Multi-stakeholder plenary meeting with a high-level governmental segment: Dialogue of Dialogues - Reflections, Complementarities and the Way Forward & Closing Ceremony

26 intervenants
Résumé

Résumé

Cette discussion constituait la plénière conclusive du « Dialogue des Dialogues » de l'inaugural Dialogue mondial des Nations Unies sur la gouvernance de l'IA, convoquée pour synthétiser les messages clés issus de deux jours de délibérations et tracer une voie vers le second dialogue mondial prévu à New York en mai 2027 . Les intervenants ont convergé vers l'idée que la trajectoire de l'IA sera déterminée non pas par la seule capacité technologique, mais par les choix humains en matière de pouvoir, d'accès et de responsabilité , et que des principes sans mise en pratique contraignante risquent d'offrir un faux sentiment de sécurité . Une préoccupation récurrente portait sur le fossé croissant entre le Nord et le Sud global en matière de diffusion de l'IA et de capacité de gouvernance. Brad Smith a noté que l'adoption de l'IA s'élève à 27 % dans le Nord global contre seulement 15 % dans le Sud global, avertissant que les tendances actuelles pourraient doubler cette disparité en l'espace d'un an . Raman Jit Singh Chima a mis en garde contre le fait que la gouvernance de l'IA doit s'appuyer sur les processus existants de gouvernance numérique multi-parties prenantes - plutôt que de les remplacer - et a souligné que les femmes, les filles et les communautés marginalisées sont touchées de manière disproportionnée et sous-représentées dans le développement de l'IA . Linda Bonyo a mis en évidence que les taux de refus de visa, affectant particulièrement les Africains, excluent structurellement les voix du Sud global des forums de gouvernance . Les panélistes ont identifié plusieurs priorités pour la coopération internationale. Guy Ryder a souligné la légitimité unique de l'ONU en tant que convocateur universel des 193 États membres , tandis que Jaan Tallinn a insisté sur l'urgence, avertissant que les capacités de l'IA progressent sur une base trimestrielle alors que la gouvernance fonctionne sur une échelle annuelle . Gaia Marcus a soutenu que la participation du public doit être traitée comme une source de données probantes, avec l'expérience vécue - en particulier celle des personnes les plus touchées - alimentant directement les structures décisionnelles . Les rapporteurs des groupes thématiques ont renforcé ces thèmes. Le Groupe Un a conclu que l'accès seul ne crée pas la prospérité et que le véritable enjeu est de construire des institutions capables de déployer l'IA avec discernement plutôt que simplement le plus rapidement possible . Le Groupe Trois a souligné que la gouvernance de l'IA doit devenir adaptative, que l'interopérabilité requiert des normes techniques communes, et qu'une supervision humaine significative doit empêcher l'autonomie de devenir un vide de responsabilité . Le Groupe Quatre a soulevé des préoccupations concernant la gestion algorithmique, le financement de la société civile, ainsi que la nécessité d'une sécurité juridique et d'une intégrité de l'information dans les contextes démocratiques . La session s'est conclue par des intervenants offrant des aspirations en un seul mot - équité, progrès collectif, confiance, urgence et compréhension partagée - reflétant un large consensus selon lequel le dialogue a établi une base importante, mais qu'une coopération internationale soutenue, inclusive et réactive est essentielle si l'IA doit servir l'humanité plutôt qu'approfondir les inégalités existantes .

Points clés

Objectif général

La discussion constitue la plénière conclusive de l'inaugural Dialogue mondial des Nations Unies sur la gouvernance de l'IA, tenu à Genève. Son objectif est de synthétiser les messages clés issus de deux jours de dialogue multi-parties prenantes, d'identifier les domaines de convergence entre les groupes thématiques et de tracer une voie vers le second Dialogue mondial prévu à New York en mai 2027. ---

Principaux points de discussion

- Le fossé Nord-Sud en matière d'IA et le besoin urgent d'un accès équitable et de renforcement des capacités. Une préoccupation centrale tout au long du dialogue portait sur le fossé croissant entre le Nord et le Sud global en matière d'adoption de l'IA, d'infrastructure et de capacité de gouvernance. Brad Smith a noté que la diffusion de l'IA s'élève à 27 % dans le Nord global contre seulement 15 % dans le Sud global, avertissant que les tendances actuelles pourraient voir le fossé plus que doubler en l'espace d'un an. La ministre Theofelus de Namibie a appelé à traduire les principes mondiaux de l'IA en lois nationales adaptées aux contextes locaux, à investir dans les infrastructures numériques et de données, et à développer les compétences et la capacité scientifique dans les pays en développement. Les co-présidents des groupes ont réaffirmé que l'accès seul est insuffisant - la capacité, l'adaptation locale et les approches centrées sur les communautés sont essentielles. - Le fossé entre les principes de gouvernance de l'IA et leur mise en œuvre pratique. Plusieurs intervenants ont souligné que si de nombreuses institutions ont articulé des principes d'IA orientés vers la sécurité, l'inclusion, la transparence et la responsabilité, ceux-ci ne se sont pas encore traduits en garde-fous partagés ou en normes contraignantes. La présidente islandaise Tómasdóttir a averti que « des principes sans pratique peuvent inspirer un faux sentiment de sécurité ». Brad Smith a observé qu'au cours des 12 derniers mois, « l'IA a avancé à toute vitesse et la gouvernance de l'IA n'a pas suivi le rythme ». Les co-présidents du Groupe Trois ont appelé à une gouvernance adaptative, à une interopérabilité pratique grâce à des définitions communes et des normes techniques, et à des bacs à sable réglementaires transfrontaliers pour passer du principe à la mise en œuvre. - Le rôle de l'ONU en tant que convocateur universel et légitime pour une gouvernance inclusive de l'IA. Un thème récurrent était le rôle indispensable des Nations Unies pour rassembler l'ensemble des 193 États membres aux côtés de la société civile, du secteur privé et du monde académique. Brad Smith a déclaré que seule l'ONU peut mobiliser le pouvoir nécessaire pour rassembler les gens et obtenir des résultats concrets à l'échelle mondiale. Le Secrétaire général adjoint Guy Ryder a souligné « l'avantage comparatif inégalé » de la légitimité universelle de l'ONU, tout en reconnaissant la nécessité d'une cohérence interne entre les agences onusiennes et d'une ouverture vers d'autres forums internationaux tels que le G20 et le G7. Gaia Marcus de l'Ada Lovelace Institute a soutenu que l'inclusion doit aller au-delà des États pour intégrer des voix communautaires diverses en tant que véritable mécanisme de gouvernance, et non comme simple exercice de consultation. - Les risques liés à l'accélération rapide des capacités de l'IA dépassant la gouvernance, notamment les préoccupations en matière de sécurité et de contrôle. Jaan Tallinn du Future of Life Institute a souligné que les principales entreprises d'IA poursuivent désormais ouvertement l'auto-amélioration récursive - utilisant l'IA pour accélérer le développement de l'IA - risquant une perte de contrôle à mesure que les systèmes deviennent de plus en plus autonomes. Il a noté que des méthodes fiables pour maintenir le contrôle sur des systèmes d'IA hautement autonomes font actuellement défaut, et que les pressions commerciales et géopolitiques concurrentielles accélèrent le développement plus vite que les considérations de sécurité ne le permettent. Il a appelé à une coopération internationale pour assurer un progrès à un rythme sûr, soulignant « l'urgence » comme son mot unique à porter vers New York. - Les droits humains, l'inclusion des communautés marginalisées et la participation multi-parties prenantes comme exigences fondamentales de gouvernance. Le représentant de la société civile Raman Jit Singh Chima a averti que les systèmes d'IA ne sont pas neutres et sont conçus au sein de structures qui reflètent les inégalités existantes, les femmes, les filles, les personnes de genre divers et d'autres groupes marginalisés étant touchés de manière disproportionnée et sous-représentés dans les discussions sur la gouvernance. Il a appelé à une réglementation guidée par les droits humains et à une reconnaissance universelle de la nécessité du consentement aux données. Linda Bonyo a mis en évidence l'exclusion des voix du Sud global en raison de barrières structurelles telles que les refus de visa - notant que l'Afrique représente 7 refus de visa sur 10 - et a appelé à la transparence dans les systèmes d'IA utilisés pour les décisions de visa. Gaia Marcus a soutenu que la participation du public doit être traitée comme une source de données probantes, avec des mécanismes de responsabilité clairs reliant les voix communautaires au pouvoir décisionnel. ---

Ton général

Le ton général de la discussion est prudemment optimiste et constructif, avec une forte note d'urgence sous-jacente. Les remarques d'ouverture de la présidente islandaise et de Brad Smith étaient aspirationnelles et fondées sur des valeurs, présentant la gouvernance de l'IA comme une question de choix humain et de responsabilité collective.

Le ton est devenu plus analytique et parfois préoccupant lors de l'allocution de Raman Jit Singh Chima au nom de la société civile, qui a introduit une prudence historique et mis en évidence les inégalités structurelles et les risques de répéter les échecs passés en matière de gouvernance. Lors de la discussion en panel, le ton a équilibré l'optimisme quant au rôle de convocateur de l'ONU avec une reconnaissance franche de l'ampleur des défis à venir, notamment autour du rythme du changement technologique et du déficit de gouvernance.

Les rapports des groupes thématiques et les interventions du public ont maintenu un registre collaboratif et orienté vers les solutions, les intervenants du Sud global injectant une note d'urgence morale autour de l'équité et de l'inclusion. À la clôture, l'atmosphère était celle d'une détermination collective, résumée dans les cinq mots offerts par les panélistes : optimisme, progrès, équité, urgence et compréhension.

Intervenants

- Ayah Bdeir - PDG de Current AI (un partenariat public-privé à but non lucratif axé sur le développement de l'IA d'intérêt public) - Nafeesa Alshaala - Représentante de Diplotech Solutions - Raman Jit Singh Chima - Directeur mondial des programmes à l'Association for Progressive Communications (APC) - Alexander Pröll - Secrétaire d'État à la Chancellerie fédérale d'Autriche - Linda Bonyo - Fondatrice et PDG de Lawyers Hub Africa ; a pris la parole au nom du Groupe thématique Quatre (représentant également le ministre de l'Espagne) - Grigoriy Borisenko - Vice-ministre du Développement numérique, des Communications et des Médias de masse de la Fédération de Russie - Egriselda López - Co-présidente du Dialogue mondial sur la gouvernance de l'IA - Marie Chatardová - Vice-ministre des Affaires étrangères de la Tchéquie ; ancienne présidente de l'ECOSOC de l'ONU - Pavel Popescu - Vice-président de l'Autorité nationale roumaine pour la gestion et la régulation des communications - Ximena Lincolao - Ministre des Sciences, de la Technologie, de la Connaissance et de l'Innovation du Chili - Anacláudia Rossbach - Directrice exécutive d'ONU-Habitat - Halla Tómasdóttir - Présidente de l'Islande - Samba Diouf - Ministre des Communications et de l'Économie numérique du Sénégal ; Co-président du Groupe thématique Deux - Emma Theofelus - Ministre des Technologies de l'information et des communications de Namibie ; Panéliste - Brad Smith - Vice-président et président de Microsoft - Seydina Moussa Ndiaye - Responsable du groupe de recherche Alive, Université numérique ; Co-facilitateur du réseau mondial soutenu par l'ONU de centres d'échange et de coopération pour le renforcement des capacités en IA - Gaia Marcus - Directrice de l'Ada Lovelace Institute ; Panéliste - Claire Melamed - Vice-présidente, UN Foundation ; Modératrice/Animatrice de la session plénière - Mark-Alexandre Doumba - Ministre de l'Économie numérique et de l'Innovation du Gabon ; Co-président du Groupe thématique Un - Jaan Tallinn - Co-fondateur du Future of Life Institute ; Panéliste - Rashid Khan - Co-fondateur de Yellow AI ; Co-président du Groupe thématique Un - Rein Tammsaar - Co-président du Dialogue mondial sur la gouvernance de l'IA ; Maître de cérémonie - Jovan Kurbalija - Directeur exécutif de DiploFoundation ; Co-président du Groupe thématique Deux - Paula Bogantes Zamora - Ministre (Costa Rica, selon le contexte) ; Co-présidente du Groupe thématique Trois - Guy Ryder - Secrétaire général adjoint chargé des politiques aux Nations Unies ; Panéliste - Rebecca Finlay - Co-présidente du Groupe thématique Trois ; ses domaines d'expertise comprennent la gouvernance de l'IA et la participation multi-parties prenantes

Intervenants
HT
Halla Tómasdóttir
94 wpm · 7 min
BS
Brad Smith
138 wpm · 7 min
RJ
Raman Jit Singh Chima
187 wpm · 7 min
PB
Paula Bogantes Zamora
110 wpm · 2 min
AP
Alexander Pröll
118 wpm · 2 min
ET
Emma Theofelus
153 wpm · 5 min
NA
Nafeesa Alshaala
114 wpm · 3 min
GR
Guy Ryder
154 wpm · 7 min
JT
Jaan Tallinn
150 wpm · 6 min
GM
Gaia Marcus
159 wpm · 5 min
RF
Rebecca Finlay
148 wpm · 2 min
MD
Mark-Alexandre Doumba
131 wpm · 2 min
RK
Rashid Khan
139 wpm · 2 min
SD
Samba Diouf
119 wpm · 2 min
JK
Jovan Kurbalija
118 wpm · 2 min
LB
Linda Bonyo
160 wpm · 4 min
AR
Anacláudia Rossbach
122 wpm · 3 min
PP
Pavel Popescu
139 wpm · 3 min
XL
Ximena Lincolao
135 wpm · 2 min
AB
Ayah Bdeir
143 wpm · 3 min
MC
Marie Chatardová
95 wpm · 3 min
GB
Grigoriy Borisenko
116 wpm · 2 min
SM
Seydina Moussa Ndiaye
113 wpm · 3 min
RT
Rein Tammsaar
116 wpm · 4 min
CM
Claire Melamed
144 wpm · 13 min
EL
Egriselda López
121 wpm · 5 min

Dialogue des Dialogues : Séance plénière de clôture du premier Dialogue mondial de l'ONU sur la gouvernance de l'IA

#

Objectif et structure de la séance

La séance plénière de clôture du premier Dialogue mondial de l'ONU sur la gouvernance de l'IA a été convoquée à Genève dans le but explicite de synthétiser les messages clés issus de deux jours de délibérations multipartites, d'identifier les domaines de convergence entre les groupes thématiques, et de tracer une voie vers le deuxième Dialogue mondial prévu à New York en mai de l'année suivante . Intitulée « Dialogue des Dialogues », la séance était conçue pour rassembler différentes perspectives et réfléchir à ce qui avait émergé des discussions précédentes . Rein Tammsaar, co-président du dialogue, a ouvert les travaux en soulignant que la séance visait à identifier les messages clés, à examiner les domaines de convergence et de complémentarité, et à explorer comment différentes initiatives, expériences et approches pourraient s'informer et se renforcer mutuellement . La séance s'est déroulée à travers une série d'allocutions de haut niveau, une table ronde animée, des comptes rendus des groupes thématiques, et des interventions depuis la salle par des États membres et d'autres parties prenantes.

---

#

Discours d'ouverture : La présidente de l'Islande sur le choix humain et la responsabilité partagée

Claire Melamed, vice-présidente de la Fondation des Nations Unies et modératrice de la séance, a ouvert les travaux en accueillant Son Excellence Halla Tómasdóttir, présidente de l'Islande, pour prononcer le premier discours . Tómasdóttir a donné un ton philosophique et moral à l'ensemble de la séance, soutenant que si l'IA recèle un potentiel transformateur pour la découverte scientifique et les réponses aux défis environnementaux, ses impacts ne seront pas déterminés par la seule capacité technologique, mais par les choix humains faits en matière de pouvoir, d'accès, de responsabilité et du type de société que l'humanité choisit d'être .

Elle a reconnu que les gouvernements, les institutions internationales, les entreprises et la société civile ont formulé d'importants principes en matière d'IA orientés vers la sécurité, l'inclusion, la transparence, la responsabilité, la durabilité et les droits humains, mais a averti qu'un alignement sur des garde-fous communs, des normes partagées et une responsabilité pratique n'avait pas encore été atteint . Son observation la plus percutante était que « des principes sans pratique peuvent inspirer un faux sentiment de sécurité » et qu'il ne suffit pas de paraître responsable tout en évitant les choix difficiles qu'exige la responsabilité . Elle a reformulé le défi de la gouvernance dans ses termes les plus fondamentaux : « La question centrale n'est pas simplement de savoir comment nous gouvernons l'intelligence artificielle, mais si nous sommes capables de nous gouverner nous-mêmes à l'ère de l'intelligence artificielle » .

S'appuyant sur l'expérience de l'Islande qui a su transformer la contrainte en créativité - en exploitant l'énergie volcanique et hydroélectrique pour le bien commun - Tómasdóttir a soutenu que le test pour l'IA n'est pas simplement de savoir si elle peut créer de la valeur, mais si elle peut créer des capacités partagées, des opportunités partagées et une prospérité partagée sans sacrifier la dignité humaine, les relations et le sens de l'existence . Elle a été explicite sur le fait que l'accès sans agentivité n'est pas de l'inclusion, et que les bénéfices de l'IA doivent atteindre les communautés partout dans le monde, les personnes les plus éloignées des centres de pouvoir actuels devant être co-autrices de l'avenir plutôt que bénéficiaires passives . Le succès, a-t-elle soutenu, devrait se mesurer non pas au nombre de projets pilotes annoncés, mais à la question de savoir si les personnes et les communautés disposent d'une plus grande agentivité, d'une plus grande dignité et d'un plus grand espoir grâce à l'IA . Elle a également averti que l'humanité ne doit pas normaliser un avenir dans lequel des décisions irréversibles - y compris le fait d'ôter la vie humaine - seraient déléguées à des machines, et que la responsabilité humaine significative doit demeurer là où les enjeux sont les plus élevés .

---

#

Deuxième discours : Brad Smith sur la fracture numérique et le déficit de gouvernance

Brad Smith, vice-président et président de Microsoft, a ouvert ses remarques en affirmant que « nous serons jugés par les choix que nous faisons » et que le moment d'agir est venu . Il a ancré la discussion dans des données concrètes, notant qu'il s'était écoulé moins de quatre ans depuis la sortie de ChatGPT qui avait inauguré l'ère de l'IA générative, et s'interrogeant sur ce que l'état actuel du monde exige en réponse .

Smith a introduit ce qui est devenu l'un des points de données les plus cités de la séance : la diffusion de l'IA dans le Nord global s'établit à 27 %, tandis qu'elle n'est que de 15 % dans le Sud global, et si les tendances actuelles se maintiennent, l'IA sera utilisée par plus de deux fois plus de personnes par habitant dans le Nord global que dans le Sud global d'ici un an . Son évaluation était sans ambiguïté : « Nous ne sommes pas en bonne posture. Nous ne sommes pas sur la bonne voie, et ce que nous observons au contraire, c'est une nouvelle tendance technologique où la technologie numérique, la nouvelle technologie, va élargir plutôt que réduire l'écart entre le Nord et le Sud » . Il a identifié une série d'étapes fondamentales nécessaires pour y remédier : achever l'électrification du monde, étendre la connectivité internet, permettre aux modèles d'IA de fonctionner aussi bien dans de nombreuses langues plutôt que dans quelques-unes, et faire de la formation aux compétences numériques une réalité quotidienne .

Sur le déficit de gouvernance, Smith a été tout aussi direct, déclarant que « au cours des 12 derniers mois, l'IA a avancé à toute vitesse. Et la gouvernance de l'IA n'a pas suivi le rythme » . Il a reconnu que les risques liés à l'IA - notamment en matière de cybersécurité et, à l'avenir, en biologie - sont croissants, et que la seule façon de rattraper le retard est de travailler ensemble . Il a soutenu que le secteur privé ne peut pas agir seul, pas plus que le secteur à but non lucratif et les ONG, mais qu'en agissant ensemble, de meilleurs résultats peuvent être obtenus . Il a appelé à l'interopérabilité entre les systèmes de gouvernance, reconnaissant que différents pays et cultures exerceront leur souveraineté et feront des choix différents, mais avertissant qu'on ne peut pas se permettre un patchwork impossible à naviguer pour les personnes, les entreprises ou les gouvernements . Il a conclu en affirmant le rôle indispensable des Nations Unies et de leurs agences dans la mobilisation du pouvoir de rassembler les gens, de produire des résultats concrets à travers des programmes réels, et d'offrir au monde ce dont il a besoin et ce qu'il mérite : l'espoir que l'IA peut construire un monde meilleur .

---

#

Troisième discours : La prudence historique de la société civile et les préoccupations structurelles

Raman Jit Singh Chima, directeur mondial des programmes à l'Association for Progressive Communications (APC) - une organisation internationale à but non lucratif fondée en 1990 avec plus de 74 membres organisationnels et 44 associés dans 60 pays - a adopté un registre nettement différent, ouvrant son intervention en reconnaissant un sentiment de malaise . Il a mis en garde contre l'idée qu'il est rare de trouver une véritable page blanche, et que le développement de l'IA et de ses architectures de gouvernance s'inscrit dans des histoires dont il faut tirer des leçons . Il a rappelé aux participants que l'IA n'est pas un phénomène nouveau - elle a traversé des cycles d'essor et de déclin - et que ce qui est particulièrement différent dans le moment actuel, c'est la mesure dans laquelle elle impacte les communautés et les sociétés au-delà des laboratoires et des environnements de test, affectant des individus vulnérables et des institutions de gouvernance déjà fragilisées .

Singh Chima a averti que la communauté internationale s'efforçait encore de bien gérer la gouvernance numérique et se trouvait maintenant sous une pression considérable pour faire un bond en avant vers la gouvernance de l'IA, risquant de répéter les erreurs du passé . Il a soutenu que les processus mondiaux de gouvernance numérique, y compris ceux nés du Sommet mondial sur la société de l'information (SMSI), du Forum sur la gouvernance de l'internet (FGI) des Nations Unies, et de la CNUCED, ont démontré la valeur d'un engagement soutenu entre forums, et que lorsque des succès ont été observés dans la coopération numérique, c'est lorsque les gouvernements, la société civile, la communauté technique, le monde académique, le secteur privé et les organisations internationales ont travaillé ensemble . Sa conclusion était claire : « La gouvernance de l'IA devrait s'appuyer sur cela, et non le remplacer » .

Il a également soulevé des préoccupations structurelles concernant les systèmes d'IA eux-mêmes, soutenant qu'ils ne sont pas neutres et sont conçus et exploités dans des structures qui reflètent les inégalités existantes . Les femmes et les filles, les personnes de genre divers et d'autres groupes marginalisés sont touchés de manière disproportionnée par les systèmes d'IA et sous-représentés dans les conversations sur la gouvernance . Faisant écho à un point soulevé plus tôt dans la discussion sur les droits humains - selon lequel les communautés ne sont pas un problème de politique de dernier kilomètre mais doivent être une priorité et une partie prenante de premier kilomètre - il a appelé à une réglementation guidée par les droits humains et à une reconnaissance universelle de la nécessité du consentement aux données, avertissant que les protections de la vie privée post-Snowden, notamment à l'occasion du treizième anniversaire de ces révélations, risquent d'être compromises par la ruée effrénée vers l'utilisation des outils d'IA et la collecte de données . Il a également mis en évidence une asymétrie profonde : les communautés qui manquent de connectivité voient néanmoins leurs données collectées pour les modèles d'IA et font l'objet d'expérimentations dans le déploiement de l'IA, sans pouvoir exprimer leurs demandes, révéler ce qui se passe ou contribuer à la gouvernance .

---

#

Table ronde : Le rôle de l'ONU, les critères de succès et la voie vers New York

La table ronde animée a réuni le Secrétaire général adjoint Guy Ryder, la ministre Emma Theofelus de Namibie, Jaan Tallinn du Future of Life Institute, et Gaia Marcus de l'Ada Lovelace Institute. Melamed a ouvert en demandant à Ryder comment l'ONU peut aider à relier les nombreux processus et initiatives différents contribuant à la gouvernance de l'IA pour parvenir à la complémentarité, l'inclusivité et l'interopérabilité .

Ryder a commencé par contextualiser les origines du dialogue : c'est seulement en août de l'année précédente que les 193 États membres de l'ONU ont décidé de convoquer ce dialogue, lui conférant une importance et une légitimité particulières . Il a reconnu les contributions extraordinaires des co-présidents, les 1 500 soumissions écrites reçues, et le travail du groupe scientifique indépendant . Sur le rôle de l'ONU, il a identifié son « avantage comparatif inégalé » comme étant la légitimité de son universalité - le seul endroit où 193 pays membres peuvent se réunir dans une conversation pleinement inclusive impliquant non seulement les États mais aussi les entreprises et la société civile . Il a été candide sur le défi interne d'assurer la cohérence et la complémentarité au sein du système des Nations Unies , et a noté la capacité de l'ONU à servir de point de référence pour d'autres processus internationaux tels que le G20 et le G7 . Il a également noté qu'un document de résultats serait préparé par les co-présidents en temps utile à l'issue du dialogue, et a suggéré que le dialogue pourrait devenir un rendez-vous annuel pour faire le point sur les progrès et l'orientation. Il a défini le succès d'ici le prochain Dialogue mondial à New York comme le maintien de la position de l'ONU en tant que convocateur universel unique, tout en bénéficiant de l'engagement des États membres et des autres parties prenantes pour donner une forme opérationnelle aux grands objectifs émergeant du dialogue - notamment faire en sorte que l'IA serve l'humanité . Il a également fait référence aux propositions du Secrétaire général issues de la séance, notamment un engagement pour la sécurité des enfants et une initiative de transparence .

---

#

La perspective de la Namibie : Traduire les principes en pratique

La ministre Theofelus de Namibie a proposé un cadre en trois volets pour la coopération internationale pratique. À court terme, elle a appelé à aider les pays - en particulier ceux du Sud global aux prises avec des défis structurels - à traduire les principes mondiaux en matière d'IA en lois et institutions nationales efficaces adaptées à leur contexte, notant que « le contexte détermine le résultat » . À moyen terme, elle a souligné la nécessité d'investir continuellement dans les infrastructures numériques et de données, reconnaissant que certains pays apprennent encore à leurs citoyens à faire confiance à la technologie et à l'utiliser, et que des financements doivent être rendus disponibles pour soutenir cette progression . À long terme, elle a appelé à des programmes visant à renforcer les compétences et les capacités scientifiques, notamment compte tenu du fait que près de la moitié de la population mondiale proviendra du continent africain d'ici 2050, ce qui rend essentiel le développement de programmes de formation sur mesure pour le Sud global .

Interrogée sur ce que le succès ressemblerait d'ici le prochain Dialogue mondial à New York, Theofelus a identifié trois indicateurs : un cadre de référence commun mais flexible pour la gouvernance de l'IA qui offre aux pays comme la Namibie un modèle pratique mais souverain pour la législation nationale ; des gains de capacité visibles dans le Sud global, notamment davantage de régulateurs formés, davantage de stratégies nationales d'IA déployées, et davantage d'initiatives d'IA interrégionales favorisant une croissance inclusive ; et des partenariats mondiaux entre gouvernements, industrie, monde académique et société civile qui échangent régulièrement des données probantes, mettent à jour les normes et co-développent des outils ouverts . Elle a souligné que les décideurs politiques dépendent de tous ces acteurs pour être correctement conseillés et mettre en place les garde-fous nécessaires pour protéger les citoyens .

---

#

L'argument de la sécurité et de l'urgence : Jaan Tallinn sur l'auto-amélioration récursive

Jaan Tallinn a introduit l'argument le plus techniquement précis et le plus urgent sur le plan existentiel de la séance. Il a reconnu l'immense bien que l'IA peut faire dans des domaines tels que la santé et la science, mais a averti que même les meilleurs experts commencent seulement à comprendre que les progrès vont s'accélérer par rapport à leur niveau actuel . La raison principale, a-t-il expliqué, est que les principales entreprises d'IA poursuivent désormais ouvertement l'auto-amélioration récursive - utilisant leur propre IA pour accélérer le développement de l'IA - et que cela risque de faire perdre le contrôle, car l'autonomie qui rend ces systèmes utiles est aussi ce qui les rend plus difficiles à contrôler . Il a cité l'observation du Secrétaire général selon laquelle les systèmes d'IA ne sont plus des outils attendant des instructions, et a noté que le rapport du groupe scientifique indépendant confirme que des méthodes fiables pour maintenir le contrôle sur des systèmes d'IA hautement autonomes font actuellement défaut .

Tallinn a soutenu que la raison de la course vers des seuils dangereux réside dans la dynamique concurrentielle qui anime le développement de l'IA, et que même les dirigeants des principales entreprises d'IA expriment leur prudence, déclarant qu'ils aimeraient ralentir mais ne le peuvent pas en raison des pressions commerciales et géopolitiques . C'est précisément là que l'ONU et la coopération internationale peuvent intervenir, a-t-il soutenu, pour créer une pression internationale afin de progresser à un rythme sûr, de sorte que les bénéfices de l'IA puissent être captés sans succomber à une perte de contrôle . Il a également proposé une distinction conceptuellement clarificatrice : il n'existe pas de véritable dilemme entre innovation et sécurité, car la plupart des préoccupations en matière de sécurité se concentrent sur la première frontière où les entreprises s'aventurent dans l'inconnu, tandis que la plupart des innovations et des bénéfices proviennent de la deuxième frontière d'adoption et de diffusion où les capacités peuvent être mises à bon usage . Le mot unique qu'il a choisi pour être porté au prochain Dialogue mondial à New York était « urgence », avertissant que les développements de l'IA se produisent désormais sur une base trimestrielle et que la gouvernance ne peut pas continuer à fonctionner sur un cycle annuel .

---

#

L'inclusion, la confiance et la participation comme mécanismes de gouvernance : Gaia Marcus

Gaia Marcus de l'Ada Lovelace Institute a soutenu qu'on ne saurait sous-estimer le caractère historique du fait que tous les États membres de l'ONU ont désormais leur place à la table de l'IA, mais que l'inclusion ne peut pas s'arrêter au niveau des États . Pour que le dialogue soit légitime, a-t-elle soutenu, il doit passer d'une situation où l'on n'entend que les futurs que quelques entreprises, investisseurs ou acteurs puissants souhaitent voir, à une situation où l'on entend les espoirs, les aspirations et les craintes des personnes du monde entier à mesure que l'IA s'entremêle avec la vie quotidienne . Intégrer la confiance et l'inclusion dans le dialogue ne peut pas signifier traiter la participation comme un événement secondaire ou un exercice de consultation .

Marcus a défini la confiance comme étant construite à travers la fiabilité, la réactivité, l'intégrité, l'équité, l'ouverture et surtout la responsabilité - et a soutenu que ce sont des qualités du processus de gouvernance lui-même, et non de simples sentiments à cultiver . Elle a appelé à ce que la participation du public soit traitée comme une source de données probantes, reconnaissant que ceux qui sont le plus affectés par les technologies d'IA - les travailleurs, les communautés vulnérables, les femmes, les filles et les enfants - sont souvent les véritables experts de leurs impacts . Elle a soutenu que l'inclusion des personnes marginalisées ne conduira pas automatiquement au changement à moins que leurs expériences et leurs attentes ne soient reliées à des leviers de pouvoir clairs, et a appelé à des points d'ancrage clairs dans les institutions et le pouvoir . Pour le prochain Dialogue mondial à New York - qu'elle a mentionné comme devant se tenir en mai 2027 - elle a appelé à une architecture de participation intégrée et institutionnalisée autour du dialogue qui apporte de la substance à un plancher de gouvernance mondiale reconnu, respecté et défendu - un plancher qui reconnaît l'expérience vécue comme une donnée probante, offre aux communautés des voies claires vers la prise de décision, et crée des obligations transparentes dont le dialogue est responsable . Son argument final était que « c'est ainsi que l'inclusion passe du statut de thème à celui de mécanisme de gouvernance et que la confiance du public devient quelque chose qui se mérite, et non quelque chose qui s'affirme simplement » .

---

#

Comptes rendus des groupes thématiques : Principales conclusions des groupes de travail

À la suite de la table ronde, les co-présidents des quatre groupes thématiques ont fourni des comptes rendus condensés des discussions des deux jours précédents. La co-présidente Egriselda López a géré les travaux au cours de cette section, présentant les intervenants et fournissant des orientations pratiques tout au long.

Le groupe un, co-présidé par le ministre Mark-Alexandre Doumba du Gabon et Rashid Khan de Yellow AI, a examiné les dimensions sociales, économiques, culturelles, linguistiques et techniques de l'IA . Doumba a soutenu que l'instinct de rivaliser sur la vitesse passe à côté de l'essentiel : « Le véritable enjeu est de savoir qui construit les institutions et le discernement nécessaires pour déployer l'IA de manière efficace, plutôt que de savoir qui la déploie le plus rapidement » . Il a noté que la majeure partie des opportunités offertes par l'IA proviendra de l'adaptation des grands modèles d'IA aux contextes locaux - permettant à un agriculteur africain ou à un opérateur de services publics des îles du Pacifique d'utiliser réellement la technologie - et que l'IA est un changement de paradigme exigeant la refonte des systèmes avec de nouvelles incitations, de nouvelles mesures du succès et de nouvelles hypothèses sur qui participe . Khan a renforcé l'idée que l'accès seul ne crée pas la prospérité ni les emplois - c'est la capacité qui le fait - et que l'avertissement le plus percutant de la salle était que « le plus grand risque est que la majeure partie du monde reste spectatrice pendant que d'autres captent la valeur » . Il a soutenu que la souveraineté et l'ouverture sont des partenaires plutôt que des opposés, et que les résultats - un enfant en meilleure santé, une meilleure récolte, une réponse gouvernementale plus rapide - constituent le mandat, à mesurer honnêtement, y compris l'empreinte environnementale de l'IA elle-même .

Le groupe deux, co-présidé par Jovan Kurbalija de la Fondation Diplo et le ministre Samba Diouf du Sénégal, a débuté avec Kurbalija notant que 43 intervenants avaient prononcé 12 084 mots au cours de la discussion, avec trois mots - capacité, aptitude et contexte - résonnant le plus fortement . Il a dressé une taxonomie des lacunes en matière d'inclusion numérique : accès, financement, participation, compétences, connaissances, langue et confiance. Il a noté que le mot « connaissance » avait été mentionné 35 fois dans la discussion contre 63 fois pour « données », et a suggéré - comme Kurbalija lui-même l'a formulé - que la connaissance pourrait être plus importante que les données dans ce contexte, avec la création locale, la co-création et l'expertise locale comme thèmes clés . Diouf a souligné que la fracture s'applique non seulement à l'IA mais aussi aux infrastructures, à la connectivité et aux appareils, et que pour les pays d'Afrique de l'Ouest, résoudre ces défis fondamentaux est un prérequis pour bénéficier de l'IA . Sa conclusion était directe : « Nous avançons ensemble. Nous partageons les connaissances ensemble. Nous partageons la plateforme ensemble. Nous partageons tout ensemble » .

Le groupe trois, co-présidé par la ministre Paula Bogantes Zamora du Costa Rica et Rebecca Finlay, a confirmé qu'une gouvernance efficace de l'IA n'exige pas que chaque pays suive le même modèle, mais qu'elle nécessite de relier différentes approches tout en respectant le contexte national, l'autonomie réglementaire et les différents niveaux de capacité institutionnelle . Les messages clés comprenaient : la gouvernance de l'IA doit devenir adaptative, car des règles statiques ne seront pas suffisantes pour des systèmes de plus en plus autonomes ; l'interopérabilité doit devenir pratique grâce à des définitions communes et des normes techniques partagées ; la coopération doit passer du principe à la mise en œuvre à travers des bacs à sable réglementaires transfrontaliers, des méthodes d'évaluation partagées et des mécanismes de signalement des incidents ; la confiance exige la transparence, une évaluation indépendante et la responsabilité tout au long du cycle de vie de l'IA ; les cadres d'évaluation mondiaux doivent refléter la diversité linguistique, culturelle et démographique ; et à mesure que les systèmes d'IA agentiques acquièrent une plus grande autonomie, la responsabilité doit rester clairement attribuée aux développeurs, aux déployeurs et aux opérateurs, avec une supervision humaine significative garantissant que l'autonomie ne devient jamais un vide de responsabilité . Finlay a ajouté que le rapport du groupe scientifique indépendant donne le plancher des préoccupations, et non le plafond, et a appelé à renforcer la base de données probantes, à développer un niveau de référence minimum commun pour l'interprétabilité, à aligner les sept groupes de travail du groupe scientifique pour mieux informer les groupes thématiques du prochain dialogue, à créer une continuité entre les dialogues, et à affirmer que la participation multipartite est une nécessité, et non un luxe .

Le groupe quatre, représenté par Linda Bonyo, fondatrice et directrice générale de Lawyers of Africa, qui s'est exprimée au nom d'elle-même et du ministre espagnol co-président absent, a couvert un large éventail de préoccupations relatives aux droits humains et à la gouvernance. Le groupe a convenu que les enfants doivent être protégés mais ne constituent pas un groupe homogène, les enfants immigrés et handicapés nécessitant une attention spécifique . Il a soulevé le défi du financement des droits humains à l'ère numérique, notant que les organisations de la société civile perdent des financements et qu'il y avait moins de représentants de la société civile à ce dialogue que précédemment . Il a mis en évidence l'absence des voix des travailleurs de l'IA dans les conversations sur la gouvernance . De manière particulièrement frappante, Bonyo a présenté des données montrant que 51 % des conversations sur la gouvernance de l'IA au cours des six derniers mois ont eu lieu à Genève, et que l'Afrique représente 7 refus de visa sur 10, excluant structurellement les voix du Sud global des forums qui façonnent leur avenir . Elle a appelé à la transparence et à la responsabilité dans les systèmes d'IA utilisés pour déterminer les refus de visa, et a proposé - à titre de point de vue personnel plutôt que de consensus du groupe - le concept d'un passeport IA pour faciliter la mobilité aux fins de participation à la gouvernance de l'IA . Le groupe a également discuté de la diversité linguistique, de la philosophie Ubuntu de la communauté, de la situation des communautés de la diaspora et des réfugiés, des droits des artistes et des travailleurs créatifs, de la nécessité de la sécurité juridique, de l'intégrité de l'information et de la relation entre l'IA et la démocratie . Bonyo a noté que les États signent des traités internationaux mais ne les intègrent fréquemment pas dans leur droit interne, et a appelé les États à soutenir et à ratifier les conventions de l'OIT sur le travail décent pour les travailleurs de l'IA .

---

#

Interventions depuis la salle : États membres et parties prenantes

Une série d'interventions depuis la salle a renforcé et élargi les thèmes de la séance. La vice-ministre des Affaires étrangères de la Tchéquie, Marie Chatardová, a salué le rapport du groupe scientifique indépendant et a appelé à une approche centrée sur l'humain au cœur du développement de l'IA, à l'interopérabilité plutôt qu'à la fragmentation, et à l'ONU comme plateforme universelle rassemblant toutes les parties prenantes . Elle a souligné que les entreprises qui développent et déploient l'IA ont une responsabilité clé pour garantir une utilisation sûre et responsable .

Le secrétaire d'État autrichien Alexander Pröll a soutenu qu'après avoir écouté les gouvernements, les scientifiques, l'industrie et la société civile de toutes les régions, sa conviction que le monde a besoin d'un cadre contraignant pour la gouvernance de l'IA n'avait fait que se renforcer . Il a cité l'expérience propre de l'Autriche - notamment une plateforme GovGPT au service de 180 000 fonctionnaires, une initiative de compétences numériques, un âge minimum de 14 ans pour les réseaux sociaux, et une Charte européenne sur la souveraineté numérique - comme preuve qu'une utilisation responsable de l'IA est possible sans ralentir l'innovation . Son appel était sans ambiguïté : « Nous avons besoin d'un cadre mondial pour la gouvernance de l'IA... fondé sur des principes communs, soutenu par de véritables garde-fous, incluant chaque pays dans cette salle » .

La directrice exécutive d'ONU-Habitat, Anacláudia Rossbach, a souligné que les villes et les gouvernements locaux sont là où l'IA est déjà appliquée chaque jour, prenant des décisions qui affectent directement la vie des gens et générant des données probantes cruciales pour l'élaboration des politiques en matière d'IA, mais qu'ils restent largement absents des délibérations mondiales sur la gouvernance de l'IA . Elle a ancré ses remarques dans des données concrètes, notant que 3 milliards de personnes n'ont pas accès à un logement adéquat et abordable et qu'1 milliard de personnes vivent dans des établissements informels, soulignant l'urgence de veiller à ce que l'IA serve équitablement les communautés urbaines. Elle a annoncé une initiative conjointe avec l'UIT pour participer au Smart City Expo World Congress 2026 à Barcelone, incluant le lancement d'un groupe de travail sur les principes directeurs et les cadres pour des villes intelligentes centrées sur les personnes à l'ère de l'IA .

La ministre chilienne des Sciences, de la Technologie, de la Connaissance et de l'Innovation, Ximena Lincolao, a identifié cinq conclusions du dialogue : l'urgence, la préservation de l'agentivité humaine, l'IA comme infrastructure critique nécessitant une gouvernance systémique, la nécessité d'investir dans les talents, et l'importance du calcul informatique comme moteur de la nouvelle industrie, accompagnée d'une responsabilité en matière d'énergie renouvelable .

Ayah Bdeir, directrice générale de Current AI, a soutenu qu'il existe déjà un consensus entre les gouvernements, les consommateurs et les entreprises sur le fait que l'IA ne peut pas être concentrée entre les mains de quelques-uns et doit servir toute l'humanité, mais que la question est de savoir comment construire l'IA dans l'intérêt public . Elle s'est appuyée sur des modèles historiques de collaboration technique internationale - le World Wide Web et Linux - et a décrit comment son organisation avait recensé plus de 24 000 outils et composants du monde entier pour identifier les lacunes dans l'écosystème d'IA open source. Elle a annoncé le lancement prochain d'un chatbot ouvert et souverain construit entièrement à partir de composants open source provenant de plusieurs pays - avec le modèle de fondation de Suisse, le calcul informatique de Finlande, les ensembles de données d'Inde, les garde-fous de sécurité d'une organisation à but non lucratif américaine, et une plateforme de données préservant la vie privée d'une autre organisation à but non lucratif - comme démonstration qu'une telle collaboration est possible . Elle a soutenu que la collaboration technique internationale à la vitesse de l'IA nécessite des bases de code ouvertes plutôt que la collaboration traditionnelle par le biais de mémorandums d'accord et d'accords juridiques .

Le vice-ministre russe du Développement numérique, Grigoriy Borisenko, a déclaré que la Russie dispose d'une chaîne technique complète pour créer des modèles d'IA avancés sans composants étrangers, notamment les grands modèles de langage GigaChat et Yandex GPT, et a exprimé sa disponibilité à partager son expérience avec d'autres pays, dans le respect de leurs codes culturels, et à travailler sur des normes communes qui réduisent le risque que l'IA devienne le privilège de quelques-uns .

Seydina Moussa Ndiaye de l'Université numérique, co-facilitant le réseau mondial de centres de renforcement des capacités en IA soutenu par l'ONU, a décrit un réseau qui ne crée pas une nouvelle institution mais relie des centres d'excellence existants dans un cadre collaboratif couvrant le développement des compétences en IA, l'accès au calcul informatique et aux données, le soutien aux stratégies nationales d'IA, et le partage de solutions pratiques d'IA . Il a invité les États membres à désigner des centres d'excellence pour rejoindre le réseau et a encouragé les partenaires de développement et l'industrie à y contribuer .

Nafeesa Alshaala de Diplotech Solutions a appelé à trois priorités : institutionnaliser l'harmonisation réglementaire avec un langage diplomatique commun pour la gouvernance de l'IA ; démocratiser l'expertise technique à travers un fonds mondial de capacités en IA et des pôles trisectoriels décentralisés ; et traiter les angles morts, notamment les coûts environnementaux, les menaces du contenu synthétique pour l'intégrité de l'information, et la protection des centres de données mondiaux en tant qu'infrastructure critique non militaire.

Le représentant de la Roumanie, Pavel Popescu, invoquant le principe attribué à Nicolae Titulescu selon lequel la paix ne doit pas seulement être affirmée mais organisée - et l'appliquant à la gouvernance de l'IA - a plaidé pour une architecture de transition centrée sur l'humain qui garantit que le passage à une économie façonnée par l'IA ne laisse pas les plus vulnérables derrière, appelant à des programmes de reconversion et à des voies vers de nouveaux emplois . Il a également appelé à la retenue là où les enjeux sont les plus graves, se félicitant du dialogue bilatéral entre les principales puissances de l'IA, et a souligné la protection des enfants comme une question de principe .

---

#

Synthèse de clôture et mots d'aspiration

Dans sa synthèse de clôture, Melamed a rassemblé les thèmes récurrents : la nécessité d'un cadre flexible pouvant s'adapter aux changements rapides de la technologie, aux besoins des différents pays et communautés, et à l'évolution du domaine de la gouvernance de l'IA lui-même ; la nécessité d'une approche inclusive englobant tous les États membres de l'ONU, les communautés et les entreprises ; la nécessité que la gouvernance ait une application pratique reliant les grandes conversations à la mise en œuvre nationale et aux changements réels dans la vie des gens ; et le rôle critique de l'ONU comme lieu où tous ces différents attributs du dialogue peuvent se réunir .

Chaque panéliste a proposé un mot ou une expression à porter au prochain Dialogue mondial à New York. La ministre Theofelus a choisi « équité » et « progrès collectif » . Le Secrétaire général adjoint Ryder a choisi « confiance » - non pas une confiance irrationnelle, mais une confiance fondée sur le fait que l'IA commence à fonctionner pour les gens et que l'équilibre de la peur se déplace vers l'optimisme parce que la gouvernance démontre sa capacité à exercer un contrôle . Tallinn a choisi « urgence », réitérant que les développements trimestriels de l'IA ne peuvent pas être gouvernés sur un cycle annuel . Marcus a choisi « compréhension partagée » - des données probantes qui vont au-delà du scientifique pour examiner l'impact sur les personnes et les sociétés, de la responsabilité, et de l'ONU comme lieu pour faire avancer les normes minimales de gouvernance . Melamed a synthétisé les contributions de ces panélistes en cinq mots qui lui sont propres - optimisme, progrès, équité, urgence et compréhension - les décrivant comme un excellent programme et une aspiration à porter dans le prochain dialogue mondial .

La co-présidente Egriselda López a conclu en affirmant que la table ronde avait clairement montré que la conversation sur la gouvernance de l'IA ne peut pas s'arrêter avec ce dialogue, et que les expériences et les idées partagées au cours des deux derniers jours soulignent qu'une gouvernance efficace de l'IA nécessitera une collaboration continue entre les secteurs, les régions et les initiatives . Rein Tammsaar a remercié tous les intervenants et a invité les participants à soumettre des déclarations écrites à l'adresse électronique officielle - AI dialogue at UN.org - pour s'assurer que toutes les contributions sont consignées dans le compte rendu , avant d'inviter les participants à se lever pour une photo de groupe marquant la conclusion du premier dialogue .

---

#

Évaluation globale

La séance dans son ensemble a révélé un degré élevé de consensus de surface sur les principes fondamentaux - l'inadéquation de la gouvernance actuelle de l'IA par rapport au rythme du développement technologique, le risque que l'IA élargisse plutôt que réduise les inégalités mondiales, le rôle indispensable de l'ONU en tant que convocateur universel, la nécessité de la participation multipartite, et la centralité des droits humains comme fondement de la gouvernance de l'IA. Sous ce consensus, cependant, se trouvaient des tensions significatives : entre des cadres de gouvernance contraignants et flexibles ; entre la priorité accordée aux infrastructures fondamentales et aux capacités spécifiques à l'IA ; entre la construction sur les processus de gouvernance numérique existants et la création de nouvelles architectures spécifiques à l'IA ; et entre l'urgence d'accélérer la gouvernance et l'importance de construire des institutions avec soin et discernement . Le défi structurel le plus pointu a été introduit par l'observation de Bonyo selon laquelle les taux de refus de visa et les crises de financement de la société civile compromettent activement l'inclusivité multipartite à laquelle le dialogue aspire - un défi que d'autres intervenants, malgré leur engagement en faveur de l'inclusion, n'ont pas directement abordé. Les cinq mots de clôture proposés par Melamed comme synthèse des contributions des panélistes - optimisme, progrès, équité, urgence et compréhension, tirés respectivement de la « confiance » de Ryder, du « progrès collectif » de Theofelus, de l'« équité » de Theofelus, de l'« urgence » de Tallinn et de la « compréhension partagée » de Marcus - reflètent à la fois l'ambition de ce qui a été accompli et l'ampleur de ce qui reste à faire avant que le deuxième dialogue mondial se réunisse à New York .

Rein Tammsaar
has been designed for bringing together different perspectives and reflecting on what we've heard during this dialogue, first and second day. It will provide an opportunity to identify key messages emerging across the dialogues, consider areas of convergence and complementarity, and explore how different initiatives, experiences, and approaches can inform and reinforce one another. It is our hope that this conversation will help draw together the many perspectives shared throughout the dialogue and contribute to a stronger understanding of where further cooperation, coordination, and collective action may be most valuable. To begin with this session, it is my great pleasure to invite Claire Melamed, Vice President, UN Foundation, to deliver the opening high -level remarks. remarks. Thank you.
Claire Melamed
thank you very much indeed ambassadors Doreen thank you all thank you everybody for a hugely stimulating and engaging few days I say it feels like very much longer already and let me assure you that that is an entirely good way because it has been so packed full of content of stimulating ideas and of commitments to as the intervenants just now said to make sure that this dialogue carries forward with concrete conversations with specific ideas and takes us sets us well onto the road for the second global dialogue Brace Yourselves to be held in New York next May I'm absolutely delighted to begin this next segment the dialogue of dialogues which will start to chart the way forward and think about how the UN can play a key role in ensuring that AI governance is built upon the foundations of human rights and of the values espoused by the UN I'm delighted to begin with high -level opening remarks and as befits a multi -stakeholder dialogue we are delighted to have remarks from different stakeholders here and I'm beginning and let me begin by welcoming to the stage Her Excellency Halla Tómasdóttir, the President of Iceland.
Halla Tómasdóttir
Thank you. Thank you very much. Thank you. scientific discovery and our response to environmental challenges. But its impacts will not be determined by technological capacity alone. It will be determined by the human choices we make about power, access, accountability and ultimately the kind of society and humanity we choose to be. Around the world, governments, international institutions, companies and civil society have put forward important principles for artificial intelligence. Many of them point in the same direction towards safety, inclusion, transparency, accountability, sustainability and human rights. But we have not yet aligned on shared guardrails. common standards or the practical accountability that this technology demands. And principles without practice can inspire false comfort. It is not enough to sound responsible while avoiding the difficult choices that responsibility requires. The central question is not simply how we govern artificial intelligence, it is whether we can govern ourselves in the age of artificial intelligence. I come from Iceland, a small island nation that has long turned constraint into creativity. For centuries, we have carried our history, imagination, and values through stories, through the sagas, poetry, music, and our very own language. They are our roots. They are our roots. They gave us a sense of identity, belonging, and continuity across generations. We have also learned to draw energy from the magma beneath our feet and from the power of our waterfalls, transforming natural forces into warmth, light, and opportunity. We tend to believe that even the most powerful forces should be put to work for the common good. That, it seems to me, is the test for artificial intelligence as well. Not simply whether it can create value, but whether it can create shared capability, shared opportunity, and shared prosperity without sacrificing our human dignity, relationships, and sense of purpose. Access without agency. Agency is not inclusion. The benefits of AI must reach communities everywhere, not only those already closest to capital, computing power, and decision -making. Those furthest from today's centers of power must be co -authors of this future, with a meaningful role in building the infrastructure, shaping the rules, and sharing fairly in the value created. We should measure success not by the number of pilots we announce, but whether people and communities have greater agency, dignity, and hope because of AI. AI is not only a technological question. It is a social, economic, and democratic question, and increasingly a question of peace and security. What happens in a world where young people conclude that the world is a place of peace and security? That the safest response to the digital world is to trust nothing and no one. And we must not normalize a future in which human beings delegate the irreversible decision to take a human life to a machine. Meaningful human responsibility must remain where the stakes are highest. This autumn, Iceland will invite people across the country into a national dialogue in which young people will play a leading role. We need a future -fit leadership that neither confuses speed with progress nor technical capability with wisdom. We need leadership that brings governments, business, educators, researchers, civil society, parents, and young people into genuine partnership. We need leadership that asks not only, what can this technology do? What can this technology do? but also how will it affect the lives of our children. I come from a country where we have learned to live alongside volcanoes, earthquakes, and pretty unpredictable weather. Perhaps that is why we tend to believe that powerful forces should be studied, understood, respected, and put to work for the common good. I believe this is the task before all of us, to ensure that artificial intelligence expands the horizons of young people, of all of us, without diminishing our humanity. To ensure that it helps us become more capable, more creative, more connected, and more hopeful for the future. And to ensure that the intelligence we create remains. In the service of the very humanity
Claire Melamed
Thank you so much, Your Excellency, and thank you for reminding us, first, that we have key choices to make here, policy choices, individual choices. We're not powerless in the face of technology, but we can choose the path. And secondly, that the path must be guided by the impact on people. That must be the criteria that guides the choices that we make and how we're judged for them. Thank you very much for those powerful remarks. And let me now welcome for our second keynote of this segment, Brad Smith, the Vice Chair and President of Microsoft. You're very welcome.
Brad Smith
Well, thank you. It's a real pleasure for me to be here this afternoon. And let me begin where you just left off. We will be measured by the choices we make. And the time to start making them is now because we come together at what I believe is a critical moment for the future of artificial intelligence. Let's take stock. It was less than four years ago at the end of November in 2022 when chat GPT was released and in effect a new era with generative AI began. Where are we today? What are the issues we need to address? Well, I think we should start where many of the intervenants this afternoon have started as well. Let's look at the state of the world and the divides that characterize the planet on which we live. Let's look at the greatest divide of all, the economic divide between the global north and south. At Microsoft, one of the things we do is measure every quarter. The state of AI diffusion, its usage or adoption on a per capita basis among the working age population in countries around the world. And what we see today is that AI diffusion in the global north stands at 27%, while in the global south it is only 15%. And what we know is that if current trends persist by this time next year, AI will be used by more than twice as many people on a per capita basis in the global north as in the global south. What does that tell us? It tells us that we are not in good shape. We are not on the right course, and instead what we are seeing is yet again a new technology trend, where digital technology, new technology, is going to widen rather than narrow the gap between north and south. We have important decisions to make, and you've heard intervenants capture its critical elements. One cannot speak. Let's simply deploy AI in the cloud as if the cloud is the space that it exists. We need a foundation. We need infrastructure. We're trying to invest in that as a company. But we still need to complete the electrification of the world. We have even more distance to travel to complete the Internet connectivity of the world. We need to expand the ability of AI models to work equally well in many languages rather than a few. And we need to make digital skilling an everyday reality rather than a dream that is being realized by far too few. These are the kinds of steps we need to take. And we need to start by recognizing that we're going to need to bend the arc of history if we're going to make progress. Similarly, I think we need to take stock and acknowledge that around the world, not everywhere, but in many countries, many people have many more questions and even concerns today than they did just a year ago. They ask, will AI bring me and my family, my neighbors and my community more economic opportunity or less? Will it create better jobs for my children or will it instead create technology that creates the jobs we already have? This is not a decision that machines will make. It is a decision that people and governments and countries will make. And we need to make them with the right kind of forethought and the right kind of values if we are going to ensure that AI serves its purpose and creates a better world. And then finally, I think there is in some ways the most fundamental question of all. Where are we when we look at the state of AI governance around the world? I think what we can say is this. Over the last 12 months, AI has raced forward. And AI governance has not. And I think that's what we need to do. And I think that's what we need to do. kept pace. We've reached a point of recognizing quite rightly that if not governed well, the risks of AI can grow, as they are this year with respect to cybersecurity, as they will in future years when it comes to biology with technology that can do both enormous good as a tool but enormous damage as a weapon. We have to catch up. And as you've heard other intervenants talk today, there is only one way to catch up when it comes to AI governance. We must work together. The private sector cannot act alone. The nonprofit and NGO sector cannot act alone. But if we act together, we can do better. And as you've heard, we need interoperability. We need to recognize that different countries and cultures will quite rightly exercise their sovereignty and make different choices. That is something that is always part of international relations. But as you've heard, we cannot afford a patchwork that is impossible for for people or companies or governments to navigate. We need to build the interoperability among governance systems if we're going to turn the current stalemate into progress. That requires the United Nations. It requires other institutions as well, and there will be many. But I continue to believe that the United Nations, the ITU, the other agencies that are here this week and every day of every week, have an indispensable role to play. Only by harnessing the power to bring people together, to solve problems through stronger public policy, to deliver real results through real programs to countries around the world can we offer what the world needs and what the world deserves. Hope. Hope that we can turn AI into... into what it should be. A new technology that can build a better world. certainly on behalf of Microsoft on behalf of I think so many people who come from so many different sections of society in so many parts of the world I am happy to be here I am happy that all of you are here and I am happy that we now have the opportunity to go forward together thank you very much
Claire Melamed
Thank you so much to Brad Smith for reminding us that while there are many divides here, digital divides, governance divides, it's absolutely critical that we bridge those, that we bridge those through institutions and that we bridge those through what is rapidly becoming one of my favorite words of the week, interoperability of approaches. For our final plenary segment, for our final keynote speaker of this segment, I'm very delighted to now welcome Mr. Raman Jit Singh Chima , who is the Global Program Director at the Association for Progressive Communications. You're welcome.
Raman Jit Singh Chima
Thank you so much, Claire. I also wanted to thank the joint organizing team of the AI Dialogue and the agencies. We've been working so hard for these conversations here in Geneva and online this week. Your Excellencies, Colleagues, I speak to you on behalf of the Association for Progressive Communications. For those of you who may not know us, we are an international non -profit organization born in 1990 with over 74 organizational members and 44 associates across 60 countries. From the beginning of our existence, APC's network has recognized that it is crucial for the international system to engage with civil society, public interest technologists, and grassroots communities on their design and conversation around technological systems, their development, and governance. And I stand before you today sharing the fact that I feel a bit uneasy. It would be tempting for us here to focus on conversations about the future, how should we approach AI and its governance. But it's rare to find a truly blank slate. The development around AI and its governance, and the architectures in particular around institutional development, come from histories that we must learn from. We must not forget that the development of AI is not inevitable and indeed not as a new phenomenon. As many of you might be aware, AI and the range of technologies that constitute artificial intelligence and the term we use today come from development and development cycles of AI booms or what sometimes people call AI springs and AI winters. What might be of course particularly different and unique about today's boom or today's AI spring is the extent to which it is impacting our communities and societies. That is not merely in labs, in testing environments and more. And in that context it is impacting vulnerable communities and individuals and it is already trying to seek and in fact successfully achieve the attention of already strained and contested policy and governance institutions. The internet has become a mainstream force. That does not mean it is universally accessible or it is equitable in its development or in its application. Transcription by ESO. Translation by — This is the problem that our communities were engaging, and this is a problem that we need to be honest about when it comes to this space. We were still working hard on getting digital governance right, and we are now under tremendous pressure to leapfrog into a conversation around AI governance and to create effective architectures for it. Global digital governance processes, including those born out of the World Summit for the Information Society, and that includes the UNIGF as well as the processes in the UNCSTD, has demonstrated the value of sustained cross -forum engagement. When we have seen success in digital cooperation at the global level, that has been when governments, civil society, the technical community, academia, the private sector, and international organizations have worked together. Many of those who work for the ITU and other UN agencies here in Geneva, as well as those working in New York and other UN hubs, have recognized and learned this through discussion, dialogue, and sometimes debate and disagreement with civil society and other states. AI governance should build on that, not replace that. The United Nations system and the different agencies involved in these conversations must remember this as they plan for the next Dialogue New York, as they design commissions and other fora, particularly given the challenges the UN faces at a broader level over the next few months and the reality of the UN's leadership transition and changes. We need to also recognize that the current and upcoming developments around AI raise concerns that intersect with existing concerns around connectivity, human rights, standards, cybersecurity, and digital inclusion. AI systems are not neutral. They are designed and operated in structures that come from existing inequalities. Women and girls, gender diverse people, and other marginalized groups are distinctly and disproportionately harmed by the development and use of AI systems and are underrepresented in the conversations around their evolution, design, and governance. The global dialogue in AI governance is an opportunity for governments and decision makers to embed commitments in these affirmations on gender equality from existing digital policy and cooperation. AI systems are not neutral. We need to recognize that the UN is a global organization and we need to recognize that the UN is a global organization and we need to recognize that the UN is a global organization and we need to recognize that the UN is a global organization and we need to recognize that the UN is a global organization from this start. You can learn from those mistakes. You can learn from those wins. While the Internet impacts and connects to many other parts of our societies and economies, we need to recognize that there is still a reality. Your gender, your community identity, your geography and more impacts what connectivity you get and whether you get connectivity at all. Absence of connectivity impacts your agency and autonomy. You will still have your data scooped up for AI models and often experimented on in AI deployment, but you cannot voice your demands, you cannot reveal what is happening, and you cannot provide your thoughts on governance. Connectivity efforts are also being impacted by how AI connectivity conversations, sorry, AI infrastructure conversations are grabbing the attention of policymakers and are focusing on particular approaches regarding deployment of compute infrastructure without looking at what communities want. We need a community -centric approach to Internet infrastructure and one that recognizes the importance of connectivity. It recognizes the impact this has on climate change, a topic that the AI dialogue needs to engage with further and needs to connect further with environmental defenders and more. My colleague Rebecca mentioned this earlier today in the conversation around human rights saying that we need to recognize that the conversations here about communities that they are not a last mile policy problem but communities need to be a first mile priority and stakeholder I want to end with a reminder on history History has taught our communities that do not be afraid of red lines In particular, we need to be clear about what we deem and enforce as unacceptable when it comes to people's rights including their rights to their data and privacy as well as the sustainability of the planet June, last month, was the 13th year anniversary of the Snowden revelations It prompted the United Nations and the UN member states to redouble the protection of the right to privacy in a digital age to drive national reforms around data access and more But all of those stand at risk and surveillance controls, for example, in particular are completely at risk of being jeopardized due to the current relentless rush for AI tool usage for partnerships between the public sector and AI firms and other efforts to collect and use data in AI systems. We therefore fully agree with the vision that was hinted at by the Secretary General and by the High Commissioner for Human Rights today. We need regulation guided by human rights, a universal recognition around leaving no space for data collection and AI development without consent. And in closing, I want to thank you all for this opportunity and in particular know that civil society looks forward to working with everyone in the AI dialogue through collaboration, through challenge and through imagination. Thank you. Thank you.
Claire Melamed
Thank you so much, Raman and thank you for reminding us that it's a very long road that brought us here and there are many interlocking processes that we have to think about and much history here also of discussion, of debate among many different stakeholders of these issues So, speaking of many different stakeholders' discussion and debate we are now going to turn to the panel section of this session of this plenary And let me just welcome our panelists to the stage. If you could all come up on the stage as I call your name and then we will begin. So let me first welcome Her Excellency Emma Theofelus , the Minister of Information and Communications Technology of Namibia. Very warm welcome to you. Let me welcome Guy Ryder, Under -Secretary General for Policy at the United Nations. Jaan Tallinn, the co -founder of the Future of Life Institute. And Gaia Marcus, the director of the Ada Lovelace Institute. A very warm welcome. Everyone has found their seats. Everyone is comfortable. Great. Let us begin. So this is the concluding plenary panel of the inaugural UN Global Dialogue on AI Governance. and as seems appropriate we are going to start facing forward into the future and think about the next UN global dialogue to be held in May of next year in New York and to think about some of the things that we have learned over the last two days, some of the lessons that we will take and some of the work that needs to be done in order to get us to that next dialogue. So let me start with you Under -Secretary General. We've heard over the last few days of many different processes and initiatives, all of the huge amounts of work that are going on in different combinations and constellations of stakeholders which are contributing incrementally to AI governance. How can the UN help to connect these efforts to achieve the complementarity, the inclusivity, and the interoperability which I think many people in this room would like to see?
Guy Ryder
Thank you very much for that question To answer the question and to look forward as you've just invited us to do I think it's worth just seeing for a minute how we got here today It was only in August of last year that our member states decided to convene this dialogue This dialogue is happening because 193 countries of the world decided that it would be a good idea to come together and have this dialogue and it's important to remember the mandate not only what its content is but who it comes from and it gives a particular importance and a particular legitimacy to what we're all doing here together today and I have to say that I think we deserve to recognise the extraordinary contributions that people have made to making this dialogue work rather successfully We owe big thanks to our co-chairmen and co-chairs who undertook really very intense consultations to get this dialogue off the ground. We got 1,500 written submissions. And, of course, we've worked very carefully and very closely with the independent scientific panel that we heard report to us yesterday. And we'll be getting an outcome document, an outcome from this dialogue to be prepared by the co-chairs in good time. So this dialogue has set us on, I think, an important road. Now, what can the UN do to bring things together? It is obviously true that there is an enormous amount of attention and activity taking place internationally around AI issues. Some of them here this week in Geneva, not just this dialogue, but the AI for Good events, the WSIS events as well. And the point about this is the UN enjoys one unparalleled comparative advantage in the task before us, and that is the legitimacy of its universality. There is nowhere else where 193 member countries can come together in this all -inclusive conversation. And when we say inclusive, of course, it's all our member states, but it's also business, it's also civil society, as we have just heard. We have a challenge, let me be perfectly honest with you, moderator, inside the UN system. There's a lot going on inside the system. We need to make sure that what happens inside the UN system is coherent. It's complementary. We are working hand -in -hand within the system. And I think what we've seen from the dialogue and the panel this week is good evidence that we are working well on that. but I also think that we are able through the UN to try to reach out to to connect with those other international processes that will take place not universal ones but different summits that will take place no doubt the G20, the G7 and others will want to have their say and I think we can be a reference point in the gathering and looking forward, I mean we've heard some very interesting ideas for initiatives arising during the course of our dialogue we heard our Secretary General suggest that we have a child safety pledge acted upon, we heard him talk about a transparency initiative, we've just heard the President of Iceland talk about the national dialogue in that country, these are things we have to put together and I'm sure there will be many other initiatives and I think bring to this dialogue which can be I think an annual place to come together and to really take stock of what we're doing and where we need to go. So I think we've made an enormously encouraging start, but
Claire Melamed
Thank you, Under -Secretary General. And speaking of starts, what for you would success look like? And I should warn all of the other panellists, you're all getting a variation of this question. What would success look like by the time the global dialogue convenes again in New York in May 2027?
Guy Ryder
Well, we will have some things which we think with some confidence we can rely upon. We'll have another report of the International Scientific Panel, and we'll be able to chew on what they bring to the table. No doubt other things such as the initiatives I just mentioned, which have been launched at this dialogue this year, will be bringing new ideas to the table. But my view is that success will be both maintaining, the UN's position, as I've said, as a unique, universal convener. for dialogue and having the engagement of our member states, also of other stakeholders, to take forward, frankly, all of the ideas in terms of objectives, which have come out rather clearly from this dialogue. You know, having AI serve humanity, I know these are sounding rather high -level ideas, but to put those into some sort of operational form. And as I say, I think we've both got the commitment of stakeholders and the direction of travel that will help us to move towards progress in that regard.
Claire Melamed
Thank you so much, Under -Secretary General. Minister, let me now turn to you. We heard from you a moment ago, delivering your statement. Now, as you said then, as we've heard several times already this week, countries are moving at different speeds in developing AI governance frameworks. But, of course, all of the countries here share an interest and their people share an interest in ensuring that AI contributes to sustainable development and to inclusive growth. So from Namibia's perspective, what practical forms of international cooperation of the kind of debates that the Under -Secretary General just referred to are most needed to help countries build the capacities, the institutions and the partnerships that are required to govern AI effectively and deliver on the promise that we know there is?
Emma Theofelus
Thank you, VP, moderator. And thank you, Mr. President. Indeed, I concur with the Under -Secretary General that the convening of this dialogue is a great first step. Every country is grappling with, you know, how to best manage AI because it's inevitable. Nobody can wait. But I think there are three concrete steps we can take, one immediate and perhaps one moderate and some long term. And I think the first one would be looking at helping countries who are, in many instances, grappling with so many other structural issues, especially us from the global south, to translate those global AI principles into effective national laws and institutions. I think that would be a good foundation for any country to actually know, based on our context, how to handle the AI question. I was listening to one of the AI panel members yesterday who did say that context determines the outcome. And it's true. We have different regions. We have different contexts. So it's only right that we condense those based on the context of each country, whether from the global south, global north, or elsewhere. And then I think the second practical moderate approach that we can look at is continuously investing in digital and data infrastructure. The fact of the matter is some countries are still grappling with getting their citizens digitally literate. It's no secret. So to pretend that all of us are at some level the same is not true. So we must find a way to create or avail financing to countries that are still teaching their citizens how to trust technology, how to trust the ability to use technology in public service and otherwise, and then continuously elevate from there. And then finally, the long term would be creating programs to strengthen skills and scientific capacity. From a Namibian perspective of a country on the African continent, it's no secret that almost half of the world's population will come from the African continent by the year 2050. This means that we have not only the energy and agility of a young population, but we will have the numbers. So it's only right that, especially for the global south and the developing nations, that there are programs tailor -made to train human capacity and human capabilities for the future of the world. To strengthen skills and scientific capacity in the global south.
Claire Melamed
Thank you Minister and let me also ask you the same question that I asked the Under Secretary General this is a huge agenda that you have posed and one that is the work of many people in many years practically speaking what would you like to see by the time we get to the second global dialogue and we are perhaps sitting if we're invited back on another panel like this, what would you like to see, what would success look like, how would you know that we were moving forward
Emma Theofelus
Well the next one will be 10 months from now so there's a lot of work to do in less than a year but I think success in May 2027 in New York would look like one, having a shared but flexible reference framework for AI governance, I say flexible again as I've spoken about the context earlier before and this reflects on all the regions across the world and giving countries like ours small states, developing states like Namibia where I'm from a practical but sovereign template for national legislation and, of course, our institutions. The second one would be visible capacity gains in the global south. Like I said with the example earlier before, we need more trained regulators. We need more national AI strategies deployed and more cross -regional AI initiatives advancing inclusive growth and resilience, especially for countries that are compounded with AI coming up, but they do not necessarily have the capacity in -country, so that there's no dependence on other people's capabilities for one's own country. And last, reflecting on Goal 17, would be partnerships, global partnerships around this dialogue where governments, industry, academia, civil society are regularly exchanging. Preuves, updating norms. and co -developing open tools that we can use in administrating the use of AI for now and the future, and that the voices of some are not lesser than others. I'm a government representative, but many instances we depend on all these actors to properly advise policymakers such as myself to generally take a listening ear to actually put the necessary safeguards in place to protect citizens. So this is, I think, what success would look like in the next 10 months before we meet in New York.
Claire Melamed
Thank you. That is, as you say, a big agenda for 10 months. I don't want to quash your applause, please. And I'm very glad that you raised the question of partnerships. I've already heard many announcements of very exciting new partnerships, and for those that haven't yet, let me encourage you all to take a look at the partnerships. Partnerships help. that has been launched as part of this dialogue. There are some fantastic examples there of exactly what you described, practical coalitions of different stakeholders coming together to solve some of the practical problems that have been outlined here today. So it's great to see those solutions starting to come. Jan Tallinn, the Future of Life Institute. We've heard, I think, through the dialogue, recognition of the particular moment that we're in with the huge increase in AI's transformative opportunities and capable systems that we are seeing reflected. As you think about that, where would you see the greatest need for international cooperation in the face of the pandemic? In the face of this rapidly developing industry?
Jaan Tallinn
Thank you very much. Indeed, I think it's great to see this kind of awareness throughout the dialogue and like indeed in the world that finally AI is becoming very capable. I certainly see it in my own own field of software development, but clearly health care sciences, there's so much good that can be done with AI. But I think even the best experts are only now starting to grasp that the progress will actually accelerate from here. The big reason is that the top AI companies are now openly pursuing what's known as self recursive self -improvement that is using their own AI to accelerate the development of AI. Just yesterday, Jack Clark, co -founder of Anthropic, said that their top model is now did a new world record in AI. In one of the crucial tasks of machine learning, the GPU current development. now the problem with this recursive self -improvement is that you're risking losing the control because you're giving machines more and more autonomy and the autonomy that makes those more generally the autonomy that makes these systems useful is also what makes them harder to control as the secretary general said these systems are actually no longer tools awaiting instruction and our intuitions and institutions are not ready for machines that decide and in the worst case we indeed lose control altogether and as the panel's report makes it clear reliable methods for retaining control over highly autonomous AI systems are lacking but of course that's not inevitable the choice, as the secretary general puts it is between governing by design and drifting and that's not inevitable by default probably drifting in the in the wind of economic pressure. But in the end, it is the path that we, humanity, are choosing, and we can choose a better path. Because the reason why we are rushing towards these dangerous thresholds is because of the competitive dynamics driving this AI development. And indeed, even the leaders of the top AI companies now are voicing their caution. They are saying that they would like to slow down, but they really can't because of this commercial and geopolitical pressure. And this is where UN and more general international cooperation can step in to make sure that there is international pressure to progress at a safe pace so we can actually capture the AI benefits, the increasing capabilities that the increasing capabilities are bringing. And... And... And not kind of succumb to this loss of control that's driven by the race. and I think with this dialogue and the scientific panels independent evidence the UN obviously has the most legitimate organization on the planet is uniquely placed to build that cooperation Thank you, thank you Jan and and looking ahead to the next dialogue we've heard that there are many risks ahead and particularly as you've just said with the developments in the technology but we've also heard throughout the last few days excitement about innovation and opportunity how do you think that we should be collectively prioritizing to anticipate those risks but while also supporting the innovation that we want So I think it's again it's great to have this report because it's easy to just look at it and say that the report points like several key problems that we that they're going to face one is that the as i already mentioned ai capabilities are advancing faster than our ability to govern them second the companies now actually have the highest level of knowledge they know more far more about how capable and how safe their systems are really than the governments and the public and the third which is like now becoming very clear most countries including many advanced economies lack the technical expertise to assess the most capable ai or have a real say in governing it and because of these risks are actually cross borders even if you don't have a say you're still going to be subject to the risks the countries left out are exposed just like everyone else so kind of preaching the ai divide and governing the risks are not two separate agendas the one i think there's like one way of putting it is that there's like two frontiers that are being pushed one is making AIs more and more capable and that's kind of happening behind two gates one is what the AI companies are deciding to release and the other is what the host country is enabling the rest of the world to access and even the companies themselves as I mentioned are trying to go like help us let's be more careful here and the other frontier is basically the AI adoption diffusion and actually putting the capabilities into good use so there is really no dilemma anymore between innovation and safety because most of the safety concerns are in the first frontier where the companies are racing into unknown and most of the innovation and benefits actually come from the second frontier where the companies can companies and the community in general can actually tap into those capabilities.
Claire Melamed
Thank you very much and in the spirit of the AI and in the spirit of the dialogue of dialogues the title of this session I think you know bringing what you have said together about the the fact that these different issues although they may sometimes be discussed in different fora are not at all in contradiction with each other and under secretary general rider's comment about the un being the inclusive forum where these things can come together i think again this speaks to very much to what we are all doing in this room today and the value of this forum so thank you for that gaia coming coming to you last gaia marcus the director of the ada lovelace institute we've heard i think already from several panelists from the the president from her excellency the president of iceland earlier the critical importance of inclusion of public trust um and that has been a recurring theme across the different cluster conversations that have um that have been happening in the past and that have been happening in the past and that have been happening in these rooms what what practical mechanisms do you think can can help to ensure that the diverse voices that we need continue to shape AI governance between these two dialogues and beyond?
Gaia Marcus
Thank you. It is truly an honour to be addressing you all today. In reflecting what we have heard, it strikes me that we cannot understate how historic it is that all UN member states now have a seat at the AI table. And inclusion cannot stop at the level of states. For this dialogue to be legitimate, we need to move from hearing only about the futures that a few companies, investors or powerful actors want to see, as we have heard, to hearing the hopes, aspirations and fears of people around the globe, as AI becomes intertwined with everyday life. So embedding trust and inclusion into this dialogue cannot mean treating participation as a side event or a consultation exercise. Trust is built through reliability, responsiveness, integrity, fairness, openness and above all, trust. Above all, accountability. And these are qualities of the governance process itself. So the question becomes, how do we make sure that diverse voices are not just heard, but they truly shape decisions between Geneva and New York in 2027 and then far beyond? So there are two things that we would like to see, to see trust and inclusion really sit as the bedrock of this dialogue. First, we need to treat public participation as a source of evidence. Global AI governance should be shared, as we know, by a shared understanding of evidence, not only the technical evidence about AI systems, but evidence about how impacts are being experienced by real people across the world. Scientific evidence is essential, but in AI, consensus can take time and the evidence often lags behind the pace of change. As we heard yesterday, while incredibly laudable innovations, And we need to be aware of the importance of the role of the government in the development of the world. And we need to be aware of the importance of the role of the government in the development of the world. And we need to be aware of the importance of the role of the government in the development of the world. And we need to be aware of the importance of the role of the government in the development of the world. And we need to be aware of the importance of the role of the government in the development of the world. the independent scientific panel gives us the floor of our concern not the ceiling and while our focus must be on human rights protecting safe and trusted ai for all it is often the people most affected by these technologies that are the true experts in their impacts workers really understand the effects and implications of automation the most vulnerable women girls children bring lived experience of the hard edges of technologies that are deployed before they are governed and deployed before they are safe so second to know that we're making progress on inclusive and trusted AI participation it needs clear docking points into institutions and power inclusion of marginalized people and of those who have the most to gain from AI will not automatically lead to trust and it will not lead to change you unless their experiences and expectations are connected to clear levers of power. That means the dialogue must be the place where different mechanisms, be they a citizen's track of affected communities, national and regional dialogues, or participatory incident reporting mechanisms, the dialogue must be the place where these mechanisms contribute to outcomes that have clear accountability. Participation does not just mean collecting r representing views, but making sure that those views shape agendas, evidence, priorities and recommendations. So now I'd say is a moment for courage, not a moment to stand in beautiful contemplation of the various problems that we are all aware of. We need to move forward with a shared expectation that this dialogue can set a global AI governance floor with clarity on what is and what isn't acceptable, with respect for the rule of law, and with the full spectrum of international human rights as its foundation. So for 2027 we need to act decisively. developing an embedded and institutionalised participation architecture around the dialogue that adds substance to a recognised, respected and defended global governance floor. A floor that complements existing processes, but adds a shared global layer. A floor that recognises lived experience as evidence, gives communities clear routes into decision -making and creates transparent obligations that the dialogue is accountable for. That is how inclusion moves from being a theme to becoming a governance mechanism and how public trust becomes something that is earned, not something that is simply asserted.
Claire Melamed
Thank you, Gaia. Thank you so much. I mean, we've heard from all of you, I think, some recurrent themes and also some very strong directions for the next... dialogue. We've heard of the need... for a flexible framework, one that can accommodate the rapid changes in technology, the rapid changes in the needs of different countries and communities, and the rapidly evolving field of AI governance itself. We've heard of the need for an inclusive approach, inclusive of all UN member states, and inclusive of communities and of those voices that often are not heard in these debates, and of course inclusive of the companies who are developing AI and bringing it through the products into people's hands. We've also heard about the need for governance to have a practical application, for the link from a broad AI governance conversation to national implementation, to practical changes in people's lives, and of the need to link up with the needs of the world. So I think that's a really important point. the different conversations that are happening around the world. whether those be thematic in terms of addressing specific parts of the agenda, safety, innovation, and so on, or whether they be the conversations that are happening in different fora where not all are represented, although progress and new ideas can be made. So we need, I think, to recognize, as Under-Secretary General Ryder said, the critical role here of the UN as the place where all of these different attributes of dialogue can come together. We have a few more minutes, I think, of the panel. So I would like to ask all of you one final wrap-up question. As you have been listening to the conversations, participating in panels, having your, I'm sure, your many bilateral conversations over the last few days, what do you think, as we approach the final sessions of this dialogue, what in, well, it says in my script in one word, I might allow you three or four, what do you hope is carried forward when from here, from Geneva, from the UN, from the to New York when we meet again in May 2027. Let me just start with you, Minister, and go down the row.
Emma Theofelus
Thank you so much, Claire. I think for me, immediate word that comes to mind is equity. But another couple of words that comes to mind is collective progress. So equity and collective progress.
Claire Melamed
thank you thank you
Guy Ryder
guy i'll certainly agree with both of those uh those ideas but it seems to me from the conversation you know that if we are successful in what we are trying to do advance on the three sort of lines of action which which actually flow from the global digital compact uh which give birth to this dialogue. We need to advance on the governance discussion. It doesn't mean we'll have a fully fledged agreement around governance, but the parameters of the things we need to do around governance, as Minister, you said in your first intervention, need to be established. We'll continue to advance our understanding of the science and do our work on capacity building, as the Minister's emphasised. So, to get to your question, it seems to me that we will have been successful, and we will be successful to the extent that people believe, not just as an abstract proposition, but as their lived experience, as the last speaker said, that AI is beginning to work for them, that the balance of fear, concern, in the face of AI is actually moving towards optimism, not because people by nature get more hopeful, but because we're able to demonstrate that we are beginning to exercise that process of control, to access to access not just drift towards the futures of AI that we want. So I think it comes around to confidence, but not irrational confidence, founded confidence that
Claire Melamed
Thank you. Jan?
Jaan Tallinn
Urgency. Because I think the situation in AI is now on an exponential where the developments certainly happen on a quarterly basis. Soon they will be monthly and then weekly. And if the governance continues to work on a yearly basis, there's a problem.
Claire Melamed
Thank you. Gaia?
Gaia Marcus
I think we all know that we've made true progress where we have a shared understanding of evidence that goes beyond the scientific to look at the impact on people and societies. We have a shared understanding of accountability and we have a shared belief that the UN is the place to drive minimum governance standards that we can all be head to so that we can all reach the moment.
Claire Melamed
Thank you. I have five words from our four intervenants. They are optimism, progress, equity, urgency and understanding. That sounds to me like an excellent agenda and aspiration to take forward into our next global dialogue. So I thank you all very much for that. It now gives me great pleasure to turn back our program to our master of ceremonies, the co -chairs of the global dialogue on AI governance and giving them a moment to walk onto the stage. Let me also add my congratulations to them for a wonderfully organized and stimulating few days and for all of the huge amount of work that they've done. I know it's been done by them and by their teams who are also, I think, in the room over the intervening few months. I congratulate you on a fantastic dialogue.
Egriselda López
Thank you so much. Claire, thank you. Thank you for moderating this important conversation, but also to your insightful remarks. Definitely this panel has made clear that the conversation on AI governance cannot end with this dialogue. And the experiences, the ideas shared over the past two days underscore the advanced effective AI governance will require continued collaboration across sectors, regions, and initiatives. And we all have a role to play in carrying these critical messages forward to the next global dialogue. So now we're going to the next part of the segment. And then we will invite all the Thematic Clusters co -chairs of the four breakout groups, to come up to the podium to provide a report back. from all the breakout groups. So I kindly invite you to come here. Are we all here? Almost there. Okay. So I have some housekeeping guidelines for all of you. Are we okay? One. Two. three and four yes fantastic okay so this is what we're going to do we're going to be very efficient in time and i will ask for your help on this i know that the discussions were really interesting i know that you have a lot to say and we're so looking forward to it and we also have a summary to write right so everything will be really really useful but now i will give you each of you only two minutes yeah i i heard the side like so only two minutes for each of you so in total for each of the thematic cluster you're going to have four minutes why are we doing this is because as you know then we have the possibility to hear from you and we're going to be able to hear from you hearing back from member states and other stakeholders also on those discussions and some of the people that were going to listen after that, they were not able to speak throughout these two days. So I kindly ask all of you, dear co -chairs, to help the co -chairs of the Global Dialogue on this. Are we good? Sounds great. Okay, let us begin with the Thematic Cluster One, co -chair Mark-Alexandre Doumba , Minister of Digital Economy and Innovation of Gabon, and Rashid Khan, co -founder of Yellow AI.
Mark-Alexandre Doumba
Excellencies, distinguished guests, thank you very much. Yesterday with my co -chair, Mr. Rashid Khan, our cluster examined AI's opportunities and implications related to social, economic, cultural, linguistic, and technical dimensions. And I'll share with you a few takeaways. The first is that we have a lot of information and the first one is to not race to be first on AI. Race to use AI wisely The instinct to compete on speed misses the point The real contest is who builds the institutions and judgment to deploy AI well as opposed to who deploys it fastest even if this can feel counterintuitive in the world that we live in Second, most of AI's opportunity will come from adapting big AI to small AI AI today is shaped by Western infrastructure, language and values but the cognitive economy only works if we meet people where they are and that means the frontier isn't getting bigger model It's about local adaptation that lets an African former or a Pacific Island utility operator actually use this technology in their own context That's where the disproportionate value sits especially for economy whose knowledge has increased AI has always been tacit as opposed to codified The third point relates to the fact that AI is a paradigm shift. It's not business as usual. It doesn't ask us to optimize existing systems. It demands that we redesign them with new incentives, new measures of success, new assumptions about who gets to participate and how. There has never been more capital, talent, technology assembled than today. Yesterday, we agreed that we are only limited by our imagination. Will we use our imagination to distribute this abundance fairly or let the gap between who builds AI and who benefits from it become this century's defining inequality? It's really now up
Egriselda López
Fantastic. Thank you, Minister. Rashid.
Rashid Khan
Excellencies. Distinguished. Yeah, please go ahead. Yeah, please. Excellencies. I will give you 15 seconds more, Rashid. Go ahead. Yeah. Distinguished delegates, colleagues, a very good evening to you all. Mrs. Dumba has set out where cluster one is going and let me share what the room shared yesterday with all of you starting with first, access is not the finish line AI is probably the fastest growing technology in the history of the world yet access alone does not create prosperity or jobs capacity does our room's sharpest warning was the biggest risk is that most of the world remains spectators while others capture the value second, strategy is enablement not a compute race digital public infrastructure the foundation of AI as a literacy for national priority and serious attention to distribution and workforce transition getting AI to people and people through the change is where the real work lies third, sovereignty and openness are partners and not really opposites local models in local languages on data that belongs to the user, working with frontier models, because in AI, context matters the most, is what we realized. Finally, no citizen wakes up and demands an algorithm from us. They demand a healthier child, a better harvest, a faster way for the governments to respond, and AI is that means to get there. Outcomes are the mandate for all of us today, and those outcomes must be measured honestly, including AI's own footprint across training and inference. So cluster one's leaving remarks, we need to create more value than we capture. Thank you.
Egriselda López
Thank you, Rashid. That was awesome. Okay, now let's go to cluster two. Thank you.
Jovan Kurbalija
Go ahead. 43 intervenants pronounced 12 ,084 words, and three words, three C's have been resonated. Capacity, capability, and context. Next, during the discussion of three hours, we mapped some sort of taxonomy or mapping of the digital inclusion gaps and divides. First, access. This is the old one. Infrastructure, cables, if I can call it. Funding gap, second one. The third one, participation gap. This is a good example of having louder voices, but we still need louder voices for many communities which are affected by AI. Skills gap, which is also highlighted. Knowledge. Knowledge came as an interesting word, mentioned 35 times during the discussion in comparison to data 63 times. Therefore, knowledge is possibly more important than data in this context. Around the knowledge, there are a few keywords. Local creation, co -creation, and local expertise. Again, context, context, context. Language has been mentioned. There is another gap. Trust. standards and the open source. Therefore we got some sort of overall mapping of the question of inclusion gaps and the need for the second steps. I'm sure that this careful reading of this transcript together with other discussion can lead us to have really focused creative and impactful developments triggered by discussion during the global dialogue. I gave you 22 minutes, seconds. Thank you so much.
Samba Diouf
This divide, we applied on the AI, but it goes beyond that. We applied on the infrastructure. We applied on the connectivity. We applied on the device. We applied on the knowledge. So what needs to be done? We need connectivity. we need electrification we need investment all this grab it together will help us to benefit on the AI so bridging the gap is not only the AI, it's going beyond the AI for our countries like in West Africa we are not facing the same situation around the world if you compare to the US or Asia, for us we are on the connectivity level we are on the infrastructure level we are on the devices level and to go beyond the AI we need to fix those challenges so yesterday's discussion was about how to close the gap, we applied in our situation where we have connectivity issues infrastructure issues knowledge sharing issues We have people ready to learn. We have people ready to understand. We have people that accept to be teaching. So what we need is all the guys together. We go together. We share knowledge together. We share the platform together. We share everything together. This is the main thing that we can do to close the gap. Thank you.
Egriselda López
Thank you. Can I please ask you a round of applause to Jovan Kurbalija, Executive Director of Diplo Foundation and Minister of Communications and SELSI Communications of Digital of Senegal, His Excellency Samba Diouf. Thank you so much for this. Okay, now let's go to thematic cluster three. Can I hear from Minister Bogantes and also Rebecca.
Paula Bogantes Zamora
Today's discussion confirmed that effective AI governance does not require every country to follow the same model. It requires us to connect different approaches while respecting national context, regulatory autonomy, and different levels of institutional capacity. Several key messages that I would like to highlight. First, AI governance must become adaptive. Static rules will not be sufficient for systems that are even more autonomous, interconnected, and capable of acting across multiple steps. Second, interoperability must become practical. This requires common definitions, shared technical standards, and clear links between legal requirements, technical control, and evidence of compliance. Third, cooperation must move from principle to implementation, cross -border regulatory sandboxes, shared evaluation methods. and incident reporting mechanisms that can help countries' test systems learn from failures and prevent repeated harms. Fourth, trust requires transparency, independent evaluation, and accountability throughout the AI systems' life cycle, especially when AI systems affect access to health care, education, employment, credit, public services, and democratic participation. Fifth, global evaluation frameworks must reflect linguistic, cultural, and demographic diversity. Countries and communities currently underrepresented must help shape the benchmarks and evidence used to evaluate AI systems. And finally, as agentic AI systems gain greater autonomy, responsibility must remain clearly assigned across developers, deployers, and operators. Trustability and accountability. And meaningful human oversight must ensure that autonomy never becomes an accountability gap. Thank you very much.
Rebecca Finlay
Dear co -chairs, first and most importantly, I want to assure you we dialogued. We dialogued so much that I cannot begin to cover the additional comments building on those excellent summary remarks from my co -chair. I'll give you four priorities and related actions that came out of them. First, strengthen the independent scientific evidence base. The scientific panel's report informs the dialogue. It is crucial that we support it, that we cohere the body of evidence already gathered, link it up between the UN, the State of Safety Reports globally, and the International Network of AI Safety Institutes. And one recommendation we received was to think about how we align the seven working groups of the scientific panel to better inform the dialogue. And to form directly the four dialogue clusters in the next dialogue. Second, develop a shared minimum baseline for interpretability that builds on existing global frameworks and reinforces trust and openness, as my co -chair said. We have global commitments in place. We have the Hiroshima process, international law, human rights frameworks, the body of pledges from the AI summits. Now we need to open that with more open benchmarks, verification, independent testing, making sure that models reflect linguistic diversity, open science, and technical standards. The UN dialogue could be the place where that coherence happens, and we could dig into some of those challenges that are required for full equity and inclusion. Third, create continuity between the dialogues to advance progress. The work must continue after today. And fourth, finally, we heard, and I could not agree more, that multi -stakeholder participation is not a nice to have. It is necessity moving forward. The next few years. Are critical in shaping the future of AI. Let us make sure to take this opportunity to act together. Thank you.
Egriselda López
Thank you so much, Minister Bogantes, and also Rebecca for this important wrap -up. Okay, now let's go with the Matthew Cluster Four. We're going to hear from Linda Bonyo, founder and CEO of Lawyers of Africa. Linda, you were also with the Minister of Spain that couldn't join us, but you will be speaking on behalf of two,
Linda Bonyo
so please go ahead. Excellent. I'm speaking on behalf of the Minister who's headed to Spain to celebrate his win last night. And we had a really good time dialoguing and talking about football, and tonight you know what's going to happen to Switzerland. We did have conversations spanning from children. We think that it's agreeable that we must protect children, but two, there was consensus that children are not homogeneous. We have immigrant children, disabled children that we must take care of in this age. Number two, we talked about financing human rights. How do we finance human rights? in the digital age. Many civil society organizations or organizations just do not have the funding anymore. In this room, there are fewer civil society organizations this time. We must look at disruptive and new ways to do that. Number three, we talked about AI workers and who's missing in the room. We do not have voices from the AI workers. And for the fourth reason that I'm going to talk about, we spoke at length on why there's little participation from the global north. We agree with the evidence from the scientific panel that AI governance globally is being run by a few states. The power is concentrated there. And we know why. It is because of visa refusals. The last six months, 51 % of conversations on AI governance happened in Geneva. The visa refusal rates for Africa, Africa accounts for 7 out of 10 visa refusal rates. I can mention an AI worker from the commonwealth they could not make it here today because of visa refusals, we think that it is time that we open up that black box there must be transparency and accountability for the AI systems that are being used to determine visa refusals we also think that this is my personal view, it wasn't from the group but we need an AI passport, it must be easy for people to move when they want to invest in AI or they want to take part on AI governance, we should not pay for visas to come here, the average African has spent 2500 euros to get to this place, some of you spent only 200 we must rectify those to ensure that the global south participates we also talked about linguistic diversity and community let's build for communities we talked about Zimbabwe and Ubuntu, Zimbabwe's national strategy talks about Ubuntu that we must look at ourselves as a community and not look at human rights violations as something that happens to the other side but it happens to all of us we also talked about diaspora and those that are immigrants and refugees what happens in terms of this particular violations that are happening we also dialogued on artists that you need to put artists first musicians and creative because we sang before we spoke we also talked about legal certainty we need some legal certainty around this because rules are changing every single day but the law must be able to be predictive in this sense and finally we also talked about information integrity what about democracy and that's something that maybe we're dancing around it but we need to talk about air and democracy and what an individualist information ecosystem would look like I'd also, I have 35 seconds thank you, I'd also want to talk about the general sentiment in the room around what to do with algorithmic management and how to domesticate and implement all these principles that we talk about globally. Lawyers have launched an AI governance index for Africa and what we're seeing everywhere is that states will sign these treaties but do not domesticate them in their country. ILO just passed, you know, I think 193, you need to read it up on decent work for these AI workers that are crying every day on algorithmic management. It is time that states support and ratify these conventions to be part of
Egriselda López
Thank you so much, Linda. Thank you to the co -chairs of the thematic discussions. I kindly ask you to stay here while we hear from member states and stakeholders about their own reflections as well. So we will move to the interventions from the floor. As you know, this was based on prior inscriptions. Please come up to the lectern when your name is called. But before that, for those speaking on behalf of a stakeholder, can you kindly go to my right here to the secretariat to identify yourself? Having said that, I pass to my co -chair who will be guiding us and we will follow the same dynamic as the others to ensure fairness and also allow role participations. The mic will be automatically cut off. Co -chair.
Rein Tammsaar
Yes. Let's go. Let's move forward. So but before I give a floor to Deputy Minister of Foreign Affairs of Czechia, I would like to say that whether you would be able to. To speak or not, we really ask you, please do not forget to send your statement. valuable statements to the email address AI dialogue at UN .org Thank you very much. So first speaker Deputy Minister of Foreign Affairs
Marie Chatardová
Thank you Thank you Chair and good evening everyone. It's my great pleasure to speak within this plenary. I really appreciate the dialogue with different stakeholders and it was one of my top priorities during my UN ECOSOC presidency. The Czech Republic welcomes the publication of the first report on the independent international scientific panel on AI. Artificial intelligence has enormous potential. to drive innovation, economic growth, and societal resilience. Realizing this potential will depend not only on the rules we set, but also on our ability to develop talent, research capacity, and expand advanced computing capabilities. Building these capacities should therefore be an integral part of the global AI agenda. The Czech Republic is actively contributing to Europe's AI ecosystem through the Czech AI Factory, part of a growing European network supporting AI research, innovation, and access to advanced computing capabilities. As we discuss the future, we should focus on interoperability rather than fragmentation. Different countries and regions will pursue different approaches, but we should work towards shared principles and compatibility that facilitate international cooperation, innovation, research and trade, including through technical standards. A human -centric approach focusing on the needs and perspectives of human beings must remain at the heart of AI development and use, ensuring that people and their rights remain central throughout the lifecycle of AI systems. At the same time, AI governance must remain inclusive and multi -stakeholder. Public trust in AI depends. It depends on transparency, accountability and meaningful human oversight. AI governance requires global cooperation. The United Nations has an important role as a universal platform, bringing together governments, international organizations, the private sector, academia, the technical community, and civil society. Companies developing and deploying AI systems also bear a key responsibility for ensuring that these technologies are used safely and responsibly.
Rein Tammsaar
I thank Deputy Minister of Foreign Affairs of Czechia. Next speaker is Diplo Tech Solutions, Nafeesa Alshaala.
Nafeesa Alshaala
Thank you. Madam Co -Chair, Mr. Co -Chair, Excellencies and Distinguished Delegates, I'm honored to speak on behalf of Diplotech Solutions. To ensure this historic global dialogue succeeds, we must urgently move from abstract ethics to a functional inclusive framework. Today, a widening technical literacy gap risks turning the global South into passive rule -takers rather than rule -makers of the digital age. To prevent a digital iron curtain, Diplotech Solutions urges this dialogue to prioritize three immediate actions. First, we must institutionalize regularity harmonization. We need a common diplomatic language and shared ethics. We must ensure that the dialogue is in line with the technical guidelines. We must ensure that the dialogue is in line with the technical guidelines. in line with the technical guidelines. We must ensure that the dialogue is in line with the technical guidelines. We must ensure that the dialogue is We must ensure that the dialogue is such as the EU AI Act and the African Union's AI strategy to eliminate regularity heavens for unethical AI. Second, we must democratize technical expertise. True inclusion requires concrete capacity building through global AI capacity fund and decentralized tri -sectoral bibliotech hubs with diplomats, private innovators, and civil society co -design policy. Finally, we must address critical emerging blind spots. The environmental cost of the large compute systems, the threat of synthetic content to information integrity, and, crucially, the protection of of global data centers, recognizing them as a critical, non-military infrastructure. I think the unique value of this UN-mandated platform is its universal legitimacy. So let's utilize it to ensure that the Global South has an equal seat as architects, not just consumers of the global AI governance. Thank you so much. Thank you.
Rein Tammsaar
I thank the distinguished representative of founder. Next speaker is State Secretary in the Federal Chancellery of Austria.
Alexander Pröll
Excellencies, distinguished guests, yesterday I came to this dialogue with one clear conviction. The world needs a binding framework for the use of AI. Having listened to governments, scientists, industry, and civil society from every region, I can say this, that message has only grown stronger. We heard about the opportunities AI brings in health, education, agriculture, productivity, and many other fields. We also heard about the risks to human rights and to democracy. Both are real. Managing both at once is exactly why voluntary guidelines are not enough. Good intentions do not stop harm. Rules do. Austria's own experience shows that responsible AI use is possible without slowing down innovation. Four examples. First, our GovGPT platform already serves 180,000 public servants. Second, our digital skills initiative has reached thousands of citizens with free training. Third, our new minimum age of 14 for social media protects our children while keeping Austria open to technological progress. Fourth, and our European Charter on Digital Sovereignty proves that 27 countries can agree on shared digital rules very fast. If Europe can do this, the world can too. So as this dialogue concludes, Austria's call is we need a global framework for AI governments. The joint understanding that human rights apply at all stages, built on common principles, backed by real safeguards, including every country in this room, also those not yet fully at the table. This is not about slowing AI down. It is about making sure AI serves the people, not just where regulation happens to be strong. Austria stands ready to keep building this framework, together with all of you. Thank you so much.
Rein Tammsaar
I thank our distinguished representative for Austria. Next speaker is... UN-Habitat, Executive Director.
Anacláudia Rossbach
Thank you, co -chairs, excellencies, distinguished delegates. As Executive Director of UN-habitat, I would like to emphasize that cities and local governments are where artificial intelligence is already being applied every day. They make decisions that directly affect people's lives and generate crucial evidence for AI policymaking. Yet, despite this reality, they remain largely absent from global AI governance deliberations. This matters because AI must respond to the reality cities face today. Yet, AI is only inclusive as the data on which it is built. In many cities, people living in informal settlements and other vulnerable communities remain isolated. They are underrepresented. or entirely absent from official data sets. If these gaps are not addressed, AI risks reinforcing existing inequalities by overlooking those who are most in need of housing services and public investments. We are confronting a global housing crisis with 3 billion people lacking access to adequate and affordable housing and 1 billion people living in informal settlements. And AI can help cities address these challenges. This is precisely why I am pleased to share that together with ITU Secretary General Doreen Martin, I signed a joint invitation to United Nations entities to participate in the United Nations Pavilion at the Smart City Expo World Congress 2026 in Barcelona. The initiative also marks the 10th anniversary of the United Nations, the first ever in the history of the United Nations. initiative and includes the launch of the Working Group on Guiding Principles and Frameworks for People -Centered Smart Cities in the AI era, joint led by UNHAPTAT and ITU under this initiative. The Working Group will co -develop a People -Centered Smart Cities Toolbox to support policymakers, city leaders, and practitioners in designing and implementing human -centered digital transformation strategies. Looking ahead, UNHAPTAT can help ensure that cities' perspectives, particularly from rapidly urbanizing regions in Africa, Asia, Latin America, and Arab states, are reflected in the global dialogue through our regional and country offices. Through the Mayor's Alliance for Digital Cooperation, in collaboration with ODET and partners of the Global Urban Data Coalition, and by leveraging the World Urban Forum, this 14th session
Rein Tammsaar
I thank the distinguished representative of UN Habitat. Next speaker is Her Excellency Ximena Lincolao, Minister of Science, Technology, Knowledge and Innovation of Chile.
Ximena Lincolao
Excellencies, authorities and representatives, in conclusion of this dialogue there is a conviction which is very clear. The challenges presented by AI cannot be addressed in a silo fashion. We need common standards and strategic partnerships which are able to connect the competitive advantages across all our countries. This conversation, there are at least five conclusions that have emerged. The first is the urgency. AI is rapidly evolving that no one was expecting countries to have to act decisively and can broaden the capacities for their citizens and therefore modernize their economies. Secondly, is that we still have agency. AI is human creativity. Whilst there is more opportunities that citizens have in order to understand and use AI, then we'll have better capacity to gear the ethics around this. Also, self -governance informs societies to act towards crafting policies. AI is not only a technology, but also a critical infrastructure. Therefore, its governance requires systemic vision, which includes digital and the physical. Technology is also policy and energy and around industry as well. Talent, we have to invest in continuing accountability. Capital to have entrepreneurship programs such as the one that Switzerland has also deployed. Also, this these days has shown us that compute is one of the drivers for the new industry. looks at this very optimistically and there has to be responsibility to have renewable energies that are abundant and technical capacities which allow us to have an ideal society for this new stage we'd like to thank you for being able to first be able to participate in this dialogue and we have to move from intention
Rein Tammsaar
Minister of Chile next speaker is CEO of Current AI
Ayah Bdeir
Good evening we are at a very unique moment in time where surprisingly government and consumers and businesses all have consensus consensus that AI cannot be concentrated in the hands of the few AI must be sovereign and AI must serve all of humanity but the question is how do we build this AI in the public interest not how do we govern it, not how do we control it but how do we build AI in the public interest we already have models of international technical collaboration by looking at history the world wide web, the other transformative technology of modern times was built through an international technical collaboration Linux, which powers 95 % of servers in the world was built through international technical collaboration I'm the CEO of Current AI we are a non -profit, public -private partnership whose sole mission is to build public interest AI but our goal is not to build it ourselves entirely our goal is to see what others have built have put in the public domain and then invest in the connective tissue in between these tools and components in order to stitch together an end -to -end vertically integrated AI and an integrated AI stack We are a public -private partnership, meaning we have government partners, we have non -philanthropy partners, and we have private sector partners who all are invested in building different components of the technical AI stack. All over the world, there are thousands and thousands of builders, of governments, of nonprofits, of companies that are creating building blocks for an open source AI stack. Data sets, foundation models, evaluation toolkits, benchmark systems. The problem is that they are not stitched together and they are not interoperable. So that is what we are investing in. We've surveyed over 24 ,000 of these tools and components all over the world. We've identified the gaps, and now we're starting to work. And this is not theoretical. On Thursday, we'll be launching the MVP of the first example of an open, completely open, public and sovereign chatbot made entirely out of open source components that exist. this requirement of international technical collaboration that AI demands of us right now cannot happen through traditional collaboration through MOUs and lawyers we have to collaborate at the speed of AI and open code bases offer us this ability if the chatbot I was telling you about the foundation model is from Switzerland the compute comes from Finland the data sets are coming from India the safety safeguards are coming from a non -profit in the US privacy preserving data set platform is coming from another non -profit it's only about stitching them together
Rein Tammsaar
I thank distinguished representative of current AI next speaker is deputy minister of digital development communications and mass media of Russian Federation excellencies
Grigoriy Borisenko
distinguished guests you may need your headphones for next couple minutes because if you don't know Russian so far уважаемые коллеги ladies and gentlemen, colleagues R2C intelligence is changing the rules of the game companies and countries deploying AI are receiving tangible benefits however, access to frontier technologies are not universal the key players are the developers of models the cloud providers, the producers of chips and these players sometimes apply restrictions as a result, some get exclusive access whereas others face obstacles Russia has a full technical chain to create advanced models without needing to use any foreign components we have two large language models GigaChat and Yandex GPT we've also trained deep models based on national values and our history and traditions we are ready to share our experience with other countries to respect their cultural codes and cultural specificities what's important is that we will not be scraping and not be using this data to for our benefit that is a decision of principle on top of large language models we have a whole specter of classical technology is face recognition medicine biometrics and others we are ready to work as partners to elaborate projects with all countries present we're working with other countries and this kind of common initiatives gives access to instruments to open models and and to shape common standards Colleagues, this reduces the risk that AI will be the privilege of a small circle of people Russia is ready for collaboration
Rein Tammsaar
Thank you I thank the Honourable Deputy Minister of Russian Federation Next speaker is Head of Alive Research Group, Sheikh Hamidou Kaneh
Seydina Moussa Ndiaye
Digital University Thank you Mr. Chair Allow me to build on the statement delivered this morning This afternoon by Senegal As one of the co -facilitators of the UN -supported global network of centres for exchange and cooperation on AI capacity building I have had the privilege of working over the past month with outstanding institutions from every region To translate the vision of the global digital compact And the Secretary General's recommendation into a practical Mechanism for international cooperation What makes this initiative unique is that it is not creating another institution. It is connecting existing centers of excellence into a collaborative network where countries can exchange expertise, share knowledge, develop talent, strengthen national AI ecosystems, and support one another according to their respective strengths. In only a few months, participating centers have jointly developed a cooperation framework that defines common principles, governance arrangements, and priority areas for collaboration, including AI skills development, access to compute and data, support for national AI strategies, local innovation ecosystem, and the sharing of practical AI solutions. This demonstrates an important role. Global AI capacity building cannot be delivered through funding alone. It also requires trusted partnerships. peer learning, and long -term institutional cooperation. The network is therefore designed as a platform where governments, universities, research institutions, the private sector, and philanthropic organizations can combine the comparative advantages to accelerate AI capacity worldwide. As AI continues to evolve at unprecedented speeds, the question is not longer whether we should invest in capacity building, but whether we can do so collectively and at sufficient scale. We therefore warmly invite member states to nominate centers of excellence to join this growing community, and we encourage development partners, philanthropic organizations, and industries to contribute expertise and knowledge. Infrastructure and resources so that every country can become an participant, not merely a consumer. in the AI era. Finally, I would like to thank the United Nations Office of Digital Emerging and Technology for its continuous support throughout this process. We look forward to working with all member states as the network enters
Rein Tammsaar
I thank the Honorable Head of the Live Research Group. Next speaker is Vice President of the Romanian National Authority for Management and Regulation in Communications.
Pavel Popescu
Excellencies, distinguished delegates, Romania takes the floor in a spirit that has long defined our diplomatic tradition, the conviction that no era of transformation should be governed by force or by fortune alone, but by shared values. Nicolae Titulescu, former Romanian MFA and the long -standing president of the League of Nations Assembly, one of the architects of the interwar international order taught us that peace should not be only asserted, it should be also organized. AI is our generation's test of the same principle, and it is the work of organizing, not merely asserting, that brings us to this room. Romania comes forward with a single guiding proposition. When it comes to AI, we must be pro -innovation and responsible at the same time. These are not competing instincts, but the same duty seen from two sides. So how do we anchor these instincts? Through three main points. First, we need to encourage a human -centric transition. The move to an economy shaped by AI must be not allowed to leave the most vulnerable behind. Progress that concentrates. It concentrates its rewards and distributes only its disruptions. It's not progress. It's instability postponed. We therefore need a deliberate transition architecture through retraining and pathways into the new jobs this technology creates. Fortunately, this is not only a task for governments alone. We salute private actors investing in free, cross -sector training credentials and seek to scale this model through multi -stakeholder cooperation. Second, we believe in restraint where the stakes are the gravest. We therefore warmly welcome the bilateral dialogue underway between leading AI powers, and we hope to see them deepen. Where the destructive potential of AI is greatest, human judgment must remain sovereign, and restraint must be a matter of principle, not merely of presence. Third, protection of the children. This is one of the most important things. AI increasingly shapes what children see, learn, and are exposed to online, and we need to protect them above. This is a matter of principle, not only of presence. Romania believes that AI race should not be settled only by those who train the most capable models. Romania pleads for an approach that balances opportunity with risk and that keeps the door open for every country to be co -authors of this age, not their spectators. Tito Lescu warned that peace must be organized, and so this must be organized. I thank you and I encourage to shape the new era in the way that Mr. Tito Lescu.
Rein Tammsaar
I thank the distinguished Vice President of the Romanian National Authority for Management and Regulation in Communications. Thank you all for... Thank you, thank you to all of our distinguished intervenants and to everyone who participated in this very... very rich exchange. Your contributions have truly brought the spirit of this session to life, giving the real meaning to its title, Dialogue of Dialogues. Special thanks to fellow co -chairs of the thematic sessions. Many, many thanks for your contribution and for your efforts. And before we move to a closing ceremony, may I ask all of you to stand up and we will do a family photo first before you leave. Thank you.

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.