Forum francophone de la gouvernance du numérique et de l ia

Calls for Action Across Sessions

Throughout the various sessions of the forum, participants issued a wide range of concrete calls for action spanning infrastructure, regulation, AI governance, linguistic diversity, and international cooperation. Below is a unified synthesis of the key calls made.

Digital Infrastructure and Environmental Sustainability

In Atelier 2: Infrastructures numériques et empreinte environnementale, Ben Youssef called for better consumer education about environmental costs, warning that "si la transition numérique ne résout pas le problème environnemental on n'a pas une terre habitable dans 15 ans." The ARCEP representative called for pushing operators to adopt better codecs, promote reconditioned terminals, and favor fixed over mobile networks. Henri Monceau called for multilateral cooperation, stating "il n'y a pas de solution qui ne soit pas une solution de coopération internationale, multilatérale, au minimum plurilatérale." Speaker 3 called for creating at minimum a regional or francophone cooperation framework: "il faut qu'elle arrive à avancer dans cette optique de créer ne serait-ce qu'un cadre qui permet de mettre en place une première forme de coopération ne serait-ce que de façon régionale ou de façon francophone."

Sustainable and Sovereign Infrastructures

In Atelier 1: Des infrastructures durables au-delà des évolutions technologiques, Speaker 1 (ICANN staff) called for greater community participation in ICANN and IETF processes: "il faut que tout le monde en profite tout le monde le fasse parce qu'on est habitué à un rôle un peu consommateur de l'internet c'est pas du tout ce qui était prévu dès le départ." He also called for countries to develop local DNS infrastructure and use ccTLDs, and for states to ensure minimum local resilience: "l'État doit être capable de pouvoir avoir des accords avec les opérateurs pour pouvoir garantir un niveau de résilience au niveau local." Speaker 2 (ITU) called for governments to implement the 40 recommendations developed by the ITU's submarine cable advisory forum.

In Atelier 3: La souveraineté infrastructurelle est-elle primordiale ?, Emmanuel Agbenonwossi called for African governments to prioritize critical state infrastructures and invest in human capital, while warning against "romanticizing sovereignty" and calling for clearer pedagogical work on what sovereignty means. Rodrigue Guiguet-Md called for exploring the 'data embassy' concept as an interim solution, and Madame Lydie Oumanga called for basic digital literacy training for decision-makers and civil society engagement at the grassroots level.

In the closing session Des infrastructures durables et souveraines: une équation impossible ?, Luc Missidimbazi called for a common African vision and shared infrastructure platform: "il faudrait qu'il y ait un consensus. Il faut qu'il y ait une plateforme qui permet de créer ce cadre commun d'infrastructures de demain." Pierre Bonis called for investment or regulation to prevent eviction of intermediary actors: "investissez ou légiférez ou régulez, mais faites en sorte que le déploiement de l'IA ne crée pas de l'éviction des acteurs des infrastructures." Dr. Xianhong Hu called on participants to share and implement the UNESCO-ICANN policy note on Universal Acceptance: "Je vous invite à la partager, à la soutenir et à la finalement mettre en œuvre dans votre pays."

AI and Territorial Governance

In Atelier 4: Territoires et IA: comment les villes francophones passent à l'acte, Céline Colucci called for collective governance frameworks for AI deployment in local governments, including elected officials, management, and employees, and for working on 'sovereign AI' and smaller models respecting GDPR and public service values. Zacharia Oulad called for cities to take ownership of their own development using technology as a transversal tool and to leverage 'city diplomacy.' The representative from Global Citizens Hub called for cities and regions to be included in international AI negotiations, while the speaker from Côte d'Ivoire called for more equitable knowledge and resource sharing to prevent AI from deepening existing inequalities.

Digitization of Content and Linguistic Diversity

In Atelier 4: La numérisation des contenus inexploités: stratégies et outils, Frédéric Lemers called for increased digitization of heritage collections, harmonization of practices, and reconscientizing public authorities: "il y a une opportunité de reconscientiser les autorités publiques et de la nécessité de financer ces campagnes de préservation et de numérisation du patrimoine." Jérémy Petit called for creating trust with rights holders and local communities, ensuring transparency of AI systems, and implementing robust semantic infrastructure: "intégrer dans les stratégies, d'assurer, de créer la confiance avec les ayants droit, mais également les communautés porteurs de tradition orale, d'assurer la transparence des systèmes d'IA, s'agissant des contenus exploités, et finalement, de mettre en place aussi un système d'infrastructures, de couches sémantiques qui soient robustes."

In Atelier 2: La diversité linguistique dans les données d'entraînement de l'IA, Emmanuel called for digitizing analog knowledge stored in university libraries: "il y a des millions de thèses qui sont produites chaque année...qui sont restées dans les bibliothèques de façon analogique...il faut d'abord organiser ce corpus." Speaker 2 called for training 'digital librarians': "il faut des bibliothécaires parce que moi aussi je considère que les bibliothécaires c'est ceux qui savent gérer des données." Emmanuel also called for using existing translated documents such as the Bible as training data: "en Guinée par exemple, moi j'ai conseillé la Bible parce que les langues les plus parlées en Guinée, la Bible a été traduite dans ces langues."

In Atelier 3: Quels outils pour développer les contenus multilingues ?, Speaker 1 issued a broader call stating that "il ne faudrait pas qu'on essaye de penser que l'IA va résoudre le problème des humains et de notre langue. C'est aux humains de résoudre le problème de la langue française." Speaker 4 (Isidore) called for mutualizing efforts: "c'est pas de standardiser ou d'uniformiser, mais de mutualiser les efforts qui sont faits."

AI and Access to Knowledge

In Atelier 1: Quels mécanismes pour empêcher l'IA de restreindre l'accès à la connaissance ?, Mehdi Snene called for developing sovereign AI models through fine-tuning open-source models: "il faut vraiment se mettre dans cette situation et se dire que c'est des acteurs économiques qui opèrent sur un marché libre... prendre, fine-tuner, c'est-à-dire adapter des modèles qui existent sur notre propre langage." Jonas Erin called for inverting the burden of proof in AI procurement for education: "On ne veut pas que ça repose sur l'expert en éducation... Nous, on veut que ce soit le contraire. On veut que ce soit le prestataire de solutions qui veut intégrer le marché de l'éducation." Sébastien Machorel called for engaging civil society in AI governance: "si vous ne savez pas ce que veulent les utilisateurs il y a des organisations qui sont là pour ça la société civile est organisée mal financée mais elle est organisée donc demandez lui."

In Comment l'IA modifie l'accès à la connaissance?, Stéphane-Eloïse Gras called for defending open and transparent AI: "il n'y a pas de souveraineté sans ouverture... il n'y aura pas de souveraineté si on ne défend pas la possibilité d'avoir des intelligences artificielles ouvertes et libres." Xavier Merlin called for reaffirming the open internet principle: "réaffirmer principe de l'Internet ouvert, justement à l'heure de l'IA générative dans un double objectif. Un, de garantir la liberté d'accès à la connaissance et deux, la liberté d'entreprendre." He also called for new techno-economic interconnection models via open standards, and for expanding the Digital Market Act to include cloud providers as gatekeepers. Seydina Ndiaye called for reinforcing each country's capacity to produce their own AI models: "le capacity building, c'est vraiment le cœur des discussions, je pense."

Regulation of AI

In Atelier 1: Pourquoi réguler ?, Oumar Agné called for data-centered AI regulation: "La question aujourd'hui, l'IA, la question c'est les données. L'IA n'existe pas sans l'IA. Et la question de régulation de l'IA, il faut faire des paliers." Billy Mikaël from Côte d'Ivoire called for anticipatory regulation: "il faut non seulement réguler pour limiter la fracture, mais aussi il faut anticiper. Il faut réguler pour anticiper aussi ce qui va arriver plus tard."

In Atelier 4: Quelle régulation pour encourager l'innovation ?, Dawit Bekele called for preserving the fundamental properties of the Internet and for multi-stakeholder collaboration: "il faut créer un partenariat avec des gens qui s'y connaissent." Laurent Ferrali called for reducing administrative burdens and opening public procurement to startups. The vice-president from DRC called for inter-ministerial coordination: "comment ne pas mettre autour de la table en Afrique un ministre de l'énergie...un ministre de l'éducation."

In Atelier 3: À quelle échelle réguler ?, Laurent Werly called for parliaments to develop a shared legislative corpus and exchange best practices: "déjà d'échanger les bonnes pratiques," and for feeding AI systems with Francophone content: "on n'oublie pas de nourrir les IA...de contenus francophones." Colonna Dissra called on non-Council of Europe states to engage with the Convention framework: "je dis aux petits États qui ne sont pas membres du Conseil de l'Europe...de se rapprocher du Conseil de l'Europe, de se rapprocher de la Convention de cadre."

In Atelier 2: Comment réguler, Patricia Amand called for African states to develop their own digital content and strategies: "il faut que déjà, nous, états africains, nous commencions à penser nos contenus, à penser nos stratégies de développement numérique et de gouvernance numérique." Thomas Dautieu called for regulators to be credible through understanding the technology and engaging in dialogue with industry and civil society.

Overarching Calls for Francophone Cooperation and AI Governance

In the closing plenary Quel équilibre entre règlementation et innovation ?, Amandeep Singh Gill called on the francophone space to "be exemplary, make the francophone space a living laboratory, with open corpora, shared data, multilingual models and talents trained on five continents." Henri Verdier called for making foundational models a common good of humanity and treating AI literacy as important as alphabetization. Lacina Koné called for proactive regulation centered on citizens rather than technology, and for African states to develop data governance frameworks. Lena Dargham called for greater francophone participation in international standardization bodies and for the OIF to inventory existing regulations among member states and work toward harmonization. Anne-Marie Jean called for francophone states to support the addition of a protocol to the 2005 UNESCO Convention on cultural diversity to address the digital dimension, and for convergence of francophone positions in international forums. Henri Monceau called for developing native language models and announced a workshop in Alexandria in October on this topic.

Commitments and Pledges Across Sessions: A Unified Summary

Across the various sessions, formal commitments and pledges were relatively rare, with most discussions remaining exploratory and advisory in nature. However, several concrete commitments did emerge, ranging from institutional agreements to legislative references and upcoming initiatives.

Institutional and Organizational Commitments

The most concrete and formal commitment was the signing of a Memorandum of Understanding (MOU) between AFNIC and Smart Africa, described in the Quel équilibre entre règlementation et innovation ? session. Pierre Bonis described the goal as "to pool the networks of national registries on the African continent... and the actions of Smart Africa", focusing on DNS infrastructure sovereignty and south-south competence transfer. In the same session, Henri Monceau committed the OIF's resource center to developing an AI tool to help member states navigate regulatory and normative texts, built with a laboratory based in Dakar.

In the Des infrastructures durables et souveraines: une équation impossible ? session, Oracle (Annick Sakho) committed to having all data centers run on renewable energy by 2035, noting existing data centers in South Africa, Morocco, and soon Kenya. France (Ambassador Céline Jurgensen) referenced its commitment to the International Coalition for Sustainable AI, which grew from 90 to 270 members, including 15 governments, 70 companies, and 9 international organizations. Additionally, UNESCO and ICANN committed to launching the French version of their joint policy note on Universal Acceptance, with plans to translate it into all UN official languages and Portuguese, and potentially indigenous languages. Luc Missidimbazi described a concrete infrastructure initiative: "Nous nous sommes associés avec la centrale électrique du Congo, avec le centre d'excellence, avec l'ONUDI et les opérateurs, pour étudier le financement d'une infrastructure qui va à la fois mutualiser les infrastructures de télécommunications et du numérique."

National and Governmental Commitments

In the Atelier 1: Pourquoi réguler ? session, several national commitments were mentioned. Côte d'Ivoire (Billy Mikaël) mentioned the development of a national AI strategy with an Objectif 2030. Mauritania (Malick Ndiaye) described an education reform already underway, stating: "on a fait une réforme de l'éducation qui a entré le 1er janvier 2026. C'est les langues nationales qui sont à l'école." In the Quel équilibre entre règlementation et innovation ? session, Anne-Marie Jean referenced Quebec's Law 109 (adopted December 2025) on discoverability of French-language cultural content as a legislative commitment, and mentioned that the OIF Secretary General announced the nomination of a special envoy on discoverability of francophone cultural content.

In the Atelier 3: À quelle échelle réguler ? session, Monaco (Colonna Dissra) described ongoing commitments including implementing a sovereign cloud, developing a sovereign AI platform with a European partner, and running a 'Maison du numérique' for citizen digital training.

Upcoming Events and Reports as Commitments

Several speakers announced upcoming deliverables as forms of commitment. In the Atelier 1: Des infrastructures durables au-delà des évolutions technologiques session, the ITU representative (Speaker 2) announced that "en la prochaine semaine, les 10 de juillet, nous allons approuver un rapport de 200 pages qui explique nos recommandations" on submarine cable resilience.

In the Atelier 2: La diversité linguistique dans les données d'entraînement de l'IA session, Speaker 5 (Henri) announced an upcoming workshop in Alexandria on October 6-7, stating: "nous aurons le 6 et le 7 octobre prochain à Alexandrie...un atelier...centré sur cet enjeu et qui visera à essayer de comprendre comment on peut dans l'espace francophone, et en particulier en Afrique francophone, développer des modèles de langage natifs."

In the Atelier 1: Quels mécanismes pour empêcher l'IA de restreindre l'accès à la connaissance ? session, Jonas Erin (Council of Europe) announced: "Nous allons publier, ce sera adopté prochainement très vraisemblablement, une recommandation sur la littératie de l'IA dont la spécificité est d'inclure un volet humain", along with an AI evaluation tool for education systems.

In the Comment l'IA modifie l'accès à la connaissance? session, Seydina Ndiaye announced: "on va lancer dimanche, avant le dialogue, le réseau international des centres pour le renforcement de capacité, qui est supporté par les Nations Unies et qui était également l'une des recommandations qu'on avait faites dans le groupe consultatif."

In the Atelier 1: Pourquoi réguler ? session, WIPO committed to organizing a forum on intellectual property and AI in November, and the OIF (Wassim Baccari) mentioned organizing a workshop in October dedicated to linguistic diversity and AI.

Ongoing Institutional Work Presented as Commitments

In the Atelier 3: Quels outils pour développer les contenus multilingues ? session, Speaker 4 (Isidore Sen, OIF) mentioned establishing a reference framework for cultural policies across member states and initiating training for public decision-makers on discoverability issues in partnership with the French Ministry of Culture. Speaker 8 mentioned an exploratory project from Quebec in collaboration with the Francophonie Numérique network to build a database feeding AI models with francophone and indigenous content. In the Atelier 3: À quelle échelle réguler ? session, Speaker 1 (Laurent Werly) described the APF's ongoing work on a shared legislative corpus in partnership with UNESCO, based on a signed convention between APF and UNESCO.

Sessions with No Formal Commitments

Several sessions produced no formal commitments or pledges, with discussions remaining deliberative and advisory: Atelier 2: Infrastructures numériques et empreinte environnementale, Atelier 3: La souveraineté infrastructurelle est-elle primordiale ?, Atelier 4: La numérisation des contenus inexploités: stratégies et outils, Atelier 4: Quelle régulation pour encourager l'innovation ?, and Atelier 2: Comment réguler. In the Atelier 4: Territoires et IA session, while no formal pledges were made, Zacharia Oulad described Agadir's commitment to using AI-supported hypervisors for urban management and water reuse as an ongoing initiative.

Technology Transfers Across Sessions: A Unified Summary

The issue of technology transfer was a recurring undercurrent across many sessions at the forum, though it was rarely addressed as a formal agenda item. Discussions ranged from infrastructure capacity-building and computing access to AI model deployment and regulatory frameworks. The following synthesis draws together the key themes and concrete suggestions that emerged.

Infrastructure and Capacity-Building

The most direct and substantive discussion of technology transfer occurred in Atelier 3: La souveraineté infrastructurelle est-elle primordiale?, where Speaker 1 (Pierre-Louis) recounted a cautionary tale about failed skills transfer in African registries: training was conducted on DNSSEC and IPv6, but without a functioning local registry to absorb the trained engineers, they ended up working for foreign organizations. His recommendation was clear: investment and training must be planned simultaneously. As he noted, "There was a time when we did skills transfer... we had 20 people, we launched DNSSEC, IPv6... Great, but there was no registry. So these people were trained, and once trained, they couldn't implement their training. So what did they do? They ended up going to work for foreign organizations." Emmanuel Agbenonwossi went further, recommending that Guinea establish a local school of excellence: "You need to have your school of excellence in Guinea. Train people, bring expertise, do the transfer of competencies and knowledge." Speaker 4 further noted that African countries lack experts to even write technical specifications for infrastructure projects, making them dependent on foreign consultants.

Access to Computing Resources

In Atelier 2: Infrastructures numériques et empreinte environnementale, the inaccessibility of GPU computing resources for African developers was raised as a critical barrier. Speaker 3 expressed concern about the dependency on foreign infrastructure: "Des fois, on développe en intégrant quelques agents, et quand on voit, on voit dire non, il faut aller louer ça au Canada... même louer du GPU, ça n'existe pas. C'est ça qui me fait peur." No concrete recommendations were made in this session, but the concern implicitly called for greater technology access and transfer.

This theme was echoed in Quel équilibre entre règlementation et innovation?, where Lacina Koné questioned why technology holders do not come to Africa given its resources: "If we have the energy and we have the coltans, why don't these technology holders come to Africa?" He also discussed the 'Compute Embassy Law' initiative by Smart Africa to democratize access to computing power across African countries through harmonized cross-border policies. Pierre Bonis described the AFNIC-Smart Africa MOU as aiming for 'south-south competence transfer' to strengthen DNS infrastructure sovereignty. Bénédicte Wechsler highlighted Switzerland's collaboration with Data Science Africa in Kenya through the ICANN (International Compute for AI Network): "they have very good researchers in AI and data science, but they have neither the infrastructure nor the data. But the collaboration is fantastic."

Cloud Technology and Licensing Models

In Des infrastructures durables et souveraines: une équation impossible?, Annick Sakho described Oracle's model of bringing cloud technology to the African continent: "la technologie de nos clouds, de nos data centers qu'on a de par le monde a été également ramenée sur le continent." She also mentioned white-label licensing arrangements as a potential model: "mettre même la technologie dans le pays qui sera gérée par un partenaire sous licence blanche comme on dit ou c'est le partenaire lui-même qui va exploiter cette technologie là." Luc Missidimbazi raised the concern that Africa has historically been a downstream client in infrastructure design: "l'Afrique vient comme un client en bout de chaîne." Aurélie Simard referenced open-source infrastructure principles and Digital Public Infrastructures (DPI) as means of enabling technology access.

AI Models and Linguistic Transfer

In Comment l'IA modifie l'accès à la connaissance?, Stéphane-Eloïse Gras highlighted model transferability as a technical mechanism that could help underrepresented languages: "sur ce qu'on appelle la transférabilité des modèles, va permettre potentiellement sur un set de données, un jeu de données qualifié en Kabil ou en Wolof, de procéder à des traductions, procéder à de la production de contenu, sans avoir besoin de toute la masse des infrastructures."

In Atelier 2: La diversité linguistique dans les données d'entraînement de l'IA, Speaker 2 (Martin) discussed the need for sovereign AI capacity through local deployment: "si on est vraiment fourrin, il faut pouvoir déployer chez soi."

In Atelier 1: Quels mécanismes pour empêcher l'IA de restreindre l'accès à la connaissance?, Mehdi Snene drew a parallel with the COVID vaccine situation to highlight the asymmetry in technology and knowledge transfer: "on a partagé les données, on n'a rien reçu." He also noted the absence of knowledge transfer from large AI developers: "ni un savoir-faire qui est véhiculé pour nous pour pouvoir aussi faire ça." The suggestion to fine-tune open-source models locally was the closest concrete recommendation in this session.

Urban and Territorial Dimensions

In Atelier 4: Territoires et IA: comment les villes francophones passent à l'acte, Mickaël Bilet raised the concern about inequality: "comment ne pas créer plus d'inégalités qui existent déjà." Zacharia Oulad suggested that cities should network and share experiences through platforms like AIMF rather than waiting for richer cities to assist poorer ones: "des réseaux comme l'AIMF peuvent vous mettre, justement, en réseau avec d'autres qui souffrent du même problème." Céline Colucci suggested sharing best practices and building local alliances as a form of knowledge transfer.

Regulatory Frameworks and Co-construction

In Atelier 4: Quelle régulation pour encourager l'innovation?, Yadem Saoudi explicitly referenced technology transfer in the context of co-construction with African countries: "une co-construction avec des données qui sont à l'intérieur, parce qu'on parle de souveraineté nationale...une volonté d'appartenir à cet état-là et de faire un transfert de technologie, de connaissances qui est aussi important."

In Atelier 3: À quelle échelle réguler?, the Ambassador of Mauritius called for shared knowledge and benefits from AI: "il faudrait aussi avoir un partage des savoirs pour que tous les bénéfices que peut générer l'IA soient partagés et que les connaissances et les savoirs soient partagés avec tous les autres pays au même pied d'égalité."

Knowledge Sharing and Francophone Solidarity

In Atelier 4: La numérisation des contenus inexploités: stratégies et outils, Speaker 1 emphasized solidarity within the francophone network: "il y a cet enjeu de solidarité au sein du réseau francophone numérique qui consiste à développer des formations de manière à permettre de faire relever le niveau moyen de maturité de chacun des stades d'une pyramide de la maturité numérique."

In Atelier 3: Quels outils pour développer les contenus multilingues?, Speaker 8 suggested that Francophonie institutions could serve as vehicles for sharing good practices: "on pourrait peut-être tirer mieux parti de ces instances-là pour partager, par exemple, les bonnes pratiques qui vont certainement inspirer les autres, que ce soit au Québec, au Camp de Tivoire, en RDC ou ailleurs."

Overall Assessment

Across all sessions, technology transfer emerged as a persistent concern rather than a resolved agenda item. The most concrete recommendations came from Atelier 3: La souveraineté infrastructurelle est-elle primordiale? (simultaneous investment and training, local schools of excellence) and Quel équilibre entre règlementation et innovation? (the Compute Embassy Law, south-south competence transfer through the AFNIC-Smart Africa MOU, and international research collaborations). Elsewhere, discussions remained at the level of identifying disparities — in computing access, infrastructure ownership, AI capacity, and knowledge flows — without proposing formal implementation mechanisms. A shared theme across sessions was the risk of African and francophone countries remaining passive recipients of technology rather than active co-creators, and the need for structural changes to reverse this dynamic.

Across multiple sessions, discussions consistently highlighted deep coordination gaps in digital infrastructure investment and stark geographic imbalances, particularly affecting Africa and the Global South.

Concentration of Physical Infrastructure

The uneven distribution of data centers and connectivity infrastructure was a recurring concern. In Atelier 2: Infrastructures numériques et empreinte environnementale, the moderator noted that Africa has only about 220 data centers in 38 countries, representing less than 0.02% of global installations, with South Africa, Kenya, and Nigeria concentrating 40% of this capacity. In Atelier 3: La souveraineté infrastructurelle est-elle primordiale ?, Emmanuel Agbenonwossi illustrated the consequences of this dependency: "Each time, for example, there is a power cut in a key state in Portugal, all public services, all banking services, etc., are blocked in these coastal countries." Speaker 4 in the same session noted that African countries had no participation in the architecture of submarine cables: "Today, the last one deployed, 2Africa, no African participated in the architecture. No article on the design."

In the closing plenary Quel équilibre entre règlementation et innovation ?, Henri Monceau cited the independent scientific group's finding that "75% of global computing power dedicated to AI is in the United States, 15% in China. In other words, the vast majority of countries share a residual fraction of this strategic capacity." Lacina Koné added: "How much data is controlled on AI servers from Africa? Only 2%."

GAFAM Dominance and Telecom Operator Decline

In Atelier 1: Des infrastructures durables au-delà des évolutions technologiques, Speaker 3 raised the issue of value concentration: "les GAFAM aujourd'hui concentrent toute la valeur qui est créée. Et que les opérateurs tel qu'on ont de moins en moins de valeur, donc ils ont de moins en moins de capacité à investir dans l'infrastructure." Speaker 2 noted that 90% of submarine cable ownership has shifted to hyperscalers. This was echoed in Des infrastructures durables et souveraines: une équation impossible ?, where Pierre Bonis observed: "les principaux constructeurs d'infrastructures physiques... ne sont plus les fournisseurs d'accès... C'est les GAFAM." Luc Missidimbazi further highlighted Africa's structural disadvantage: "si vous avez fait un câble qui part du Portugal à l'Afrique du Sud, on vous met en bout de chaîne et dans le segment, en fait, ça vous coûte cher." He also noted that "le régis de nos identités numériques de pays, les CCTLD, sont quasiment gérées beaucoup même à l'étranger."

Intra-Francophone Geographic Disparities

In Atelier 4: Territoires et IA: comment les villes francophones passent à l'acte, the gap between well-resourced and under-resourced municipalities was starkly illustrated. Sébastien Bachelet cited the Nièvre department's 27 million euro investment, arguing that "le véritable enjeu n'est plus seulement d'avoir la fibre, c'est de faire un levier d'attractivité, de compétitivité, d'amélioration du quotidien." The speaker from Côte d'Ivoire contrasted this with his city's reality: "toute échelle respectueuse c'est pour ça que je parle de fractures."

In Atelier 4: La numérisation des contenus inexploités: stratégies et outils, Speaker 3 captured the infrastructure paradox facing digitization efforts: "tout le monde dit aujourd'hui il faut numériser, derrière il faut des serveurs des data centers pour les stockings et tout le monde dit ça ne va pas, il n'y a pas d'électricité."

Governance Failures and Capacity Gaps

In Atelier 3: La souveraineté infrastructurelle est-elle primordiale ?, Rodrigue shared the experience from Burkina Faso where a Government Cloud project failed partly due to lack of local expertise to even write terms of reference, while Speaker 4 noted that bandwidth on existing cables is often only 10% utilized even as countries seek to invest in additional cables. In Atelier 2: Comment réguler, an audience member summarized the structural problem: "on a très bien constaté que l'Afrique, les États-Unis, la Chine ont le monopole, j'allais dire le quasi-monopole sur les infrastructures. L'Afrique n'est pas maître de l'infrastructure."

Proposed Solutions

Several sessions pointed toward potential remedies. In Quel équilibre entre règlementation et innovation ?, Lacina Koné proposed the 'Compute Embassy Law' to allow cross-border sharing of computing resources among African states, and argued that Africa's untapped energy assets — 3,000 hours of sunshine per year — could host AI data centers. In Des infrastructures durables et souveraines: une équation impossible ?, Annick Sakho pointed to Oracle's data centers in South Africa, Morocco, and Kenya as steps toward addressing geographic gaps, while encouraging regional resource sharing. Henri Verdier noted in Quel équilibre entre règlementation et innovation ? that unlike the internet, the AI revolution "is born in companies worth 1,000 billion dollars. It is born private at birth," making deliberate policy intervention to correct geographic concentration all the more necessary.

The question of proposals for the implementation of the "Global Digital Compact" was not discussed in any of the sessions covered in this meeting.

The question of proposals to strengthen countries' digital sovereignty was not discussed in any of the recorded sessions.

Across the various sessions, no comprehensive or formal proposals were made specifically to strengthen the participation of women, youth, and civil society in AI and digital negotiations. However, several scattered references and partial initiatives emerged from different workshops.

Civil Society Participation

The most direct call for civil society inclusion came from Atelier 1: Quels mécanismes pour empêcher l'IA de restreindre l'accès à la connaissance?, where Sébastien Machorel stressed that civil society is organized and ready to contribute: "la société civile est organisée mal financée mais elle est organisée donc demandez lui nous sommes là pour ça." Similarly, in Atelier 2: Comment réguler, Thomas Dautieu underlined the importance of dialogue with civil society for regulatory credibility, though without making concrete proposals.

Women's Participation

The most notable reference to women's inclusion came from Atelier 3: La souveraineté infrastructurelle est-elle primordiale?, where Lydie Oumanga described her grassroots initiative: "With the Association des femmes d'exception du Congo that I set up, this year, it's digital inclusion. So we do workshops just to explain what your phone is useful for beyond taking selfies." However, this initiative focused on digital literacy rather than formal participation in AI negotiations. A brief mention of gender equity appeared in Atelier 4: La numérisation des contenus inexploités, but without substantive proposals.

Youth Participation

Youth were referenced primarily as end-users or beneficiaries rather than active participants in governance. In Atelier 4: Territoires et IA, a speaker from Côte d'Ivoire raised concerns about young people's superficial engagement with AI: "On a une grosse population jeune qui utilise aujourd'hui l'IA générative...mais ils ne vont pas réellement dans l'outil afin de pouvoir en tirer ses réels bénéfices." In Atelier 1: Quels mécanismes pour empêcher l'IA de restreindre l'accès à la connaissance?, Jonas Erin mentioned the need to include students as active participants in AI governance within education, though without a formal proposal.

Multi-Stakeholder Approaches

A broader multi-stakeholder ethos was expressed in several sessions. In Atelier 4: Territoires et IA, Sébastien Bachelet argued: "je suis absolument persuadé qu'aujourd'hui, on ne peut plus compter que sur les gouvernements", calling for users' voices to be heard. Citizen participation initiatives such as Arras's "mur de l'IA" and Rennes's permanent citizen council on digital issues were described by Céline Colucci as practical examples of inclusive governance at the local level. In Des infrastructures durables et souveraines, Aurélie Simard referenced participatory approaches: "des infrastructures d'apprentissage où on associe les communautés, que ce soit les étudiants, les entreprises, les organisations non gouvernementales qui peuvent donner à remonter de l'information."

Overall, while the importance of inclusive participation was acknowledged across multiple sessions, no session produced concrete, actionable proposals specifically targeting the formal inclusion of women, youth, or civil society in AI and digital negotiations.

Unified Summary: Concerns and Proposals on the Nexus Between Development, AI, and Digital Technologies

Across the sessions, a rich and urgent set of concerns emerged about the relationship between AI, digital technologies, and sustainable development — alongside a range of concrete proposals to ensure these technologies genuinely serve human and planetary well-being.

1. Environmental and Infrastructure Costs

A recurring concern was the environmental footprint of digital infrastructure and AI. In Atelier 2: Infrastructures numériques et empreinte environnementale, Speaker 1 highlighted the full lifecycle environmental cost of digital infrastructure, from rare earth extraction to end-of-life disposal. The session's central framing was the concept of a transition jumelle (twin transition): "si la transition numérique ne résout pas le problème environnemental on n'a pas une terre habitable dans 15 ans... c'est l'une qui doit servir l'autre."

In the plenary session Des infrastructures durables et souveraines: une équation impossible?, Ambassador Jurgensen quantified the stakes starkly: "L'entraînement d'un seul modèle comme Chad GPT-4 requiert l'équivalent de la consommation électrique annuelle de 460 000 personnes en Afrique subsaharienne." Céline Colucci in Atelier 4: Territoires et IA also raised environmental concerns about data centers, while Speaker 3 in the infrastructure atelier cited a reportage about communities in South Africa being deprived of electricity due to data center prioritization: "on choisit des serveurs plutôt que des hommes."

2. Equity, Labor, and Social Dimensions

Multiple sessions raised concerns that efficiency gains from AI do not automatically translate into equity. In Atelier 4: Territoires et IA, Zacharia Oulad emphasized: "efficience et efficacité ne veulent absolument pas dire équité." Céline Colucci warned against AI being used to cut human labor rather than liberate workers: "Chez Amazon, l'IA, elle sert à tailler au maximum le temps humain et pas à libérer l'humain."

In the plenary Quel équilibre entre règlementation et innovation?, Henri Verdier raised concerns about AI replacing workers, citing the Alibaba CEO's statement at Vivatech about replacing all employees with AI subscriptions, calling it a race toward a cliff. He also raised concerns about AI being born in trillion-dollar private companies rather than as a commons.

3. Concentration of Power and Extractive Dynamics

Several sessions flagged the hyperconcentration of AI development among a small number of powerful actors. In Des infrastructures durables et souveraines, Pierre Bonis warned that "le développement de l'IA générative... c'est des gens qui s'adossent aux oligarques entre eux." Aurélie Simard noted that open-source infrastructure is collapsing due to extractive AI practices: "Un certain nombre d'elles sont aujourd'hui en train de s'effondrer."

In Atelier 1: Pourquoi réguler?, an economics professor raised the concern that "ce qui rend l'intelligence artificielle intelligente aujourd'hui, c'est l'intelligence humaine. C'est toutes les données humaines qui nous la nourrissent. Or, le problème, c'est que ces données humaines ont été prises de l'humanité sans aucune régulation."

In Atelier 4: La numérisation des contenus inexploités, Speaker 3 drew a parallel with raw materials extraction: "L'IA va arriver encore, on va produire des connaissances et on risque d'être pauvre."

4. Linguistic, Cultural, and Knowledge Exclusion

A major cross-cutting concern was the exclusion of African and other non-Western languages and cultures from AI systems. In Atelier 2: La diversité linguistique dans les données d'entraînement de l'IA, Speaker 1 noted: "95% des êtres humains utilisent 5% des langues," and cited a study showing AI systems reflect a WEIRD bias: "ça s'appelle Weird, Western, Educated, Industrialized, Rich and Democratic. C'est ça le biais des IA."

In Atelier 4: La numérisation des contenus inexploités, Speaker 1 warned: "si les données sur base desquelles elles ont été formulées, leurs élèves ne sont représentatives que d'une partie du savoir de l'humanité, forcément elles ont introduit un biais." Speaker 5 (Senegal's Digital Director) warned about falsification of history: "d'ici quelques temps, peut-être que si on tape surtout sur les informations non numérisées, qu'on se retrouve avec une falsification de l'histoire."

In Comment l'IA modifie l'accès à la connaissance?, Seydina Ndiaye noted: "tant que les courants de pensée les plus en avance sont dans des langues qui sont les plus dominantes... on gardera le même schéma continuellement."

5. Development Gaps, Infrastructure Deficits, and Political Economy

In Atelier 3: La souveraineté infrastructurelle est-elle primordiale?, Emmanuel highlighted the competing priorities African governments face: "When you start talking to him about all his projects, he sees, he budgets, he says, listen, my voters are waiting for health, hospitals, agriculture." Brain drain, low engineer-to-population ratios, and political short-termism were also cited as structural barriers.

In Atelier 1: Quels mécanismes pour empêcher l'IA de restreindre l'accès à la connaissance?, Mehdi Snene raised concerns about Africa's marginal position in AI: "l'Afrique n'a que 0,7% de flux sur l'IA génératif pour schématiser l'impact, c'est quoi l'impact économique d'exclure l'Afrique, rien du tout."

In Atelier 3: À quelle échelle réguler?, Roger Adam highlighted that African countries cannot effectively tax or regulate large AI actors: "des gens qui vont dans notre cyberespace...qui s'éclatent des chiffres d'affaires, que nos États puissent en bénéficier."

6. Cognitive and Educational Risks

In Comment l'IA modifie l'accès à la connaissance?, Yadem Benmessaoud raised concerns about AI causing cognitive decline among students: "Un décrochage cognitif qui se montre et qui est depuis longtemps de plus en plus important." In Atelier 4: Quelle régulation pour encourager l'innovation?, Yadem Saoudi noted "on a vu le niveau de connaissances s'effondrer de manière globale" when students used AI to get direct answers.

7. Proposals: Making AI and Digital Tech Serve Sustainable Development

Across sessions, a coherent set of proposals emerged:

Taken together, the sessions converged on a vision in which AI and digital technologies can only meaningfully support sustainable development if they are designed with equity, linguistic diversity, environmental sustainability, and local ownership at their core — and if the structural barriers of infrastructure deficits, talent gaps, and extractive economic models are actively addressed through both national strategies and international cooperation.

Across the various sessions, proposals and references to regional organizations such as the African Union (AU) in the context of AI and digital technology development were scattered and often critical rather than concrete, though several recurring themes emerged.

Underutilization of Existing Regional Structures

A recurring concern was that existing regional bodies exist but are not being leveraged effectively. In Atelier 2: Comment réguler, Patricia Amand pointedly questioned: "Smart Africa, c'est 40 pays aujourd'hui. Qu'est-ce qu'on en fait? On a des outils, on a l'Union africaine. Qu'est-ce qu'on en fait?" An audience member further highlighted the problem of fragmentation among overlapping frameworks, proposing consolidation into a single, clear directive rather than multiple competing structures.

Calls for a Common African Platform

In Des infrastructures durables et souveraines: une équation impossible?, Luc Missidimbazi called for a common African platform for infrastructure, while Annick Sakho referenced ECOWAS (CDAO) as a vehicle for regional data protection harmonization: "en Afrique de l'Ouest, la CDAO a des commissions informatiques et ce sera opportun que ce soit un message qui soit porté au niveau des différents gouvernements." Henri Monceau raised the question of whether a "ZLECAF or African Union initiative for AI" would be ideal, to which Sakho responded affirmatively but noted the complexity of existing laws.

Smart Africa as a Practical Vehicle

In Quel équilibre entre règlementation et innovation?, Lacina Koné proposed the Compute Embassy Law as a Smart Africa initiative to harmonize cross-border policies for computing infrastructure, and described Smart Africa's work on "Uberizing Computer Access" to democratize AI computing across African states. The AFNIC-Smart Africa MOU was also highlighted as a step toward strengthening DNS infrastructure sovereignty across Africa. Similarly, in Atelier 3: La souveraineté infrastructurelle est-elle primordiale?, Rodrigue Guiguet-Md referenced Smart Africa's role in parliamentary sensitization on internet governance and AI: "I had the opportunity to be in a session of parliamentarians who were sensitized on issues related to internet governance... a session that Smart Africa organized."

Skepticism About Regional Leverage

In Atelier 3: À quelle échelle réguler?, Speaker 7 (Roger Adam) expressed skepticism about the AU's ability to influence major tech actors: "même si on va à l'école africaine, on n'est pas suffisamment puissant pour qu'ils nous écoutent." Similarly, in Atelier 1: Pourquoi réguler?, Oumar Agné criticized AU working groups for copying regulatory models from elsewhere: "je ne sais pas si vous êtes dans le groupe de l'Union africaine, c'est mon tour, on copie les modèles aujourd'hui."

Leveraging Multilateral and Cross-Regional Networks

In Atelier 4: Territoires et IA, Zacharia Oulad proposed leveraging Agadir's position at the crossroads of multiple regional organizations — including the African Union, Arab League, Mediterranean Union, and Islamic Cooperation Organization — to advance AI cooperation. In Atelier 2: Infrastructures numériques et empreinte environnementale, Speaker 2 emphasized participation in ITU standard-setting processes: "il y a beaucoup de choses, on travaille beaucoup avec l'UIT sur des normes des trajectoires numériques... c'est très important que les gouvernements que les parties prenantes participent aussi à l'UIT et aussi les pays africains."

Linguistic and Cultural Cooperation as a Starting Point

In Atelier 2: La diversité linguistique dans les données d'entraînement de l'IA, Speaker 2 suggested that African languages sharing cultural commonalities could pool corpora as a form of regional cooperation: "l'union de quelques langues africaines qui ont une culture en commun...n'est pas une approche qui, économiquement, pourrait faciliter la tâche?"

Overall Assessment

Across all sessions, no single, concrete, and actionable proposal was formally adopted regarding the AU's specific institutional role in AI and digital technology development. The discussions revealed a tension between the recognition of existing regional frameworks and frustration at their underuse, fragmentation, and limited leverage. The most actionable proposals centered on Smart Africa as a practical implementation vehicle, ECOWAS for data protection harmonization, and calls for consolidating overlapping African digital governance structures into a more unified and effective regional voice.

This question was not discussed in any of the conference sessions reviewed. No tensions between the conference's direction and any country's stated position were identified or flagged across the sessions on sustainable and sovereign infrastructures, access to knowledge and AI, linguistic diversity, or regulatory frameworks.

Follow-up Mechanisms Established Across Sessions

Across the full range of sessions, formal follow-up mechanisms were notably absent. The overwhelming pattern was one of discussion and recommendation without binding commitments, working groups, review conferences, or reporting deadlines emerging from the meetings themselves. However, several concrete announcements and references to existing or upcoming processes were recorded.

Newly Announced Follow-up Mechanisms

Upcoming Events Referenced as Follow-up Venues

Pre-existing Mechanisms Referenced

Informal or Aspirational Proposals Not Formally Adopted

Overall Assessment

The forum generated significant discussion and surfaced important recommendations, but fell short of establishing formal institutional follow-up. The most concrete outcomes were the AFNIC–Smart Africa MOU, the announcement of the International Network of Capacity-Building Centers, and the OIF special envoy on discoverability. The forum's primary structural role was as a preparatory and deliberative space feeding into larger international processes — most immediately the UN Global AI Governance Dialogue — rather than as a decision-making body establishing its own accountability mechanisms. Real-time AI transcription and synthesis by the Diplo Foundation, made available to participants via QR code the same evening, served as the principal documentation mechanism across sessions.

Across the sessions, discussions touched on national digital-economy and connectivity strategies to varying degrees, ranging from direct references to infrastructure investment and broadband expansion to more indirect observations about the economic conditions shaping digital development.

The most substantive connectivity-related discussions occurred in infrastructure-focused sessions. In Atelier 2: Infrastructures numériques et empreinte environnementale, ARCEP's representative suggested that energy efficiency gains in telecoms could directly enable broader coverage: "si on pousse les opérateurs télécoms à être plus vertueux, à utiliser moins d'électricité, moins d'énergie, peut-être qu'on va pouvoir mettre plus d'équipement ou aller plus loin dans la couverture des populations." Satellite connectivity was also raised as an alternative to cable infrastructure.

In Atelier 3: La souveraineté infrastructurelle est-elle primordiale ?, the underutilization of existing submarine cable bandwidth was highlighted as a structural challenge, with one speaker noting: "There are countries that tell you, we want a third cable, a fourth cable, whereas when you take the first cable the country has, on international bandwidth, consumption is at 10%." High internet costs in Africa despite cable investments were also flagged.

In Atelier 1: Des infrastructures durables au-delà des évolutions technologiques, the erosion of telecom operators' investment capacity was raised as a systemic concern: "On investit à hauteur de ce qu'on gagne... les opérateurs tel qu'on ont de moins en moins de valeur, donc ils ont de moins en moins de capacité à investir dans l'infrastructure de connaissibilité." The localization of DNS infrastructure in Africa was also cited as a means to improve efficiency.

The plenary session Des infrastructures durables et souveraines: une équation impossible ? offered the broadest treatment of connectivity in a development context. Luc Missidimbazi recalled that "il y a 20 ans, la première problématique c'était la connectivité, connecter l'Afrique," while noting that new infrastructure challenges have since emerged. He also highlighted an energy-telecom infrastructure bundling initiative in Congo as a concrete investment strategy. Annick Sakho discussed Oracle's data center expansion across Africa, and Ambassador Jurgensen referenced France's 93 billion euros in foreign investment commitments, "dont une majorité dans le numérique et les infrastructures stratégiques."

In Atelier 4: Territoires et IA: comment les villes francophones passent à l'acte, Sébastien Bachelet described the Nièvre department's post-fiber strategy, investing 27 million euros in IoT networks and digital service bundles, using connectivity as "un levier d'attractivité, de compétitivité, d'amélioration du quotidien." Zacharia Oulad described how Agadir doubled its municipal revenues over four years partly through AI-assisted fiscal resource identification.

In Quel équilibre entre règlementation et innovation ?, Lacina Koné cautioned against reducing digital transformation to rural connectivity alone: "Don't make us believe that the electrification of rural areas is the issue," arguing instead for citizen-centered digital transformation. He also highlighted Africa's energy assets as a basis for hosting AI infrastructure contributing to the digital economy.

In Atelier 4: Quelle régulation pour encourager l'innovation ?, the Senegal 'initiative numérique présidentielle' was cited as a positive national strategy for startup development, and Laurent Ferrali warned against over-taxing the digital sector: "si on impose un marché en croissance, il est beaucoup moins en croissance."

In Atelier 2: Comment réguler, Patricia Amand emphasized the importance of not hindering digital development in African contexts: "sans freiner le développement économique et en particulier de nos environnements africains qui sont encore en demande de développement économique, de développement numérique et de transformation numérique."

Overall, while connectivity and digital economy themes were present across multiple sessions, they were most often addressed as contextual backdrops or illustrative examples rather than through detailed national strategies, broadband access targets, or specific ICT-GDP contribution metrics. Sessions focused on content, linguistic diversity, and AI knowledge access did not engage with these issues.

Across the sessions reviewed, no single meeting produced a comprehensive, formalized set of proposals explicitly linking digital technology, human rights, and international peace and security as an integrated agenda. However, a number of significant concerns and partial proposals emerged across several sessions that, taken together, sketch the contours of this nexus within Francophone digital governance debates.

The most direct engagement with peace and security came in Quel équilibre entre règlementation et innovation?, where Omar Zniber noted that francophone ambassadors had organized a dedicated debate on "the role of artificial intelligence in disarmament and arms control," covering both defense and humanitarian dimensions. Zniber also referenced his presidency of the Human Rights Council, where he organized debates on AI, urging participants to "never forget the human dimension when dealing with this subject of artificial intelligence." Henri Verdier warned that "innovation can also serve dystopian projects and the question for now is whether these digital technologies will allow us to build a utopian society," and raised concerns about trillion-dollar companies acting as geopolitical actors.

Geopolitical dependency on digital infrastructure was identified as a structural threat across multiple sessions. In Atelier 4: Territoires et IA, Céline Colucci warned of the risk of reliance on US cloud providers, describing "une difficulté géopolitique énorme...qui a la capacité de couper les services du jour au lendemain, parce que soit ils considèrent que c'est trop stratégique...soit parce que l'action ou la réaction ne convient pas, et c'est une forme de sanction." In Atelier 3: La souveraineté infrastructurelle, Lucien Castex illustrated this dependency concretely, noting that "The Claude 5 and Mythos 5 models were deactivated internationally following a directive from the American government, which clearly illustrates that digital infrastructures are a game of dependency." Madame Lydie Oumanga raised the human rights dimension of internet shutdowns, noting: "If you have known countries where they cut the Internet when there are elections, you are consuming the Internet, we do nothing."

The manipulation of knowledge through AI was flagged as a security and rights concern in Comment l'IA modifie l'accès à la connaissance?, where Stéphane-Eloïse Gras cited the example of Grok being pre-prompted to assert "il y a bien un génocide blanc en cours en Afrique du Sud" regardless of the question asked, framing this as a major governance and security concern. She also noted that "pas plus tard qu'il y a quelques semaines, l'Europe et toutes les autres régions du monde se sont vues interdire l'accès à un modèle dit fondationnel," framing access restrictions as a cognitive sovereignty emergency. Xavier Merlin proposed reaffirming open internet principles to "garantir la liberté d'accès à la connaissance et... la liberté d'entreprendre."

In Atelier 1: Des infrastructures durables, Speaker 1 connected internet outages to state stability, warning that infrastructure failure "lie fortement à la capacité de l'État à rester au pouvoir à l'administration en place, à rester au pouvoir et ça peut amener à des émeutes," while Speaker 2 referenced geopolitical incidents such as cable cuts in the Baltic Sea. In Atelier 2: Infrastructures numériques, Speaker 3 raised a social justice concern about communities being deprived of electricity in favor of data centers: "on choisit des serveurs plutôt que des hommes."

Human rights were invoked as a regulatory foundation in Atelier 3: À quelle échelle réguler?, where Speaker 1 stated that regulation must be "respectueuse des droits fondamentaux, de la diversité culturelle et des principes démocratiques," and Speaker 2 grounded the Council of Europe Convention in "le respect des droits fondamentaux...la démocratie, de l'État de droit, et du respect des droits de l'homme." In Atelier 1: Quels mécanismes pour empêcher l'IA de restreindre l'accès à la connaissance?, Jonas Erin framed the Council of Europe's mandate around "la démocratie, l'état de droit et les droits de l'homme."

In Atelier 4: Territoires et IA, Zacharia Oulad offered a constructive example of how city diplomacy can preserve human ties when national governments freeze relations, noting that "c'est grâce au fait que la diplomatie des villes a continué d'exister qu'au moment où les gens d'en haut se sont dit, on va redevenir copains-copains, les gens d'en bas se sont dit, nous, on n'a jamais arrêté," suggesting that digital cooperation at the subnational level can serve as a peace-sustaining mechanism.

Overall, while no session produced a consolidated set of proposals formally addressing the digital technology–human rights–peace and security nexus, the discussions collectively surfaced several recurring themes: the geopolitical weaponization of infrastructure access, AI-enabled ideological manipulation, internet shutdowns as tools of political repression, the need to anchor digital regulation in fundamental rights frameworks, and the importance of multilateral and subnational cooperation as buffers against unilateral digital coercion. The most concrete proposal to emerge was the reaffirmation of open internet principles as a guarantee of cognitive freedom, alongside the call — most explicitly voiced by Omar Zniber — to keep the human dimension central to all AI governance deliberations.

Unified Summary: Addressing the Information-Society Orientation Framework's Objective of Building an Inclusive, Rights-Respecting Information Society Organized Around Liberty, Security, and Solidarity

Across the sessions reviewed, the themes of inclusivity, rights, liberty, security, and solidarity — central to the information-society orientation framework — were addressed with varying degrees of explicitness. While no session explicitly cited the framework by name in a sustained or systematic way, its core objectives surfaced repeatedly across infrastructure, governance, regulation, linguistic diversity, and AI policy discussions. The most direct engagement came from the plenary session on balancing regulation and innovation, while partial engagement was evident across nearly all other sessions.

Liberty

The principle of liberty was most explicitly invoked in the session Comment l'IA modifie l'accès à la connaissance?, where Xavier Merlin described open internet regulation as a "règlement de liberté garantissant la capacité d'accéder librement au contenu de son choix" and called for reaffirming these principles in the age of generative AI. In Atelier 1: Des infrastructures durables au-delà des évolutions technologiques, Speaker 1 warned against dependence on proprietary platforms, stating: "autrement on dépendra que des plateformes" and called for co-construction of the internet through multistakeholder models: "on a encore l'opportunité de construire l'internet de demain, de le co-construire avec des modèles multi-acteurs." In Quel équilibre entre règlementation et innovation?, Janis Karklins explicitly referenced the WSIS legacy, stating that the multi-stakeholder approach "has allowed the Internet to evolve into a global, open, stable, secure, resilient and interoperable ecosystem serving billions of people."

Inclusivity

Inclusivity was the most consistently addressed dimension across sessions. In Atelier 2: Infrastructures numériques et empreinte environnementale, Speaker 3 raised concerns about the reversal of democratization trends: "quand l'informatique est arrivée dans les années 80... c'était de démocratiser... Et là, avec l'arrivée de l'IA, c'est l'inverse qui est arrivé. C'est-à-dire qu'en fait, en gros, le matériel va coûter de plus en plus cher, et ça va devenir inaccessible." In Atelier 4: Territoires et IA, the speaker from Côte d'Ivoire raised the risk of new digital divides, and Céline Colucci described citizen participation initiatives. In Atelier 3: La souveraineté infrastructurelle est-elle primordiale?, Madame Oumanga highlighted grassroots digital literacy workshops, and Speaker 4 emphasized digital identity as foundational: "If you want to succeed in digital inclusion, financial inclusion, everything we say, it's identity." In Atelier 3: Quels outils pour développer les contenus multilingues?, Speaker 1 illustrated the stakes of linguistic exclusion: "la constitution, elle est écrite en français. 60% de la population ne lit pas français. Et ils doivent voter la constitution. Donc, comment? On s'assure au moins qu'ils comprennent ce qu'ils votent." In Des infrastructures durables et souveraines: une équation impossible?, Dr. Xianhong Hu stated: "Il ne peut y avoir de politique de l'EA inclusive sans inclusion linguistique." In Comment l'IA modifie l'accès à la connaissance?, Seydina Ndiaye noted the systemic nature of exclusion of minority languages and cultures from AI systems, and the launch of an international network of capacity-building centers was presented as a concrete solidarity measure. In Quel équilibre entre règlementation et innovation?, Henri Verdier called for AI literacy to be treated as important as alphabetization to ensure human autonomy and dignity, and Lacina Koné emphasized citizen-centered transformation.

Security

Security was addressed primarily in terms of data protection, cybersecurity, and cognitive sovereignty. In Atelier 4: Territoires et IA, Céline Colucci discussed GDPR compliance, cybersecurity, and data protection as baseline requirements for a rights-respecting digital society. In Atelier 3: À quelle échelle réguler?, Speaker 2 described Monaco's sovereign cloud and AI platform as security measures, and both experts emphasized the balance between security and avoiding over-regulation that would stifle innovation. In Comment l'IA modifie l'accès à la connaissance?, Stéphane-Eloïse Gras raised concerns about cognitive sovereignty and the risks of AI systems defining "l'espace du croyable, l'espace du crédible et l'espace de l'interprétation." In Atelier 4: Quelle régulation pour encourager l'innovation?, data protection for minors and health data was discussed extensively, and Dawit Bekele emphasized that the Internet's fundamental properties of openness and global interconnection must be preserved.

Rights-Respecting Governance

Rights-based framing was present in several sessions, though rarely invoked explicitly. In Atelier 3: À quelle échelle réguler?, Speaker 1 emphasized "droits fondamentaux" and Speaker 2 grounded the Council of Europe Convention in human rights. In Atelier 1: Quels mécanismes pour empêcher l'IA de restreindre l'accès à la connaissance?, Jonas Erin called for empowering individuals as citizens: "il faut aussi penser une forme de gouvernance de l'IA... encapaciter l'individu en tant que citoyen pour qu'il soit en mesure de lui-même de participer au garde-fou collectif." Mehdi Snene's concerns about data exploitation without consent or benefit-sharing touched on solidarity and rights dimensions. In Atelier 4: La numérisation des contenus inexploités, Speaker 1 stated the goal of "contribuer à l'édification d'un écosystème numérique qui soit ouvert, qui soit disponible pour terrain" and emphasized trust-building: "il n'y a pas d'IA possible à mon sens sans confiance." In Atelier 1: Pourquoi réguler?, Jérôme Duberry framed regulation as a means to "mettre l'humain au centre" and to decide "qui est-ce qu'on souhaite protéger, quel type de technologie on veut dans la société."

Solidarity

Solidarity was addressed most explicitly in Atelier 3: À quelle échelle réguler?, where Speaker 1 invoked "la solidarité entre les peuples" as a Francophone value. In Atelier 4: Territoires et IA, the AIMF's role in connecting francophone cities was highlighted as a solidarity mechanism, and Zacharia Oulad's concept of valeur publique was central to his presentation. In Atelier 2: Comment réguler, Patricia Amand raised the question of the Francophone voice in global AI governance: "la voix francophone, est-ce qu'elle compte?" In Atelier 3: La souveraineté infrastructurelle, Madame Oumanga reminded participants: "We forget that we work for the peoples and that today if we lose sight of the fact that we are just passing through and that we work for these populations, we will have a big problem." In Atelier 4: Quelle régulation pour encourager l'innovation?, the DRC regulator stressed working for citizens: "est-ce que ce qu'on va faire là le Congolais lambda est-ce que ça lui parle est-ce que ça a un impact parce qu'on travaille pour eux."

Structural Gaps and Governance Challenges

Several sessions highlighted structural obstacles to realizing the information-society framework's objectives. In Atelier 2: Infrastructures numériques et empreinte environnementale, Henri Monceau raised the paradox of facing global challenges without adequate global governance structures. In Atelier 2: Comment réguler, an audience member from Cameroon raised the challenge of regulating harmful content while lacking control over infrastructure. In Des infrastructures durables et souveraines: une équation impossible?, Aurélie Simard called for "des citoyens qui ont un usage éclairé, des entreprises qui développent des pratiques responsables... et des politiques publiques qui soutiennent cet effort." Pierre Bonis emphasized that the Internet's success was built on open standards and multi-stakeholder consensus.

Conclusion

The information-society orientation framework's objectives were substantively — if implicitly — engaged across the majority of sessions. Inclusivity was the most consistently addressed dimension, particularly through the lens of linguistic diversity, digital access, and citizen participation. Liberty was addressed through open internet principles and multi-stakeholder governance. Security was discussed primarily in terms of data protection and cognitive sovereignty. Solidarity emerged through Francophone cooperation mechanisms, capacity-building initiatives, and citizen-centered governance rhetoric. The most explicit engagement with the WSIS framework occurred in Quel équilibre entre règlementation et innovation?, where Janis Karklins directly referenced the WSIS+20 review and its reaffirmation of multi-stakeholder governance. The primary gap across sessions was the absence of explicit rights-based framing that would connect these discussions to the framework's stated objectives in a systematic and actionable way.

Data Localization, Data Flows, and Cross-Border Data Transfers: A Unified Summary

Across the sessions, data localization, data flows, and cross-border data transfers emerged as recurring and substantive concerns, particularly in discussions about digital sovereignty, infrastructure vulnerability, and AI governance. The debates ranged from concrete infrastructure failures caused by offshore data storage to conceptual proposals for "data embassies" and regional harmonization frameworks.

Infrastructure Vulnerability and the Case for Localization

The most vivid illustrations of data localization risks came from Atelier 3: La souveraineté infrastructurelle est-elle primordiale?, where Emmanuel described two striking cases of African states' dependence on foreign data infrastructure. The first concerned a cyberattack on Senegal's tax authority, where data was stored in Kuala Lumpur: "They had stored this data in Kuala Lumpur, that is 17 or 20 hours of flight from Dakar. And so, it's not even on national territory that these infrastructures were attacked." The second concerned West African public services routed through Portugal: "Most of the data and exchange routes are in Portugal. And so, each time there is a power cut in a key state in Portugal, all public services, all banking services, etc., are blocked."

Rather than advocating for full data localization, Speaker 1 in the same session proposed a more nuanced approach focused on resilience: "Ensuring that the BCP and DRP are localized at home, meaning you have the capacity to restart the system from home — which is slightly different from wanting to localize everything at home."

The Data Embassy Concept

The "data embassy" model was proposed in multiple sessions as a pragmatic middle ground between full localization and unregulated offshore storage. In Atelier 3: La souveraineté infrastructurelle est-elle primordiale?, Rodrigue described it as follows: "The question of the data embassy so that states can at least use this computing power, this storage capacity that already exists in certain countries or with certain industries, through certain regulations and laws that could be harmonized to guarantee that the data they store there is protected." However, he cautioned that many African governments remain unfamiliar with the concept: "Coming from the continent and especially working with these governments, I believe they do not know. This data embassy story, we have been talking about it for years."

In Des infrastructures durables et souveraines: une équation impossible?, Annick Sakho elaborated on the same concept as a legal compliance mechanism: "avoir une ambassade dans son data center qui me permettrait de respecter ma loi qui veut que les données personnelles soient dans un environnement qui soit à l'intérieur des frontières du pays."

Bilateral Agreements and Data Classification

Also in Des infrastructures durables et souveraines: une équation impossible?, Annick Sakho stressed that cross-border data flows require bilateral agreements: "ça nécessite des accords bilatéraux pour permettre d'avoir justement ce flux de données transfrontalier qui va permettre à un pays non seulement d'utiliser la puissance de calcul, mais aussi d'accepter que ces données résident dans un data center hors de ses frontières sur la base d'accords bilatéraux." She also proposed data classification as a prerequisite for any localization or transfer policy: "Quelles sont les données qui sont considérées comme des données critiques? Quelles sont les données auxquelles on a accès... Quels sont les données qui peuvent être exportées, qui peuvent être utilisées dans des environnements d'intelligence artificielle qui ne soient pas à l'intérieur des frontières." She also referenced ECOWAS data protection commissions as a vehicle for regional harmonization.

Regional Compute and Data Sharing Frameworks

In Quel équilibre entre règlementation et innovation?, Lacina Koné proposed a "Compute Embassy Law" to enable cross-border sharing of computing resources among African states, implying managed cross-border data flows within a regional framework. He also noted that Africa currently contributes only 2% of AI training data and called on African states to establish governance over their data: "It is the right time for our states to prepare the future by putting governance on this data."

Data Sovereignty in AI and Health Contexts

In Atelier 4: Quelle régulation pour encourager l'innovation?, Laurent Ferrali raised concerns about foreign actors accessing sensitive local data, noting the risk of "le pillage des données de santé par les opérateurs étrangers." Proposals in this session included hosting data in national infrastructure, creating tiered access systems, decentralizing data responsibility to local institutions, and using on-device storage so users control when data is shared.

In Atelier 1: Quels mécanismes pour empêcher l'IA de restreindre l'accès à la connaissance?, Mehdi Snene raised the issue of African health data being exported for AI training: "on a vu aussi des projets de vente de données médicales qui partent d'Afrique pour alimenter des producteurs d'IA on l'a vu, ça a été invalidé constitutionnellement mais ça existe." He also noted the absence of accountability mechanisms: "il n'y a ni le mécanisme pour avoir des comptes à rendre sur l'exploitation de nos données pour produire leur IA."

AI Sovereignty and Local Data Retention

In Atelier 2: La diversité linguistique dans les données d'entraînement de l'IA, Speaker 2 argued that externalizing AI means losing control over data: "si on externalise, il abandonne les données" and stressed the need to retain data locally: "être propriétaire des données qui vont permettre de les utiliser."

In Atelier 4: Territoires et IA, Céline Colucci highlighted the importance of data control for effective AI: "on a un gros sujet qui est l'IA souveraine...on a besoin de données fiables et maîtrisées pour avoir des résultats efficaces."

DNS Localization

In Atelier 1: Des infrastructures durables au-delà des évolutions technologiques, Speaker 1 noted a more technical dimension of data localization — the localization of DNS queries — describing ICANN's efforts: "l'ICANN ce qu'elle a fait dans ce cadre là c'est qu'elle a installé deux structures importantes en Afrique qui ont permis de localiser des requêtes DNS qui ne sortaient plus du territoire."

Conclusion

Across sessions, a consistent set of concerns emerged: African and francophone states are disproportionately dependent on foreign data infrastructure, leaving them vulnerable to outages, cyberattacks, and data exploitation. The "data embassy" and "compute embassy" models were the most concrete proposals for managing cross-border data flows while preserving sovereignty. Bilateral agreements, data classification frameworks, and regional harmonization through bodies like ECOWAS were identified as necessary policy tools. However, several speakers noted that awareness and implementation of these frameworks remain limited, particularly among African governments.