High-level multi-stakeholder plenary segment: Harnessing the benefits of AI for all through inclusive and interoperable approaches - part 2

6 intervenants
Résumé

Résumé

La session a porté sur la manière de transformer la gouvernance de l'IA, en passant de grands principes à des systèmes pratiques et inclusifs, grâce à la collaboration entre gouvernements, industrie, monde académique, société civile et organisations internationales . Whitney Baird a cadré la session autour de l'idée que toutes les personnes concernées par les promesses et les risques de l'IA doivent être associées aux discussions sur sa gouvernance . Bosun Tijani a affirmé que les gouvernements doivent aller au-delà des principes pour mettre en place des systèmes de gouvernance opérationnels, car l'IA remodèle déjà les secteurs et les activités économiques . Il a indiqué que les pays ont besoin de stratégies nationales claires pour l'adoption et la mesure de l'IA, d'une capacité renforcée du secteur public et d'institutions fiables. Il a également souligné la nécessité d'investissements accrus dans les infrastructures, la puissance de calcul et les talents . Selon Tijani, la faiblesse de la connectivité et des infrastructures limite à la fois l'utilisation de l'IA et la capacité à la gouverner efficacement, en particulier dans les pays à faible capacité d'absorption . Tshilidzi Marwala a identifié le principal écart de mise en œuvre comme étant l'« arbitrage réglementaire en matière d'IA », où des règles fragmentées et non coordonnées entre les organisations internationales, l'industrie et les États permettent à certains acteurs d'exploiter les juridictions les plus faibles . Il a appelé à une harmonisation, à une éducation généralisée sur les données, les algorithmes, l'informatique et les normes, à une meilleure formation des législateurs, ainsi qu'à un renforcement de la coordination entre institutions au niveau opérationnel . Rumman Chowdhury a également souligné que la participation effective dépend du capital humain et de l'évaluation, faisant valoir que les pays ont besoin d'organismes d'évaluation indépendants, de normes ancrées dans les réalités régionales et de règles de passation de marchés publics liant les contrats publics à des exigences d'évaluation . Chowdhury a indiqué que cela permettrait aux pays de passer d'une réception passive des systèmes d'IA à une participation à la définition de leurs propres priorités de développement . Kate Kallot a soutenu que la véritable souveraineté exige une appropriation locale plutôt qu'une simple consultation, citant la Barbade comme exemple où les données gouvernementales critiques restent dans le pays, le transfert de capacités est intégré dans les contrats, et des outils open source permettent aux développeurs et aux communautés locaux d'étendre eux-mêmes les systèmes . Elle a mis en avant l'implication des communautés dès la phase de conception comme l'approche la plus efficace . Yi Zeng a reconnu que les principes éthiques sont largement reconnus, mais a indiqué que les tests révèlent d'importantes défaillances de mise en œuvre, tant sur le plan éthique que sécuritaire, les services actuels de modèles de langage échouant souvent à respecter même les normes fondées sur le consensus et restant vulnérables aux attaques . Il a exhorté les scientifiques et le grand public à améliorer leur culture de l'IA. Yi Zeng considère l'ONU comme la meilleure plateforme disponible pour l'action collective et soutient que le dialogue devrait produire des résultats concrets, notamment des lignes rouges pour l'IA et un recours accru à l'IA au service des objectifs de développement durable . Lors du tour de clôture, les participants ont largement convenu que la prochaine étape est l'action : une plus grande inclusion des voix du Sud global, un soutien aux innovateurs locaux, un équilibre entre opportunités et risques, ainsi qu'un accès concret aux infrastructures et aux capacités de calcul partagées pour permettre aux pays de développer une IA adaptée à leurs propres contextes . Dans l'ensemble, le panel a conclu que l'efficacité de la gouvernance de l'IA dépendra de la confiance, du renforcement des capacités, du contexte local, de la coopération internationale et d'une mise en œuvre tangible, plutôt que de discussions de haut niveau supplémentaires .

Points clés

- Un thème central était la nécessité de passer des principes généraux de gouvernance de l'IA à une mise en œuvre pratique et mesurable. Bosun Tijani a soutenu que les pays ont besoin de stratégies nationales claires en matière d'IA, d'une capacité renforcée du secteur public, d'institutions fiables, de mesures d'impact et d'investissements dans les talents et la puissance de calcul pour rendre la gouvernance concrètement significative. - Les intervenants ont régulièrement mis en évidence la fragmentation de la gouvernance mondiale de l'IA et la nécessité d'une meilleure coordination et harmonisation. Tshilidzi Marwala a décrit ce phénomène comme un « arbitrage réglementaire en matière d'IA », dans lequel des règles nationales, industrielles et internationales non coordonnées permettent à certains acteurs d'exploiter les environnements réglementaires les plus faibles, et a appelé à l'éducation, à des valeurs partagées, à des incitations et à un renforcement de la coordination institutionnelle au niveau opérationnel. - L'inclusion, la souveraineté et l'appropriation locale ont été des priorités majeures, en particulier pour le Sud global. Rumman Chowdhury a souligné que la participation effective dépend du capital humain, d'organismes d'évaluation indépendants et du lien entre évaluation et marchés publics, afin que les pays puissent façonner l'IA plutôt que de simplement la recevoir. Kate Kallot a également soutenu que la véritable souveraineté exige un contrôle des données dans le pays, le transfert de capacités contractuelles et des systèmes ouverts que les communautés et développeurs locaux peuvent étendre. Bosun Tijani a ajouté que l'accès aux infrastructures et aux capacités contextuelles est indispensable pour que les pays puissent bénéficier de l'IA. - La confiance, la sécurité et l'évaluation ont été présentées comme des fondements essentiels d'une gouvernance responsable de l'IA. Chowdhury a soutenu que la puissance de l'IA réside dans l'évaluation et dans la définition de ce qui est bon ou mauvais, tandis que Yi Zeng a averti que les modèles de langage actuels au niveau applicatif ne satisfont toujours pas de nombreuses attentes éthiques et sécuritaires et restent vulnérables aux attaques. Plusieurs interventions ont établi un lien entre la supervision humaine, la culture numérique du public et la pensée critique d'une part, et un déploiement digne de confiance d'autre part. - Le panel s'est conclu par des appels à des résultats concrets issus du Dialogue mondial de l'ONU, en particulier une coopération internationale axée sur l'action. Les participants ont exhorté la plateforme onusienne à contribuer à une refonte de la gouvernance mondiale, afin qu'elle intègre les voix du Sud global, soutienne les innovateurs locaux, établisse des lignes rouges claires et une orientation éthique, et réponde aux prérequis tels que le calcul partagé et les infrastructures. Marwala a explicitement appelé à passer « des paroles aux actes », tandis que la modératrice a conclu en soulignant qu'aucune partie prenante ni aucun pays ne peut gouverner l'IA seul.

L'objectif général de la discussion était d'explorer comment la gouvernance de l'IA peut être traduite de grands principes en une action efficace, inclusive et mondialement coordonnée. Le panel s'est concentré en particulier sur ce que les gouvernements, les universités, les entreprises, la société civile et les institutions internationales - notamment l'ONU - devraient faire pour renforcer les capacités, améliorer la mise en œuvre et garantir que tous les pays puissent participer de manière significative à l'élaboration de la gouvernance de l'IA.

Le ton général était sérieux, constructif et orienté vers les solutions. Au début, la discussion a été présentée comme collaborative et urgente, avec une forte reconnaissance à la fois des promesses et des risques de l'IA. Au fil de la session, le ton est devenu plus incisif et prudent, notamment autour de la fragmentation, de la faiblesse de la mise en œuvre, des défaillances en matière de sécurité et de l'exclusion des pays moins puissants. À la fin, il s'est transformé en un appel à l'action plus ferme, les panélistes pressant pour des résultats concrets, une inclusion plus large et une coopération internationale soutenue par l'ONU.

Intervenants

- Whitney Baird - Modératrice de la session multipartite ; Présidente et directrice générale du Conseil des États-Unis pour les affaires internationales - Bosun Tijani - Ministre des Communications, de l'Innovation et de l'Économie numérique, Nigeria ; membre de l'Africa AI Council [S8] - Tshilidzi Marwala - Recteur de l'Université des Nations Unies et Secrétaire général adjoint des Nations Unies [S1] - Kate Kallot - Fondatrice et directrice générale d'Amini - Yi Zeng - Professeur à la Gaoling School of Artificial Intelligence, Université Renmin de Chine. - Rumman Chowdhury - Directrice générale et cofondatrice de Humane Intelligence. Intervenants supplémentaires : - Aucun.

Intervenants
BT
Bosun Tijani
161 wpm · 6 min
TM
Tshilidzi Marwala
129 wpm · 6 min
YZ
Yi Zeng
149 wpm · 6 min
WB
Whitney Baird
126 wpm · 9 min
RC
Rumman Chowdhury
162 wpm · 4 min
KK
Kate Kallot
173 wpm · 5 min

Whitney Baird a commencé par présenter le U.S. Council for International Business comme la voix des entreprises américaines sur la scène internationale . Elle a ensuite ouvert le panel en soulignant que les promesses et les risques de l'IA impliquent que toutes les parties prenantes doivent être associées au processus, et elle a présenté la session comme une discussion multipartite dans le cadre du premier Dialogue mondial de l'ONU sur la gouvernance de l'IA . Elle a également présenté les panélistes issus des milieux gouvernementaux, académiques, entrepreneuriaux, de la société civile et du système onusien . La session portait sur la manière dont la gouvernance de l'IA peut progresser vers la mise en œuvre de façon inclusive et applicable dans des contextes variés .

Un thème récurrent était la nécessité de passer des principes à des systèmes concrets . Bosun Tijani a soutenu que cette transition est urgente, car l'IA est déjà intégrée dans la vie quotidienne et bouleverse des secteurs et des entreprises de toutes tailles . Il a affirmé que la gouvernance devrait commencer par l'élaboration, dans chaque pays, d'une stratégie nationale pour l'adoption de l'IA et d'un cadre permettant de comprendre précisément ce qui est gouverné . Il a lié la mise en œuvre à plusieurs conditions : les décideurs politiques doivent avoir une compréhension suffisante de l'IA pour la réguler efficacement, les capacités du secteur public doivent suivre l'évolution technique, les institutions doivent être dignes de confiance, et les pays doivent investir dans les infrastructures de calcul, les talents et l'IA en tant que moteur de croissance .

Tijani a également insisté sur l'importance de la « capacité d'absorption » . Il a soutenu que la gouvernance repose sur des fondements pratiques tels que la connectivité, les infrastructures et la capacité à développer des compétences nationales . Si la connectivité est insuffisante, si les infrastructures font défaut ou si les populations ne peuvent que consommer l'IA sans contribuer à sa création, le déploiement comme la gouvernance s'en trouvent limités . Dans ses remarques de clôture, il est revenu sur ce point et a averti que, si les pays n'accèdent pas aux infrastructures et aux capacités nécessaires pour construire des solutions adaptées à leur contexte, le fossé numérique lié à l'IA pourrait devenir encore plus profond que celui engendré par l'internet .

Baird a répondu en soulignant la confiance, la capacité du secteur public et la connectivité comme des thèmes importants soulevés par Tijani . Elle est revenue à plusieurs reprises sur la question de la confiance : d'abord après les remarques de Tijani sur la confiance dans les institutions , puis après les commentaires de Rumman Chowdhury sur l'évaluation et la certitude , et à nouveau après l'intervention de Yi Zeng sur la supervision humaine et la pensée critique .

Tshilidzi Marwala a identifié l'« arbitrage réglementaire en matière d'IA » comme un problème majeur . Il a expliqué que ce phénomène découle du manque de coordination entre les modèles de gouvernance aux niveaux international, industriel et des États membres, ce qui crée des opportunités pour certains acteurs de rechercher des régimes moins contraignants, notamment en matière de données et d'infrastructures de calcul . Baird a répondu qu'une gouvernance fragmentée peut éroder la confiance et alourdir les coûts pour les développeurs et les entreprises .

Marwala a soutenu que l'éducation constitue l'une des réponses possibles . Il a affirmé que les sociétés ont besoin d'une meilleure culture des données, des algorithmes, de l'informatique et des applications de l'IA . Il a également dit que la gouvernance doit s'enraciner dans des valeurs, et que pour l'ONU, ces valeurs devraient découler de la Déclaration des droits de l'homme et de la Charte des Nations Unies . Il a en outre soutenu que l'éducation devrait façonner les comportements et les incitations, y compris pour les entreprises, et que les normes, politiques et réglementations devraient être simplifiées et intégrées dans les environnements d'apprentissage tels que les salles de classe . Il a ajouté que les législateurs ont également besoin d'être formés pour produire des lois sensées sur l'IA, et que les institutions doivent collaborer non seulement par le dialogue, mais aussi à un niveau opérationnel concret .

Rumman Chowdhury a soutenu que la puissance de l'IA réside « non pas dans le développement, mais dans l'évaluation », car c'est l'évaluation qui définit ce qui est « bon ou mauvais, juste ou faux » . Elle a affirmé que la véritable souveraineté en matière d'IA dépend de la gouvernance, de l'expertise sectorielle et du capital humain . Selon elle, seuls les êtres humains comprennent le contexte, peuvent juger des performances des modèles et décider du rôle que l'IA devrait jouer dans la société .

Elle a ensuite décrit comment son organisation avait mené des évaluations impliquant des milliers de personnes à travers le monde, et a indiqué que le défi consiste désormais à élargir cette participation . Elle a noté que les pays investissent souvent dans des centres de données, des modèles et des talents techniques coûteux, tout en négligeant la couche d'évaluation et les infrastructures nécessaires pour agréger de nombreuses perspectives individuelles en jugements sur l'adéquation et la pertinence . Ses propositions étaient que les gouvernements devraient créer des organes d'évaluation indépendants, opérationnaliser les normes aux niveaux régional et sectoriel, et utiliser la commande publique pour exiger la conformité, de sorte que les entreprises ne puissent pas remporter des marchés publics ou des contrats dans le secteur social sans respecter ces normes . Elle a dit que cela aiderait les pays à passer d'une réception passive de systèmes d'IA externes à la définition de leurs propres priorités, et déplacerait le pouvoir vers les personnes les plus concernées par la technologie . Baird a répondu que l'évaluation est un moyen de renforcer la confiance et la certitude .

Kate Kallot a abordé la souveraineté sous l'angle de l'appropriation locale. Elle a soutenu que « la consultation n'est pas adaptée » et que le partenariat souverain exige une appropriation locale . En prenant Barbade comme exemple, elle a décrit trois pratiques . Premièrement, les données critiques doivent rester dans le pays . Elle a indiqué que son entreprise avait numérisé des millions de documents gouvernementaux, notamment des accords bilatéraux et des documents du conseil des ministres, tout en veillant à ce que cette base de connaissances reste sous juridiction barbadienne sur des infrastructures contrôlées par le gouvernement . Cela a permis au gouvernement d'utiliser les capacités des modèles de pointe sans envoyer ses données les plus sensibles à l'étranger .

Deuxièmement, Kallot a soutenu que le transfert de capacités doit être réel et inscrit dans les contrats . Dans son modèle, les contrats gouvernementaux incluent des obligations de former des gestionnaires de données et des ingénieurs locaux sur une période de deux à trois ans, avec une passation de responsabilités une fois que les équipes locales sont prêtes . Elle a rejeté un simple modèle d'assistance technique et a affirmé que le succès devrait se mesurer à la capacité croissante des équipes locales à s'approprier les systèmes . Troisièmement, elle a dit que les communautés doivent pouvoir construire sur la base de ces systèmes . À Barbade, cela a notamment consisté à mettre en open source « ChatBB » afin que les développeurs locaux puissent l'enrichir, que les jeunes et les data scientists puissent s'y former, et que le contexte local puisse alimenter les services publics et la création de valeur économique . Plus largement, elle a soutenu que les communautés devraient être impliquées avant que les systèmes ne soient codés ou déployés, afin qu'elles contribuent à définir le problème, les cas d'usage, les cas limites et les expériences vécues pertinentes . Pour elle, la co-conception est ce qui rend la gouvernance réelle .

Yi Zeng a dit qu'il existe déjà un large consensus sur les principes éthiques de haut niveau en matière d'IA, notamment à travers l'UNESCO et l'organe consultatif de haut niveau des Nations Unies, mais que la mise en œuvre demeure le problème central . Il a indiqué que ces grands principes avaient été traduits en une centaine de considérations techniques plus précises et utilisés pour tester des modèles de langage . Selon lui, les modèles pouvaient obtenir des scores très élevés sur des questions standard, mais lorsque des questions équivalentes étaient reformulées de nombreuses façons différentes, les performances chutaient fortement, à environ 58 . Il a soutenu que cela démontrait que les modèles de langage applicatifs ne satisfont pas encore, même aux attentes éthiques largement partagées, lorsqu'ils sont soumis à des tests plus rigoureux . Il a également indiqué que les systèmes actuels restent significativement vulnérables aux attaques adversariales .

Yi a ajouté que les scientifiques doivent « dire la vérité » sur l'état actuel de l'IA plutôt que de surestimer ce que ces systèmes sont capables de faire . Il a dit que l'éthique et la sécurité ne sont pas uniquement des questions techniques, et que les scientifiques comme le grand public ont besoin d'une meilleure culture de ce qu'est l'IA et de ses limites . Il a affirmé qu'il n'y a « aucune magie » dans l'IA et l'a décrite comme une machine qui traite l'information sans véritablement la comprendre . Il a également soutenu que le monde a besoin d'une plateforme capable de rassembler les États autour de ces questions, et que l'ONU reste le meilleur cadre disponible, même si elle a besoin de davantage de soutien et de ressources .

Baird a répondu que les êtres humains doivent rester dans la « chaîne de commandement » et que la gouvernance requiert non seulement des ingénieurs, mais aussi des philosophes et d'autres personnes capables d'apporter des perspectives critiques plus larges . Yi Zeng a ensuite ajouté que les voix du public, y compris celles des jeunes générations, devraient être entendues, et a suggéré que certaines applications de l'IA n'ont peut-être pas besoin d'être développées du tout, appelant à la prudence quant aux types d'IA que les sociétés choisissent de créer .

Lors du tour de clôture, Yi a dit que le dialogue devrait produire des résultats concrets, et pas seulement des discussions . Il a soutenu que l'IA devrait être utilisée de manière positive pour contribuer à la réalisation des Objectifs de développement durable dans le temps limité qui reste, tout en appelant à l'établissement de lignes rouges dans le développement et l'utilisation de l'IA . Il a dit que l'éthique et la sécurité devraient contribuer à orienter l'IA dans une direction saine plutôt que d'entraver l'innovation .

Chowdhury a dit que de nombreux participants siègent déjà dans de nombreux groupes consultatifs et jugeraient cette initiative à la valeur qu'elle crée réellement . Bien qu'elle ait réaffirmé la nécessité d'une gouvernance mondiale, elle a critiqué les formes de gouvernance qui se contentent de consolider les pouvoirs existants . Elle a dit que l'année à venir devrait apporter de réels progrès dans l'intégration des pays du Sud global et des pays de la majorité mondiale dans les processus de gouvernance . Dans le même temps, elle a mis en garde contre les formes de souveraineté qui affaiblissent la coopération internationale, plaidant plutôt pour une appropriation et un contrôle qui n'impliquent pas de fermer les portes entre les pays .

Kallot a dit que les discussions mondiales ont été trop fortement influencées par un ensemble restreint d'acteurs et par l'hypothèse selon laquelle l'IA doit suivre une trajectoire centrée sur des modèles très grands et des infrastructures de calcul massives . Elle a appelé la plateforme de l'ONU à aider à « déconstruire et reconstruire » ce récit en montrant ce qui fonctionne réellement dans le Sud global et la majorité mondiale . Elle a également appelé à un soutien pratique et à une plus grande visibilité pour les entreprises et les développeurs qui cherchent à développer l'IA selon les termes de leurs propres pays .

Marwala a dit que la communauté internationale doit « passer des paroles aux actes » . Il a également décrit la gouvernance comme un exercice d'équilibre façonné par le contexte et les valeurs, citant des exemples tels que la transparence versus la sécurité, les données synthétiques versus les données authentiques, et la recherche d'opportunités versus l'aversion au risque . Il a ajouté que les paradigmes actuels de l'IA privilégient la précision plutôt que la vérité, et a suggéré que de nouveaux paradigmes techniques pourraient être nécessaires pour remédier à cette limitation .

Tijani a conclu en revenant sur l'inclusion . Il a dit que l'accès aux bénéfices de l'IA ne découlera pas simplement de l'adoption de systèmes développés ailleurs, et que les pays ont besoin d'infrastructures et de capacités pour construire des solutions adaptées à leur propre contexte . Il a suggéré que la plateforme de l'ONU pourrait contribuer à apporter de la clarté sur des questions pratiques telles que le partage des infrastructures de calcul entre régions et les prérequis dont les pays ont besoin avant de pouvoir véritablement s'approprier l'IA . Il a réitéré son avertissement que, sans investissement dans ces prérequis, le fossé numérique lié à l'IA pourrait devenir plus profond que le fossé créé par l'internet .

Dans ses remarques de clôture, Baird a dit que la discussion avait confirmé que la gouvernance de l'IA ne peut être assurée par un seul acteur ou un seul pays, et qu'elle nécessitera une collaboration soutenue entre les gouvernements, l'industrie, le monde académique, la société civile et la communauté internationale . Tout au long du panel, les intervenants sont revenus à plusieurs reprises sur la mise en œuvre, les capacités nationales, la coordination, l'évaluation, l'appropriation locale, l'inclusion du Sud global et la poursuite de la coopération au sein de l'ONU .

Whitney Baird
Thank you for the kind and warm introduction. I'd like to introduce the U .S. Council for International Business as a voice of U .S. business in the international arena and our member companies provide solutions, innovations, and investments that are indispensable to responding to global challenges Thank you I'm very excited to be here to moderate the multi -stakeholder session of the inaugural UN Global Dialogue on AI Governance. I have really appreciated hearing from states and from civil society today. I think what is so clear with the promise and risks of AI development and AI implementation is that everyone who has a stake in the process needs to be at the table. And with that, I'm really delighted to have the opportunity to be joined by an exceptional panel of experts. And if I can introduce them one by one, the first is His Excellency Bosun Tijani, the Minister of Communications Innovation and Digital Economy from Nigeria. Minister Tijani. Our second panelist... Our second panelist is Tshilidzi Marwala. He's the Rector and Undersecretary General of the United Nations University. Our third panelist is Kate Kallot, who is the founder and CEO of Amini AI. Our fourth panelist is Raman Chowdhury, who is the CEO and co -founder of Humane Intelligence. Raman. And our fifth is Yi Zeng, who is the Professor at the Gaoling School of Artificial Intelligence (GSAI) at Renmin University of China. Welcome, Professor. I'm going to ask each of our panelists a question, and I have asked them. I recognize that we are a bit over time and have a hard stop, so we'll ask each to have their... Their remarks go no more than three minutes. Mr. Minister, let me start with you. Nigeria has been at the forefront of AI governance leadership, and you also sit on the Africa AI Council. I would note that in my previous life with the U .S. Department of State, I have visited incubators in Abuja and with coders in Lagos, and I know what the vibrancy of that community is like in Nigeria. What are the key actions and enabling conditions needed to translate AI governance frameworks and aspirations into measurable outcomes, and where is greater collaboration most urgently needed?
Bosun Tijani
Thank you so much, Madam Chair. Let me start, of course, by appreciating the United Nations for putting together this inaugural gathering. I think the Secretary General of ITU mentioned in your speech that this couldn't be at a better time, and the pace at which it's being done. is one that shows strong, strong leadership, which I think is exactly what we need when we're talking about governance frameworks in artificial intelligence. The first step is that we need to be moving away from governance principles into practical systems because AI is already something that we're living with every day. It's disrupting sectors, complicating business opportunities for both small, medium, and large businesses. So I think as governments or leaders in general, the urgent demand for us to be able to ensure that the governance frameworks that we're putting in place are actually meaningful is for us to start to look inward. As much as we co -create some of these governance frameworks together, I think each country will need to have clear strategies on things like how do we adopt AI in our country and how do we go about measuring what we are also governing as well. So I think it starts from strategy. I think there's also a strong element of the way we look at AI that is focused around capacity building as well. How do we ensure that the government officials that are responsible for implementing these frameworks do have the understanding and they can continuously be moving alongside development in this space as well? So building capacity not just in terms of talent to build AI, but also capacity in the public sector, I think is extremely important in this space as well. But beyond that, I think I've always been a fan of saying as much as we think through governance framework, we also need to think of the absorptive capacity for nations to actually put AI to good use. Because we can see all the development that we want to see in the world. In nations where connectivity is still not sufficient, meaningful, you can see limitation as to how AI is applied and then the ability of people to even govern. in the first place. In spaces where infrastructure is not available, you already limit the application of AI and the ability to also build in context and as such then govern appropriately as well. I think in cases where you also don't have the capacity to develop, where people just have to be consumers and they can't participate in development, you start to see the limitation as well. So for me, I think in practical sense, if we're going to make some of these changes, I think it starts with clear leadership, both political leadership. I think the leaders must have a clear understanding of why this is important. There will be a need for trusted institutions because you can't really have governance systems in place if people don't trust the institutions that will implement the governance framework. We obviously also need to think through carefully how we measure the impact of the framework that we have. We need to adopt whether it serves our purpose. or not. And I think lastly is investment as well. Nations must prioritize investment in artificial intelligence, whether it's investment in compute or investment in talent. The same way we can articulate how much impact good connectivity will make to a nation is the same way we now need to articulate the benefit and the impact of investment in artificial intelligence on the bottom line of every country because this is now the key driver for GDP in most countries as we're seeing. So it's important that we prioritize investment as well. Thank you.
Whitney Baird
Thank you, Mr. Minister. I'm really taken by the use of the word trust. I think that's enormously important. Also capacity building for the public sector and of course the issue of connectivity for those who remain not connected. And I know within the ITU and in the UN as a whole, that access has been a key priority. So, Professor Marwala, you've worked across government, academia, and now the United Nations system. Looking globally, where do you see the biggest disconnect between AI governance discussions and implementation? And what role should universities, scientific institutions, and international organizations play in helping countries build the knowledge, capacity, and institutions needed to govern AI effectively?
Tshilidzi Marwala
Thank you very much for joining this very distinguished panel. Where I see the biggest disconnect is what, and I will borrow from the financial world, is what I call AI governance arbitrage. You have governance models that are happening at the international level. You have governance models that are emerging from industry. You have governance models that are emerging from all our member states. And they are not coordinated. So somebody who is interested in data will go to a country where the regulations are not that strong. And the same goes for compute capability. So we need to find a mechanism in which we can be able to harmonize all this. I think the thing that we need to do is really to educate. We need to educate the people about data, about algorithms. We tend to forget about algorithms themselves. We need to educate them about computing. We need to educate them about how AI can actually be applied in all the applications. And to be able to do that, we have to reassert the values that should underpin the governance models. and for us in the UN we see the values emanate from the Declaration of Human Rights and the UN Charter. We also need to change the behavior of people and I'm not talking about social engineering here and I'm not even talking about behavioral psychology but we need to change them and I think we change them through education. We also need to develop mechanisms to incentivize good behavior around AI especially companies, including companies. What are the mechanisms that make people respect some of these models that we have adopted? And we need to simplify and educate about policies and regulations and standards. I see Dorin is here, they are doing fantastic work on standards but are we taking these standards to the classrooms? So that when people are designing, learning about the design of these things, they have the standards perspective. We also need to talk about how do we educate the legislators so that they can be able to craft legislations that make sense. I have read about the AI Act. I'm interested to see what is happening in my own continent with regards to regulations. We will loss around AI. And finally, there are many institutions that are pushing this forward. The ITU is one of them. Some of this, the United Nations system is another one. And you have all these institutions. I think we need to be able to bring them together at the working level. And this forum is good to bring them together at the dialogue level. But we need to also go to the working level. Thank you very much.
Whitney Baird
Thank you very much. Thank you very much, Professor. I'm really taken by your use of the term arbitrage because I think seriously when you have fragmented and sometimes discoordinated governance mechanisms, it both decreases trust among users, but it also increased costs for practitioners and builders, companies, and others. And there's really, again, what I hear, again, is the need for education in the public sector, legislators and executive branches to be able to effectively provide the kind of leadership. Ramon, I'm going to ask you this from the perspective of someone who is deeply involved in the discourse on responsible AI. What concrete steps can stakeholders take to ensure that all countries and communities can meaningfully participate in AI governance discussions and implementation efforts?
Rumman Chowdhury
Thank you, and thank you all for your attention. And I appreciate that question because the power of artificial intelligence lies not in development but in evaluation because what greater power than the power to define what is good and bad and right and wrong? True AI sovereignty is a combination of good governance and domain expertise, and that would require you to cultivate something that you probably all have in spades, and that's human capital. Only humans can understand context. Only humans can actually define model performance because we understand and we should be part of how we want artificial intelligence to be used in the world. A few years ago, I started Humane Intelligence, a nonprofit organization, and we conducted evaluations all around the world with thousands of people. Today, what I'm working on is Humane Intelligence Public Benefit Corporation, which builds out infrastructure to enable better evaluations. A few years ago, the question was, how can people get involved? Today, the question is, how can that involvement scale? So as every country invests in their AI stack, you think about creating expensive data centers, building expensive models, and hiring expensive talent. Often what is overlooked is the evaluations layer and the infrastructure needed to scale up individual perspectives into defining goodness of fit. Thank you. So what can policymakers do? First is creating independent evaluation bodies. The second is taking all of these standards for evaluation that are being built and operationalizing them from a regional and domain perspective. And then finally, linking these evaluations to procurement. In other words, no company or organization would be able to do that. They would be able to have a contract to build artificial intelligence. for government and social sector unless they have met these requirements. All in all, this enables countries from moving away from being simply a recipient of the AI models being built in other places to directing the priorities of development. And in doing so, it shifts the power dynamic away from a few hegemonic entities and more into the hands of regular people whose lives are impacted by the technology. Thank you.
Whitney Baird
Thank you very much. I appreciate it. And, you know, the evaluation, again, is a way that you build trust and build certainty. And I think a very useful contribution. And thank you. So, Kate, your work focuses on ensuring AI is built with local context, data, and communities in mind. And from your experience, what practical approaches have been most effective in ensuring communities are not simply consulted but become genuine partners in the design, deployment, and governance of AI systems?
Kate Kallot
Thank you for the question, and it's a pleasure to be here. I think really that question underscores a philosophy, an operating philosophy that we've had at Amini, which is that consultation is not fit for purpose and that true sovereignty or sovereign partnerships requires local ownership. And I'll give you an example of how, through our work with governments across the global south, we're able to achieve this through three very concrete things, using as an anchor our work with the government of Barbados. The first is that critical data has to stay in -country. In Barbados, we digitized millions of government records. Every bilateral agreement, every cabinet paper is dating back to 2018. And that intelligent knowledge... base lives in the country under Barbados jurisdiction on an infrastructure that the government controls. And that means that ministers and civil servants can tap into it, but critically they can now access frontier model capabilities without having to ever send their most critical data outside of their border. Countries shouldn't have to pick between capability and control or independence. Then secondly, we've talked a lot about capacity. Capacity actually has to transfer. And every contract that we sign with the government, we make sure that the transfer obligation is built in. And we're not talking about a help desk or customer service. We're actually talking about a team in country which is training data stewards within government, but also engineers on specific pillar of operation so that they can take care of the systems. And we usually do this over the transfer over two to three years. And when a team is ready to own a function completely, hands over the systems entirely. And the goal is for our role to narrow over time and that for us, that sustainability for us is actually a true measure of success. Then thirdly, the community needs to be able to build on top of it. And in Barbados, we open sourced what we call ChatBB, which is an AI assistance that's lowering the access to public services for citizens. And because it's open sourced now, any Barbadian developers can build on top of that infrastructure. We're running training programs for local youth and data scientists. They can now not just use the system, but also contribute and add services, create new applications on top of it, which is bringing more and more of the local context into the system and in turn generating economic value locally. And for me, that's the deeper point because the most effective approach is not bringing solutions and expecting them to work. It's actually bringing communities at the design stage and before every single line of code is returned. or deployed so that they can define the problem with you. They can define their use cases, their edge cases, and bring their lived experience and their local context in. And that's the difference between a community that has shaped what gets built. They don't just own the system. They're actually extending it. And I want to end by saying that local context in the Global South has always been our superpower. And right now, co -design alongside local partnership is what becomes true governance. Thank you.
Whitney Baird
Thank you, Kate. And I just continue to hear such a strong theme of capacity building so that populations can own and benefit and really use, again, the promise that AI can bring. So, Yi, many areas. Many ethical AI principles are already well established. Where do you think the greatest implementation gaps remain, and how can international scientific cooperation help translate ethical principles? into practical governance mechanisms that can be adopted across diverse countries and contexts.
Yi Zeng
Yeah, you're right. I think there are word consensus on the AI ethical principles already, not only in the UNESCO Ethics of AI recommendation, but also about the writings in the Governing AI for Humanities report from my UN high -level advisory body members. Well, when we actually translate these around 10 high -level ethical principles into technical details, we map them into 90 different finer considerations. And then for each and every one of them, actually we ask questions to language models. And you know what was really happening is that when we asked the language model about these technical details, and then they got really good, like 98. But when I populate all the questions, like one question to 100 different ways of asking the same question, what is the result? The result is that they only get 58. What does that mean? That means application -level language models does not pass, even the word consensus. This is the real world status at this point. Not only about ethics. Let me also talk about safety. When we test the common safety concerns, the best one get like for adversarial attack so that to find the safety loopholes, they get like 60 % of the time they will be attacked. This is the best one. The worst one is 40 % of the time they will be attacked. What does that mean? That means application -level language. So language model services for 40 % of the chance. that they're providing very risky services. This is what's really been happening now. So I think as a scientist, we must speak for the truth. It's not only about how do we technically, you know, dig into reality and to provide, you know, this evaluation. It's not only about technology. The way that provide, you know, groundings of these ethical principles also need people. Not only scientists that need to speak for the truth, that need to be balanced, but also we need general public to raise the literacy. We need scientists to help the world and the general public to raise the level of understanding of the current status of AI. There's no magic. It's only a machine that process information but that don't really understand information. Not only about technology, not only about the people, that information, in this stakeholders, but also... about the platform. Let me end by saying that when we wanted a more ethical and more safer AI, five years ago we've been talking every question, each and every question that we discussed today, but none of the questions have been solved within these five years. So what do we need? We also need a platform. I personally cannot find a better platform compared to the United Nations. When our UN advisory body on AI worked, Antonio Goodreads says that UN has not much money, has not much people to help us. So many people were criticizing the United Nations to do a better job. But what I'm thinking is that then offer your help, offer your money, offer your people to create a platform that can help us. No other platforms that can gather together 193 member states all together to consider about these questions in a very serious way. And I hope the next five years we could partially solve the most important challenge. Thank you.
Whitney Baird
Thank you. And I really, I think, a strong call to action, but additionally, really appreciate the focus on the absolute critical need for humans to remain in, as they say, the chain of command. We do. And the skill of critical thinking that we don't just need engineers. In fact, we need philosophers. We need others who really can approach this from the broadest possible perspective on how we deploy and how we use the information and how we trust. And, again, at the end of the day, we come back to trust.
Yi Zeng
Voices from the general public. I like the argument from the Z generation that probably some of the application, we don't have to even have to build that. We should be very cautious what kind of AI we're bringing. We should be very cautious what kind of AI we're bringing. We should be very cautious what kind of AI we're bringing. We should be very cautious what kind of AI we're bringing. We should be very cautious what kind of AI we're bringing. We should be very cautious what kind of AI we're bringing. We should be very cautious what kind of AI we're bringing. We should be very cautious what kind of AI we're bringing.
Whitney Baird
Thank you. Thank you. We have a few minutes left, and what I would really like to do is come back to the panelists for a very brief reflection. In looking ahead, how can this UN global dialogue on AI help address the most pressing gaps in today's AI governance landscape? So what is one concrete outcome you would like to see emerge from this platform over the coming year? And maybe just each person two minutes, and maybe we'll go in reverse order, and I will start with you, Ying.
Yi Zeng
Well, if I may, I think there are two things we need to consider very seriously. I'm going to continue by saying AI ethics is not only about what we shouldn't do, but also what we should do. So for the sustainable development goals, it's only 50%. And of the goals that has been achieved, we only have a very few years left. So can we use AI as an enabling technology to tackle the rest of the challenges in global sustainable development? This is something that we should do, actually, together. Let me also end by saying that for this dialogue, we need some outputs. I agree with you. And if we are talking about ethics and safety of AI, definitely development of AI really needs red lines. Ethics and safety of AI is not a backlash of the development of AI. Actually, it helps you to find the directions of developing AI. When you're getting to the wrong direction and then you're getting a little bit to another way, so that you keep a steady direction. AI development, so the ethics and safety of AI will never hinder the process of AI development and application. It helps AI to develop in a healthy and steady way. Thank you.
Whitney Baird
Maybe Rumman will come to you.
Rumman Chowdhury
Thank you so much, and thank you for the wonderful question again. So there's a few things that I think I would like to see in the next year. You know, I think for all of us sitting on this panel and in this room, at this point we're on so many advisory groups, bodies, panels, and organizations that, you know, we often find ourselves cutting down the kind of engagement that we're doing and focusing on the ones that bring the most value. So what will bring the most value in the next year? In 2023, I wrote an op -ed for Wired where I said AI desperately needs global governance, and that has actually not changed. But what I have seen is a global governance that's built on consolidating the people who are already powerful. These dialogues present an opportunity to maybe reset. the direction that global governance has taken. So one is I would love to see a concrete direction in incorporating and engaging global South and global majority countries. Again, you know, organizations like Kate's and mine have really focused on encouraging voices that are not always the voices in the room. The second thing I'd like to see is a discussion of sovereignty, yes, but sovereignty that doesn't lose the concept of cooperation. We live in a time that's very geopolitically fraught between countries, between regions, and frankly, it feels like we're heading into an era where we're actually rolling back some of the international cooperative agreements that we've seen, not just across countries, but also within technologies. So while I significantly encourage sovereignty and ownership of the models and the modalities in which AI is built, I also want to ensure that we engage in international cooperation rather than closing doors to each other. Thank you.
Whitney Baird
Thank you, Rumman. Kate.
Kate Kallot
I'm going to echo a little bit what has been said and I think we need to one of the outcomes I'd love to see over the next year is a place where we deconstruct and reconstruct the AI narrative because for the past couple of years that narrative has been singly driven from a very specific part of the world by very specific actors and they've taught us that or made us believe that the only way to do AI is large scale general purpose models and large scale gigawatt factories which is not the case and it's been a pleasure sitting through the outcomes and the evidence base of the past that were shown this morning because we are now hearing exactly what is going on and we are actually getting evidence of what is working or not on the ground. And I think we need more of those stories to be surfaced throughout the Global South and the global majority. And then the second thing, I think we need to support the companies who are trying and the builders who are trying in their own countries to push that agenda and help their countries develop themselves and look at AI in a different way. Give them a platform to surface their innovation. Support them on being on stages like this one because they are the ones who will be the guardians of what AI becomes over the next couple of years. And it's not just going to be the people in this room. It is going to be the developers, the communities on the ground, and they have to be part of the conversation. Thank you.
Whitney Baird
Professor.
Tshilidzi Marwala
Thank you very much. If there is just one thing that... I would like to see is to move from talk to action. We really have been talking quite a great deal of taking the technology to the global south, democratizing access to the technology, but there has to be action, concrete action that requires all of us to come together. And the second issue that I really think is important in taking moving from talk to action, is that governance is really about a balancing problem. Professor Yi was talking about risk aversion. It's really about a balance between opportunity seeking and risk aversion. And where that balance should be, it's not very clear. It depends on many many values that are varying and I think if we could be able to be helped to find the right balance and there are many places where this balance is necessary transparency versus security if we have to make AI explainable and transparent it comes at the expense of security where is the balance who determines it another one which was also mentioned earlier there is quite a great deal of AI that is being trained using synthetic data what is the balance where do we draw the line between the use of synthetic data versus the use of authentic data and who determines that and then there is the issue of truth truth is a very controversial subject in AI if you go to an AI course nobody talks about truth they talk about accuracy and the reason why they talk about accuracy is because the dominant version of AI which is based on back propagation assumes that you can differentiate a function and if you can differentiate a function it can't be discrete like truth and that is why as long as we do not find a new paradigm then we are always going to be talking about this we need to invest in new paradigms to deal with algorithmic limitations of AI thank you applause thank you
Whitney Baird
Minister Tijani the last word to you thank you
Bosun Tijani
so much for giving me the opportunity to chime in last I think on this one for me it will be one thing which I'll frame around inclusion and I know that inclusion can manifest itself in many ways but for me it will be inclusion when it comes to access to the benefit of artificial intelligence applause I think we need to be clear that access will not come only from just merely adopting what's been built elsewhere. We need to ensure that nations have access to capabilities, particularly infrastructure, that will allow them to be able to build for context. I think that context is what I believe will drive the benefits that we want to see countries make from access to AI. And this platform is one platform that can help bring clarity to that. Whether we're talking about clarity on access to shared compute, for instance, within regions amongst closer countries or beyond. Or we're talking about how we deepen what I mentioned earlier, the critical prerequisite for nations to use AI. Because I think we're not talking enough about the fact that there are prerequisites. There are things that must be in place for nations to truly appropriate. And if we don't invest in that, if we don't find ways to drive investment and support for this infrastructure. The gap that we're going to see with artificial intelligence will be bigger than what we were battling with when we were talking about the emergence of Internet. This is going to be bigger because you have those who have access to AI and have used it more diligently and those who don't have access just because the infrastructure is not there. Thank you. Thank you. Thank you
Whitney Baird
all. Thank you, everyone, for joining us today for our first moderated panel of the day and to all of our intervenants. I think today's conversation has reinforced that AI governance cannot be achieved by any single stakeholder or country, and progress will depend on sustained collaboration across governments, industry, academia, civil society, and the international community. So thank you all, and I hand the program back to our master of ceremonies. As I believe you all know, I'm going to be doing a lot of work on the AI governance side of things, and I'm going to be doing a lot of work on the AI governance side of things. I believe we break for lunch. Thank you very much.
AI diplomacy
Privacy and data protection are particularly pertinent, given that AI systems often require massive datasets, which can include sensitive personal information. It is crucial that developers embed privacy guarantees and d...
Searching for Standards: The Global Competition to Govern AI | IGF 2023
Regulatory sandboxes, particularly in the fintech sector, are highlighted as a useful technique for fostering innovation. The Monetary Authority of Singapore has utilized regulatory sandboxes to reduce risks and enable t...
AI, Data Governance, and Innovation for Development
She also stresses the need to harmonize existing regulations rather than creating new ones for every new technology. Preuves Example of Nigeria's national AI strategy development process involving multiple sectors i...
AI supremacy: One, two, three, go!
Legend has it that in the 13th century Dominican friar Albertus Magnus built an artificial ‘brazen head’ that could talk and answer any question asked. According to story, his student, philosopher Thomas Aquinas, eventua...
[Panel Discussion] Global AI Policy Coordination
The level of disagreement was moderate and unlikely to prevent collaboration, as most intervenants shared underlying values about human-centric AI development and global cooperation Partial agreements Partial agree...
Navigating the interplay between artificial intelligence, philosophy, education, and governance
Overview The rapid development of artificial intelligence (AI) presents profound philosophical, educational, and governance challenges and opportunities for the global community. To address these dimensions, we organis...
The year of AI clarity: 10 AI Forecasts for 2025
Most nations seek a balance, allowing the use of protected data and knowledge while acknowledging ownership, often through opt-out clauses. Algorithms: In 2025, the assumption that AI governance should focus mainly on ...
Inclusive AI governance: Universal values in a pluralistic world
These values can inform governance models that prioritise relational accountability, ethical cultivation, and social cohesion, offering alternatives to transactional, compliance-driven frameworks.This is why I dare here ...
Digital sovereignty stack: Infrastructure, services, data, and AI knowledge
In March 2026, Brazil's Digital Child and Adolescent Statute (Digital ECA) came into force, imposing structural obligations on platforms to protect children and adolescents online. Brazil is strengthening its digital sov...
The New Delhi AI Summit: Inclusive rhetoric, fractured reality
In the words of UN Secretary General, António Guterres, “If we want AI to serve humanity, policy cannot be built on guesswork. It cannot be built on hype – or disinformation. We need facts we can trust – and share – acro...
The MANAV manifesto: Reclaiming agency for the majority
In practice, this means prioritising voice-to-action interfaces that allow a person to navigate complex public services (accessing government benefits, filing a complaint, checking land records) through their local langu...
Safe, secure, and trustworthy AI: What is it and how do we get there?
The value proposition of the UN GPs on BHR https://diplo-media.s3.eu-central-1.amazonaws.com/2025/10/IMG_4122.jpeg https://diplo-media.s3.eu-central-1.amazonaws.com/2025/10/IMG_4120.jpeg https://diplo-media.s3.eu-centra...
Keeping AI in check
A ten-step guide published by CoE starts with the need to conduct a human rights impact assessment on AI systems. Technology is a social product, and as such, it integrates values that orientate the way it operates, ev...
AI in 2026: Learning to live with powerful systems
The past few years have been defined by astonishment. Each new AI release seemed to arrive faster than society could absorb its implications. Systems grew more capable, outputs more convincing, and public reactions more ...
Unpacking the High-Level Panel’s Rapport on Digital Cooperation: Geneva policy experts propose action plan
In the era of ‘trust deficit’, the legitimacy of the help desk is of high relevance especially for actors from small and developing countries. Action: Ensure that help desks are associated to, or linked with, the UN syst...
Leveraging the UN system to advance global AI Governance efforts
But my worry is that when the conversations are within a small group of countries and you leave out the majority of where the population is, you are finding solutions that people get alienated from because they're not in...
From summer disillusionment to autumn clarity: Ten lessons for AI
One example is the push to develop domestic GPU chips after US sanctions. China is in catch-up and bypass mode – using whatever it takes (including open-source innovations like DeepSeek and large state R&D programmes) to...
Global AI governance: Reflecting on 2024 and shaping the path for 2025
On 9 January 2025, DiploFoundation, in collaboration with the Permanent Missions of China, France, Kenya, Mexico, Pakistan, Switzerland, and the United States to the United Nations in Geneva – as co-sponsors – hosted an ...
1

This is supported by a prior statement from Whitney Baird describing the US Council for International Business as representing the U.S. private sector in multilateral and international organisations and companies with a global presence [S85].

2

The knowledge base consistently supports this multi-stakeholder framing for AI and digital governance, including calls for participation by governments, business, civil society, academia, and other actors [S83], [S93], [S95].

3

This aligns with repeated knowledge base emphasis on moving from principles to action and implementation, including practical pathways for translating digital governance principles into action and creating enabling policy environments grounded in real-world experience [S103], [S87].

4

The knowledge base explicitly discusses the need to go 'from principles to action' in digital governance and warns against producing too many declarations without implementation [S103].

5

The knowledge base adds useful context by noting that national strategic frameworks on AI and cybersecurity are important for governance and implementation, reinforcing the relevance of country-level strategies [S104]. It also notes that many countries are reforming institutions to address digital challenges more holistically [S81].

6

The knowledge base corroborates several of these implementation conditions: diplomats and officials need AI understanding for effective policy work [S40]; countries need broader capacity building and access to opportunities [S40]; and trust is repeatedly identified as foundational in digital governance [S100].

7

This is well supported by the knowledge base. Multiple sources state that connectivity and digital capacity are prerequisites for meaningful participation in the digital economy and for broader digital development [S83], [S87], [S99].

8

The knowledge base adds relevant nuance that lack of connectivity and digital capacity hinders full participation in the digital economy [S83], and that access to capacity-building opportunities often fails to reach communities and companies in the Global South [S40].

9

The knowledge base does not confirm the comparative prediction directly, but it strongly supports the underlying concern about widening digital and socio-economic inequalities from AI and the persistence of earlier digital divides [S82], [S92], [S95].

10

Trust, capacity, and connectivity are all prominent themes in the knowledge base: trust is described as a prerequisite for effective digital public infrastructure [S100], connectivity as a foundational enabler of digital participation [S87], and capacity building as essential for AI and digital governance [S40], [S83].

11

The knowledge base provides strong contextual support for this concern about fragmentation. It notes that cross-cutting digital issues cannot be addressed effectively in silos [S81], and that regulatory fragmentation can undermine coherence, trust, and practical implementation [S86], [S87].

12

The knowledge base supports this by stating that regulatory fragmentation undermines the open nature of the internet, innovation, and creates barriers for businesses and consumers [S86]. It also links trust to effective governance and implementation [S100].

13

This is consistent with knowledge base material stressing AI literacy, education reform, and broader inclusion of educators and non-technical communities in AI governance discussions [S68], [S104].

Leveraging the UN system to advance global AI Governance efforts — It commenced with the recognition of the urgency and significance of the subjects, particularly in light of the impending departure of some attendees.The introductory remarks set a cooperative tone for the panel, utilisi...
Shikoli Makatiani — Shikoli Makatiani Chief Technical and Operations Officer of Serianu Kenya Ltd https://diplo-media.s3.eu-central-1.amazonaws.com/2023/12/Shikoli-Makatiani.jpg Shikoli Makatiani, the Chief Technical and Operations Officer ...
South Africa to establish AI expert advisory council to shape future regulations — South Africa is about to create an AI Expert Advisory Council in an effort to address the regulatory issues surrounding the use of AI. This initiative was highlighted at a national AI summit organised by the Department o...
Katharina Ziolkowski — Katharina Ziolkowski
Jennifer Kates — Jennifer Kates
Katarina Bojović — Katarina Bojović https://dig.watch/wp-content/uploads/katarina-andjelkovic.jpg Ms Katarina Bojović is a lecturer in e-diplomacy and internet governance courses and a researcher and coordinator of the sociocultural basket...
Innovation, entrepreneurship and Africa’s digital economy: Driving AfCFTA delivery through digital innovations and technology (Dynamics Impact Advisory) — He is the owner of over 30 estates, each having between 50 to 500 acres. Topics: Entrepreneurship, Job Creation, Digital Economy Stephen Akintayo advocates for reforms in land use and ownership in Africa. Supp...
WAIGF Opening Ceremony & Keynote — Dr. Bosun Tijani, Minister of Communication, Innovation and Digital Economy, your leadership has helped elevate the role of digital transformation in Nigeria's national development. and has inspired the entire region to ...
Software.gov — The country has transitioned from physical wallets to digital wallets, and the development of the Unified Payment Interface has revolutionized the way people make transactions. These advancements have contributed to grea...
Liang Wenfeng and the rise of China’s AI industry — Liang Wenfeng, the 39-year-old founder of DeepSeek, has rapidly become a central figure in China’s AI ambitions. Previously maintaining a low profile, he gained national attention after being invited to a closed-door sym...
John le Carré: The Biography — In any case, who is perfect? These failings – if such they are – pale in comparison with Le Carré’s skills and industry as a novelist, his later acts of charity, and his willingness to court critical attack by using his ...
Rumman Faruqi — Rumman Faruqi
IGFSA | SIDE EVENT — The intervenants explored the potential of artificial intelligence (AI) and blockchain in various sectors such as healthcare, education, agriculture, and finance. They highlighted the opportunities and challenges associated ...
Work for a brighter future — Professor General for Human Resources and Social Policy Chung has also served as Member of the UN (2002–08) at Norilsk Nickel Mining and Metal- Sub-Commission on the Promotion and Prot...
Diane Whitehead — Diane Whitehead
Brian Barder — Brian Barder
Artificial intelligence in Africa: National strategies and initiatives — (2019). Emerging Digital Technologies for Kenya. The 2022–2032 Digital Master Plan contains extensive references to AI. It starts from acknowledging that ‘AI technologies and capabilities will be the in thing in the ...
Governing AI for Humanity | Final Rapport — Figure 16: Experts’ ratings of barriers to harnessing AI to drive additional economic activity and progress on the SDGs in lower-middle/lowerincome countries These results underline the tentativeness of AI’s eventua...
Open Forum #58 Collaborating for Trustworthy AI an Oecd Toolkit and Spotlight on AI in Government — Now we do need a skilled workforce so capacity development and deployment is something that we absolutely need. Now one of the big constraints that also come with that is that kids do not grow at the speed that artificia...
[Panel Discussion] Managing Digital Transition - Policy and Capacity Building — 2025 we celebrate 200 years of setup of train infrastructure and creation of trains so in 200 years we created trains we also created training incidents and the whole infrastructure to manage around the risks and impact ...
Keeping AI in check — A ten-step guide published by CoE starts with the need to conduct a human rights impact assessment on AI systems. Technology is a social product, and as such, it integrates values that orientate the way it operates, ev...
Artificial intelligence (AI) and the human condition — We already witness the emergence of new technology-enabled illegitimate powers across borders, ideologies, and classes. Power corrupts power holders! While we can assume that a possible rebellion of artificial intelligen...
Inclusive AI governance: Universal values in a pluralistic world — These values can inform governance models that prioritise relational accountability, ethical cultivation, and social cohesion, offering alternatives to transactional, compliance-driven frameworks.This is why I dare here ...
The year of AI clarity: 10 AI Forecasts for 2025 — China: China has implemented strict regulations requiring platforms to label AI-generated content, especially deepfakes, and to obtain consent from individuals before using their likenesses. Which practices do social ...
Jua Kali AI: Bottom-up algorithms for a Bottom-up economy — As artificial intelligence (AI) becomes a cornerstone of the global economy, AI’s foundations must be anchored in community-driven data, knowledge, and wisdom. 'Bottom-up AI' should grow from the grassroots of society in...
AI in Africa: Beyond the algorithm — Whose it is serving when we still have a compute divide? When you think about the infrastructure that's needed to train AI systems, data center, 90% of the world's data center capacity still sits in two countries, US and...
AI, smart cities, and the surveillance trade-off — The algorithm identifies patterns and extrapolates them into the future. This sounds rational until you consider why those patterns exist in the first place. Neighbourhoods that were historically neglected may show hig...
Africa – In the AI era — 'Africa's data infrastructure needs to be more developed. The entire continent has fewer data farms than the Netherlands, a stark indicator of the digital divide.' In this article, Dr Jovan Kurbalija looks at Africa's ...
Four seasons of AI:  From excitement to clarity in the first year of ChatGPT — How to address AI risks There are three main types of AI risks that should shape AI regulations: the immediate and short-term ‘known knowns’ the looming and mid-term ‘known unknowns’ and the long-term yet dis...
AI and the Roadmap for Digital Cooperation — Commentators, such as Mr Wolfgang Kleinwächter, issued words of caution regarding the potential of digital collaboration, arguing that the secretary general's call for a global ban on LAWS ‘is falling on deaf ears’ and t...
Rethinking learning: Hope, solutions, and wisdom with AI in the classroom — Some intellectual struggle is essential for development. This might mean specific assignments done without AI assistance, certain skills practised until they become automatic, and certain knowledge that must be internali...
From summer disillusionment to autumn clarity: Ten lessons for AI — Many argue we should place emphasis on what makes us human: critical thinking, creativity, ethics, the ability to formulate original questions and to judge the outputs of AI. The role of a teacher will shift from info...
Reclaiming the public mission of higher education in the age of AI — Beyond merely offering classes on the technology, the institution has organised itself to support AI-enabled inquiry in fields ranging from the arts to agricultural sciences, demonstrating an ambition to build foundation...
AI@UN: Navigating the tightrope between innovation and impartiality — 'We, the peoples,' the first line of the UN Charter, should guide the development of AI at the UN. Contributions from countries, companies, and communities worldwide can enhance AI’s potential to support the UN’s mission...
Ethics and AI | Part 2 — Ethics for AI beginners. Business as usual or an inflexion point? An incursion between various definitions of ethics does not help too much those who try to depict – in one single pronouncement – a clear landscape of t...
Ethics and AI | Part 4 — Principles for the Ethical Use of Artificial Intelligence in the United Nations system: New wine in old bottles? The United Nations System Chief Executive Board for Coordination (CEB)1 produced, starting from the work ...
Unpacking the High-Level Panel’s Rapport on Digital Cooperation: Geneva policy experts propose action plan — In the era of ‘trust deficit’, the legitimacy of the help desk is of high relevance especially for actors from small and developing countries. Action: Ensure that help desks are associated to, or linked with, the UN syst...
Global AI governance: Reflecting on 2024 and shaping the path for 2025 — On 9 January 2025, DiploFoundation, in collaboration with the Permanent Missions of China, France, Kenya, Mexico, Pakistan, Switzerland, and the United States to the United Nations in Geneva – as co-sponsors – hosted an ...
[Panel Discussion] Global AI Policy Coordination — written cultures, resource constraints, and different privacy concepts The traditional Global North/South divide is becoming less relevant; the real distinction is between powerful and less powerful actors in AI govern...
WS #466 AI at a Crossroads Between Sovereignty and Sustainability — This concentration limits broader access to AI innovation and development capabilities. Preuves LLMs like GPT and LAMA are developed by big tech companies, not universities or other technical institutions; researche...
AI diplomacy — AI also raises concerns regarding safety, security, and privacy. AI-driven systems, such as autonomous vehicles, must be designed to safely handle unforeseen situations, and the cybersecurity risks associated with AI tec...
From summer disillusionment to autumn clarity: Ten lessons for AI — One example is the push to develop domestic GPU chips after US sanctions. China is in catch-up and bypass mode – using whatever it takes (including open-source innovations like DeepSeek and large state R&D programmes) to...
Digital sovereignty: The end of the open internet as we know it? (Part 1) — In the context of an offensive and chauvinist turn in US policy, the popular magazine The Economist suggested a range of potential choke points that could be adopted by the EU in retaliation. They included restricting ac...
20 Keywords for the Digital 2020s: A Digital Policy Prediction Dictionary — Policymakers must recognise it too, and respond with cross-silo policy approaches. Multilevel approaches should bring digital policy issues closest to people and communities who are directly concerned with these is...
Internet Governance Forum 2025 — So even across borders, countries can keep their data locally. And then through API, you can share the specific data categories that have been agreed upon based on policy framework." Data Categorization Approaches Severa...
Network Session: Digital Sovereignty and Global Cooperation | IGF 2023 Networking Session #170 — Additionally, the collision between a country's desire for control and the need for collaboration adds another layer of complexity.The current global governance structures may not effectively handle digital cooperation i...
[Session Discussion] On Sovereign AI: From Theory to Action — This lack of disagreement may indicate either broad consensus on the need for AI sovereignty, or that the session was structured more as a presentation than a debate. The implications are that while there may be underlyi...
Artificial intelligence: policy implications — It also calls for the development of standards for the concepts of privacy by design, privacy by default, informed consent, and encryption in AI systems. Ethics As AI systems involve judgements and decision-making - repl...
Ethics and AI | Part 6 — Mechanisms should be put in place afterwards to allow for external feedback on any potential infringement of fundamental rights. Human agency should be ensured, i.e. users should be able to understand and interact with...
AI and international peace and security: Key issues and relevance for Geneva — This, in turn, may lead to more confidence in both cyber and data security. Beyond the virtual sphere, AI could play a role in augmenting physical safety and security to a greater extent than would be possible through hu...
[Panel Discussion] Regulation by Design - Engineer Accountable AI Agents — cooperative international stances to optimistic vs. cautious implementation timelines. These disagreements reflect deeper tensions between innovation imperatives and safety/democratic accountability concerns that could s...
Policy Network on Artificial Intelligence | IGF 2023 — Topics: AI education, AI Governance, Public Sector We need different kinds of AI expertise working together to ensure inclusive and fair global AI governance. Topics: Diversity and Inclusion, AI Governance Ca...
Africa and Internet governance: At the table or on the menu? — Internet governance (IG) is a particularly challenging puzzle when viewed from the perspective of countries in the Global South. Many of these countries are asymmetrically integrated and constrained by structures of depe...
Diplomatic policy analysis — Overdependence on algorithms without critical human oversight can lead to biased or incomplete conclusions, particularly in complex, nuanced scenarios. Digital divides: Not all countries have equal access to advanced an...
Day 0 Event #248 No One Left Behind Digital Inclusion As a Human Right in the Global Digital Age — Certain population groups have been identified as being at risk of digital exclusion. And this slide provides a closer look at the various groups identified as at risk of digital exclusion. And at Nordregio, we have look...
The year of AI clarity: 10 AI Forecasts for 2025 — China: China has implemented strict regulations requiring platforms to label AI-generated content, especially deepfakes, and to obtain consent from individuals before using their likenesses. Which practices do social ...
AI and the Roadmap for Digital Cooperation — Second, values that should guide the development and application of AI were touched upon in almost all interventions. Mr Max Tegmark (President, Future of Life Institute) expressed a key concern, shared by many, when he ...
AI@UN: Navigating the tightrope between innovation and impartiality — 'We, the peoples,' the first line of the UN Charter, should guide the development of AI at the UN. Contributions from countries, companies, and communities worldwide can enhance AI’s potential to support the UN’s mission...
The New Delhi AI Summit: Inclusive rhetoric, fractured reality — When India hosted the AI Impact Summit in New Delhi, from 16-20 February, it seized the moment to demonstrate its growing influence in the digital and AI field. But beyond India's own ambitions, the gathering also served...
WS #462 Bridging the Compute Divide a Global Alliance for AI — Instead, addressing the compute divide requires both expanding overall global capacity and ensuring more equitable distribution of resources.

Barriers to Equitable Access #

Infrastructure and Investment Challeng...

AI diplomacy — Privacy and data protection are particularly pertinent, given that AI systems often require massive datasets, which can include sensitive personal information. It is crucial that developers embed privacy guarantees and d...
Searching for Standards: The Global Competition to Govern AI | IGF 2023 — Regulatory sandboxes, particularly in the fintech sector, are highlighted as a useful technique for fostering innovation. The Monetary Authority of Singapore has utilized regulatory sandboxes to reduce risks and enable t...
AI, Data Governance, and Innovation for Development — She also stresses the need to harmonize existing regulations rather than creating new ones for every new technology. Preuves Example of Nigeria's national AI strategy development process involving multiple sectors i...
AI supremacy: One, two, three, go! — Legend has it that in the 13th century Dominican friar Albertus Magnus built an artificial ‘brazen head’ that could talk and answer any question asked. According to story, his student, philosopher Thomas Aquinas, eventua...
[Panel Discussion] Global AI Policy Coordination — The level of disagreement was moderate and unlikely to prevent collaboration, as most intervenants shared underlying values about human-centric AI development and global cooperation Partial agreements Partial agree...
Navigating the interplay between artificial intelligence, philosophy, education, and governance — Overview The rapid development of artificial intelligence (AI) presents profound philosophical, educational, and governance challenges and opportunities for the global community. To address these dimensions, we organis...
The year of AI clarity: 10 AI Forecasts for 2025 — Most nations seek a balance, allowing the use of protected data and knowledge while acknowledging ownership, often through opt-out clauses. Algorithms: In 2025, the assumption that AI governance should focus mainly on ...
Inclusive AI governance: Universal values in a pluralistic world — These values can inform governance models that prioritise relational accountability, ethical cultivation, and social cohesion, offering alternatives to transactional, compliance-driven frameworks.This is why I dare here ...
Digital sovereignty stack: Infrastructure, services, data, and AI knowledge — In March 2026, Brazil's Digital Child and Adolescent Statute (Digital ECA) came into force, imposing structural obligations on platforms to protect children and adolescents online. Brazil is strengthening its digital sov...
The New Delhi AI Summit: Inclusive rhetoric, fractured reality — In the words of UN Secretary General, António Guterres, “If we want AI to serve humanity, policy cannot be built on guesswork. It cannot be built on hype – or disinformation. We need facts we can trust – and share – acro...
The MANAV manifesto: Reclaiming agency for the majority — In practice, this means prioritising voice-to-action interfaces that allow a person to navigate complex public services (accessing government benefits, filing a complaint, checking land records) through their local langu...
Safe, secure, and trustworthy AI: What is it and how do we get there? — The value proposition of the UN GPs on BHR https://diplo-media.s3.eu-central-1.amazonaws.com/2025/10/IMG_4122.jpeg https://diplo-media.s3.eu-central-1.amazonaws.com/2025/10/IMG_4120.jpeg https://diplo-media.s3.eu-centra...
Keeping AI in check — A ten-step guide published by CoE starts with the need to conduct a human rights impact assessment on AI systems. Technology is a social product, and as such, it integrates values that orientate the way it operates, ev...
AI in 2026: Learning to live with powerful systems — The past few years have been defined by astonishment. Each new AI release seemed to arrive faster than society could absorb its implications. Systems grew more capable, outputs more convincing, and public reactions more ...
Unpacking the High-Level Panel’s Rapport on Digital Cooperation: Geneva policy experts propose action plan — In the era of ‘trust deficit’, the legitimacy of the help desk is of high relevance especially for actors from small and developing countries. Action: Ensure that help desks are associated to, or linked with, the UN syst...
Leveraging the UN system to advance global AI Governance efforts — But my worry is that when the conversations are within a small group of countries and you leave out the majority of where the population is, you are finding solutions that people get alienated from because they're not in...
From summer disillusionment to autumn clarity: Ten lessons for AI — One example is the push to develop domestic GPU chips after US sanctions. China is in catch-up and bypass mode – using whatever it takes (including open-source innovations like DeepSeek and large state R&D programmes) to...
Global AI governance: Reflecting on 2024 and shaping the path for 2025 — On 9 January 2025, DiploFoundation, in collaboration with the Permanent Missions of China, France, Kenya, Mexico, Pakistan, Switzerland, and the United States to the United Nations in Geneva – as co-sponsors – hosted an ...
The US is pursuing holistic digital diplomacy — Following the 9/11 terrorist attacks, US intelligence agencies were fundamentally reformed in order to address fragmentation and policy silos as major reasons for their failure to prevent these attacks. It is now the tur...
Open Forum #81 What Gets Measured Gets Addressed How the US Measures Internet Use — President Clinton and Secretary Daley released Falling Through the Net on July 8th, 1999 in Los Angeles during the President's New Markets Tour. And during the tour, the President and Secretary Daley discussed the fact t...
WS #278 Digital Solidarity & Rights-Based Capacity Building — We recognize the lack of connectivity or digital capacity, and I know all of you are working on this, and these things can hinder the ability to fully participate in the digital economy and challenges our collective ab...
The US State Department and 21st Century Statecraft — Hilary Clinton launched the US State Department's 21st Century Statecraft initiative earlier this year, using of course social media to spread the word and engage the different target audiences of Public Diplomacy. This ...
Leaders TalkX: Partnership pivot: rethinking cooperation in the digital era — Lori Schulman: Thank you thank you so much and last but certainly not least the United States Chamber of International Business with Miss Whitney Baird she is their president and CEO and you get the final word today whic...
Day 0 Event #166 Review of IGF2023 Kyoto and road for WSIS+20 — I'll be shameless and plug in a little bit about the International Chamber of Commerce if For those of you who might not know us ICC is a global business organization We represent over 45 million businesses of all sect...
Morning session — I now give the floor to the observer of the International Chamber of Commerce. International Chamber of Commerce Excellencies, distinguished delegates, ladies and gentlemen, I have the honor to speak on behalf of ...
International Chamber of Commerce — The International Chamber of Commerce is the largest, most representative business organization in the world. Its over 45 million members in over 100 countries have interests spanning every sector of private enterprise.T...
Commercial diplomacy and international business — https://www.clingendael.org/sites/default/files/pdfs/20070400_cdsp_diplomacy_kostecki_naray.pdf https://www.diplomacy.edu/wp-content/uploads/2021/06/clingendael-2018.png
Business-to-business e-commerce: Trade Implications — Among other things, the aim is to develop an inventory of case studies, to unify concepts and definitions, to develop conceptual frameworks, and to engage private industries in data supply and studies. The present data o...
E-commerce in the WTO: the next arena of Internet policy discussions? — In the second circle, there are organisations dedicated to other policy issues that affect e-commerce. To name a few, the World Intellectual Property Organization (WIPO) addresses intellectual property aspects, the Inter...
Addressing AI-driven Socio-economic Inequality — In March 2020, entrepreneur and billionaire Elon Musk stated that 'AI is far more dangerous than nukes'. This wasn't he first time the adverse effects of the scale of AI deployment has been discussed. Throughout the last...
Main Session on Artificial Intelligence | IGF 2023 — I think, as well, that maybe the International Atomic Agency is really difficult as an analogy because we have so many uses of artificial intelligence. I would like to bring back, again, the idea of more of an independen...
AI for Good Global Summit — And then we need to be the silence change agent, right? Maybe we have to say it, but do it more, right? I think that's gonna be an interesting time. We'll repeat it as in Haiti, Haitian Creole, which is not a standard la...
AI promises, ethics, and human rights: Time to open Pandora’s box — In 2021, I participated in the Artificial Intelligence online course offered by Diplo. In one of our online sessions, a passionate debate flared on the topic of the day: AI’s ethical and human rights implications. The di...
Unpacking the High-Level Panel’s report — Less than a week after the launch of the UN High-Level Panel on Digital Cooperation's report, the digital policy community in Geneva gathered to discuss ways of implementing the Panel's recommendations. During the discus...
AI/Gen AI for the Global Goals — We know that the UN is also launching an AI strategy. I'm not sure if it's been launched and it's coming up very soon. And as a result, we came up with this policy brief. The policy brief is on AI and user sustainabilit...
High-Level sessions: Setting the Scene – Global Supply Chain Challenges and Solutions — We usually say that in complex times, only complexity can save us. That is why we are very lucky to count on so many tough leaders, policymakers, CEOs, ministers, heads of agencies here with us this morning. The solution...
WSIS 2018 - Moderated high-level policy session 1 — He explained that the G15 is currently working on four main areas of co-operation: migration, intellectual property, renewable energy, and ICTs. He highlighted that these types of small-group arrangements have many benef...
Pre 11: Freedom Online Coalition’s Principles on Rights-Respecting Digital Public Infrastructure — We didn't see a view that proposes the power imbalances between different actors in the implementation of DPI...digital infrastructure can be both more malleable than physical infrastructure, but can also be more brittle...
The State of the Humanitarian System — Corruption risks were needed. For example, Harnemeijer and Meeux (2007) also highlighted in beneficiary consultations (CDA note huge differences in costs per beneficiary for 2005; 2006a; 2...
BREAK OUT ROOM 2: The Declaration for the Future of the Internet: Principles to Action — So I'm actually talking about self-awareness within technical communities that we've got to better sell ourselves and our missions and our broader missions to our stakeholders, and hopefully with the support of others wh...
Is it the future yet? — These range across all 17 of the United Nations Sustainable Development Goals and could potentially help hundreds of millions of people worldwide. Real-life examples show AI already being applied to some degree in about ...
A tipping point for the Internet: 10 predictions for 2018 — Updates: Jurisdiction | Privacy and data protection | Convergence | Copyright | E-commerce | Intermediaries 5. Artificial intelligence: Between philosophical considerations and practical applications AI trigge...
How David outwits Goliath in the age of AI? — From bigger is better to smaller is smarter. Last week, as OpenAI touted its USD 500 billion ambitions in a high-profile White House announcement, a quiet revolution unfolded thousands of miles away when a Hangzhou sta...
AI in Action: When technology serves humanity — The technology is explicitly designed to help those in greatest need first, under human oversight and accountability. AI does not decide who is worthy; it accelerates the process of learning where human help is most urge...

Avertissement : Il ne s'agit pas d'un compte rendu officiel de la session. DiploAI génère ces ressources à partir d'enregistrements audiovisuels ; elles sont présentées telles quelles, y compris d'éventuelles erreurs. En raison de contraintes logistiques (audio/vidéo ou transcriptions), les noms peuvent être mal orthographiés. Nous nous efforçons d'être aussi précis que possible.