so countries. Delhi Police manages through its centralized control room a massive network of more than 300,000 CCTV cameras. Can anyone guess what would be the rank in population of one of our state, Uttar Pradesh, if it were a country? It would be 5th in the entire world. The state police of Uttar Pradesh uses AI system in very many ways, including satellite image analysis for encroachment and environmental crimes. Every 12 years, on confluence of two holy rivers, a religious fair called Kumbh Mela is organized. People from across the world participate in this month-long celebration. If this fair were a country, What would be its rank then? With 660 million visitors taking dip in the river, it would be behind only India and China. Counter-terrorism at Kumbh is an unprecedented logistical nightmare, requiring security forces to execute hyper-vigilant and high-tech surveillance without disrupting the profound spiritual flow or triggering a catastrophic panic in the world's largest human gathering. Most advanced tool helped us manage this flow and keep it secure. What we call AI in general sense is actually large language model. As the name suggests, it works on learning from large text of a language. India, with its 22 official languages, 122+ major languages, and 1,600+ dialects pose a special challenge for AI models. Code-switching, that is seamless switching between languages, is yet another major challenge. Also, even Hindi, the most widely spoken of Indian languages, contributes only to about 0.1% of the total text of internet. Famously, AI systems cannot create an image of a person writing with his left hand. Because the majority of the data fed to it was of right-handed persons. Just imagine then, how can AI actually know India, which is like 50 countries in itself? Yes, foreigners coming to India are initially overwhelmed by the intense sensory overload and sheer chaos, but once they immerse themselves into it, they develop a deep love for the country's vibrant warmth and unity in diversity. I predict the same thing happening to AI. Project Vashini is India's AI-powered platform designed to break the language barriers by providing real-time open-source translation, speech, and text services across all 22 official Indian languages. It acts as a digital public infrastructure empowering millions of non-English speakers to access the internet, government e-services, and digital tools seamlessly in their own mother tongues. It is also the world's largest crowdsourced data co-creation initiative by native language speakers. The central pillar of India's technical counter-terrorism effort is the National Intelligence Grid, an integrated master database linking more than 20 siloed data pools including banking, immigration, telecom, and travel records. There are AI tools being used for border security and satellite imagery analysis. Indrajal is an AI-based anti-drone system. There are specialized AI systems to detect radicalization on social media platforms. There are facial recognition systems, video analytics, and CCTV monitoring systems being run by various city police. I'm thankful for giving me this opportunity to share my views. Thank you all for your time.

Thank you very much for joining us for this important event and sharing India's experience. And with this, I pass the floor again to my colleague, Oren.

Great. Thank you very much, Akhileh, and thank you to all our distinguished speakers who have joined online and in person. We're just at time, but we'd like to at least offer the opportunity to take one or two questions from the floor. So, if you'd bear with us 2, 3 minutes, we can do so. We had a couple of pre-inscribed speakers. I'm just going to see if they're still in the room, but we had a request to take the floor from Mozambique, if the distinguished representative of Mozambique would like to take the floor. No, may not be with us. And then we had a request from the distinguished representative from Yemen.

No?

Okay. And in that case, then the next I have on my list, I have a request from the Special Rapporteur for Human Rights and Counterterrorism, Mr. Saul.

If we're at time, I'm happy to pass, if that's fine by you, or—

I mean, we're just on time. If it's a quick intervention, no problem.

I'm always happy to speak. Excellencies, distinguished colleagues, in the experience of the United human rights mechanisms, the use of new technologies in countering terrorism has unleashed new waves of human rights violations, often targeting civil society, human rights defenders, journalists, and political opponents. There's good reason to expect the same from AI unless old habits change and AI is effectively regulated. My recent position paper on AI and counterterrorism demonstrates the high risks that AI poses to human rights, from surveillance and predictive policing— policing to border security and military operations, stemming from the intrusive capabilities of the technology, its flaws and limitations, and its misuse. Effective regulation requires at least the following 6 steps. First, there must be clear and explicit legal regulation with no blanket exceptions for law enforcement, security, military, and border management purposes. Laissez-faire industry self-regulation is clearly not sufficient. AI systems that cannot meet human rights standards or that pose an unacceptable risk to human rights must be prohibited from use, transfer, or export. Permissible uses must be stringently regulated by law. Secondly, developers and deployers of AI must undertake rigorous human rights due diligence assessments of AI systems throughout their lifecycle to prevent, mitigate, and remedy harms. The cumulative effects of involve multiple technologies, from facial recognition to digital surveillance, must be carefully addressed together. Thirdly, there must be strict safeguards on data quality testing and validation, uh, and personal data protection and data security, including in cross-border intelligence sharing. Fourth, AI systems must be transparent and explainable to enable effective remedies. Fifth, there must be always effective and meaningful human control in decision-making and guardrails against automation and confirmation bias. Finally, there must be effective and accessible oversight and accountability mechanisms to prevent, suppress, and remedy human rights violations. We should not fatalistically believe that we cannot collectively control the current technological arms race and the trillion-dollar investment boom. The future must be in all of our hands. Thank you.

Thank you very much. And just— we have like maybe 1 minute if there's a question from the floor, if anybody would like to make an intervention. If not, I think we can wrap. Please, we have a speaker at the end.

Thank you very much. I'm Adam Hadley from Tech Against Terrorism. And very briefly, I wanted to announce that at Tech Against Terrorism today, we've published our Counterterrorism AI Benchmark., which has worked to evaluate 27 AI models with respect to how they respond to prompts that we believe terrorists, uh, might be making. Really alarming findings in that a third of all cases, these AI models have provided significant uplift. Um, we've evaluated models by Anthropic, Alibaba, Cohere, Deepseek, Falcon, and others. Um, and we certainly welcome a platform perhaps on another occasion to share more detail about our findings.

Thank you.

Great.

Thank you, and thank you for bearing with us while we take at least some remarks from the floor. Um, so with that, I think we are beyond time. So allow me once again, on behalf of UNICRU, UNOCT, uh, to thank our distinguished hosts, um, permanent missions of India, UAE, Japan, um, and the European Union for their support. Uh, we look forward to continuing the partnership, um, as we said, as we move into the implementation stage in AI and counterterrorism. So thank you all, and thank you for joining us once a day— today. And we're closed. Thank you.