Quel équilibre entre règlementation et innovation ?

The second Francophone Forum on Digital and Artificial Intelligence Governance opens with the observation of a clear rise in the initiative’s momentum and the assertion of a more structured political ambition. Henri Monceau emphasizes that in-person participation has doubled and that more than 900 registrations were recorded, three times more than the previous year. He sees this not only as interest in AI, but also as the expression of a broader need for a Francophone space for dialogue, reflection, and proposals on digital governance. He presents this forum as a place where states, parliamentarians, local authorities, regulators, academics, standardization bodies, data protection authorities, technical actors, and civil society can jointly develop common positions. He also makes clear that the initiative is not intended to speak on behalf of the world, but to contribute to a more inclusive, legitimate, and genuinely multilingual global governance of AI.

This opening is situated within a broader international sequence. Monceau notes that the forum is being held on the eve of the first United Nations global dialogue on artificial intelligence, created in the wake of the Global Digital Compact, and shortly after the publication of the preliminary report of the independent scientific panel on AI. From this context, he draws a twofold diagnosis: never has a technology spread so quickly, but never have the capabilities that make it possible been so concentrated. In this regard, he cites the global distribution of computing power dedicated to AI, 75% of which is located in the United States and 15% in China, leaving only a residual share for the rest of the world. In his view, this concentration is at once technological, economic, cultural, linguistic, and geopolitical, hence the particular responsibility of the Francophonie in defending an innovation that does not result in uniformity. Monceau then structures the day around three major themes: the balance between innovation and regulation, the effects of AI on knowledge, languages, content, and access to knowledge, and then the question of infrastructure, its sustainability, and digital sovereignty. He specifies that no definitive answer will be found in a single day, but that each of these questions calls for collective responses. The opening tone is both solemn and mobilizing: at this stage, the issue is no longer whether AI will transform societies, but who will actually take part in that transformation. Doreen Bogdan-Martin, Secretary-General of the ITU, extends this framing by explicitly placing the forum within the immediate multilateral agenda. She recalls having followed with interest the conclusions of the first forum, focused on innovation, cooperation, inclusion, youth, and culture, and sees this second edition as an opportunity to inform the following week’s discussions on global AI governance. She takes up the major themes announced - the balance between regulation and innovation, access to knowledge, sustainable and sovereign infrastructures - and invites participants to join the first United Nations global dialogue on AI governance, co-organized by the ITU, UNESCO, and the UN. She also presents the ITU’s role as a technical and standard-setting support organization, acting through capacity-building, the development of tools, standards, and technical publications, as well as through the AI for Good initiative, conceived as a platform linking the private sector, the United Nations, NGOs, administrations, start-ups, and researchers. Finally, she places the WSIS Forum within the same dynamic, emphasizing that AI must be discussed there not only as a technological innovation, but also from a development perspective. Henri Monceau then explains that Amandeep Singh Gill, who had initially been expected in person, was unable to arrive on time due to a transport problem and recorded a video message from Rome that was played during the session. In this address, the United Nations Secretary-General’s Envoy on Technology describes AI as the century’s “cognitive infrastructure,” capable of translating, summarizing, advising, and deciding, but he raises the crucial question: in what language does it think, and with what memory of the world? He states that the major models are currently trained on only a fraction of human experience, leaving thousands of languages, as well as entire legal, philosophical, medical, and poetic traditions, outside the frame or at the margins. From this, he concludes that the risk is not merely technical, but civilizational: that of an intelligence that would mistake one part of humanity for all of humanity. He draws from this a demanding conception of universality: a truly global AI is not measured first by its computing power, but by its ability to embrace the plurality of ways of speaking, knowing, and being. Cultural diversity is therefore not, in his view, a moral add-on, but a condition of the universality of digital governance. He praises the Francophonie for having understood that language itself constitutes an infrastructure, and that defending a language means defending a way of understanding the world. In the immediate wake of this message, Henri Monceau relays a compelling call to make the Francophone space a “living laboratory” based on open corpora, shared data, multilingual models, and talent trained across five continents. He argues that diversity should not be seen as an obstacle to innovation, but as its accelerator, and suggests that what the Francophonie succeeds in doing could then be taken up by other linguistic communities.

Omar Zniber, President of the Group of Francophone Ambassadors in Geneva, in turn places the forum back within the context of international Geneva. He points out that the meeting coincides with a week marked by several major events - the global dialogue on AI governance, the World Summit on the Information Society Forum, and the AI for Good Summit - which situates the forum within a particularly dense diplomatic week in Geneva and reinforces its role as a platform for exchanges, experience-sharing, and concrete recommendations. He then recalls that digital technology and artificial intelligence are profoundly reshaping economies, administrations, education, health, and production systems, while also being able to accelerate the Sustainable Development Goals, improve public services, and boost competitiveness. He nevertheless stresses that these promises come with major challenges: protection of personal data, reduction of the digital divide, ethics, and digital sovereignty. He also mentions the uses of AI in disarmament, security, and humanitarian issues, to underscore that the human dimension must remain central. In his view, the Francophone space has strengths that can help build a safer, more sovereign, and more inclusive digital ecosystem through the sharing of knowledge and good practices. He finally insists on the importance of using AI to entrench linguistic diversity in international bodies and to enrich the capacity to “understand one another together.” Janis Karklins then adds a conceptual complement by returning to Internet governance. He defends the multistakeholder approach as one of the main reasons for the success of the Internet since WSIS, recalling that no single actor alone possesses all the necessary expertise, authority, or responsibility. Governments, the private sector, civil society, academia, the technical community, and international organizations must therefore continue to contribute complementary perspectives, because decisions on the Internet are more robust when they are grounded in the experience of those who develop it, use it, and depend on it. Karklins also recalls that ICANN coordinates the Internet’s system of unique identifiers according to a multistakeholder approach, which contributes to the security, stability, and interoperability of the global domain name system. This point establishes continuity between the historical governance of the Internet and the new debates on AI: the multistakeholder model, reaffirmed in the context of the WSIS+20 review, remains essential as the challenges become more complex. The first roundtable addresses precisely the balance between regulation and innovation, with Henri Verdier, Léna Dargham, Lacina Konaté, and Anne-Marie Jean. From the outset, Henri Monceau formulates the classic tension: some argue that regulation hinders innovation, while others believe that in the absence of rules there can be neither trust, nor sustainable investment, nor responsible innovation. The discussion will, however, quickly dismantle this overly simple opposition. Henri Verdier opens with a deliberately provocative and vivid answer: he recalls that French gastronomy is one of the most regulated sectors in the world, yet that has not prevented it from being excellent. He adds that bad regulation can obviously hinder innovation, but that there are also many examples where regulation has supported it. His most developed illustration is aviation: Europe, he says, was able to enter a market dominated by Boeing by relying on higher safety standards, which were turned into a competitive advantage. Above all, he concludes that the real question is not only whether regulation hampers innovation, but recognizing that not every innovation is synonymous with progress. Without political intervention, democracy, and collective decision-making, digital innovation can also serve dystopian projects. Léna Dargham then shifts the center of gravity of the debate toward trust and standardization. She believes that the essential point is not to oppose regulation and innovation, but to find the balance that makes it possible to create trust. To do so, it must be possible to demonstrate the reliability, interoperability, and connection of solutions to international markets. She describes standardization as an “invisible pillar” that allows innovations precisely to be recognized as reliable and interoperable, and to create a common language among innovators, investors, regulators, and markets. Her remarks broaden the notion of governance beyond the law toward technical and commercial mechanisms that are often less visible, but decisive. Lacina Konaté moves in the same general direction, while reframing the priorities from an African perspective. In his view, the debate is not about choosing between regulation and innovation, but about creating the conditions of trust that allow innovation to thrive while protecting citizens. He states that no lasting technological transformation has ever developed without trust and that, for Francophone countries, the priority should not be to pile up ever more rules, but to build smart, proportionate regulatory frameworks that are conducive to innovation. He also insists on an idea he will develop later: AI governance does not begin with the algorithm, but with infrastructure, data, skills, and trust. Anne-Marie Jean, for her part, brings in the example of Quebec’s cultural sector to illustrate that a highly regulated environment does not prevent innovation. She recalls that culture in Quebec is structured by copyright laws, associations, unions, and support mechanisms, without this preventing Quebec creators from gaining international visibility. Her argument is that this regulatory environment protects fairness and creates the conditions in which artists and creators can innovate sustainably. The discussion then shifts to the question of the real priority for African countries: access, regulation, infrastructure, or capacity. Lacina Konaté answers bluntly that, for the population, the priority may be access to the Internet, but that, for governments, the priority must be regulation. He places this position within a historical and demographic reading: Africa would not truly have participated in any previous technological revolution and is now facing population growth on an unprecedented scale, which requires different responses. But his notion of regulation is itself distinctive. Konaté believes it must be proactive and citizen-centered, whereas the last thirty years would instead have produced regulations centered on technologies. Here he develops one of the most striking distinctions of the session: digital transformation is not simple digitization. Putting forms or paperwork online does not transform the state if services remain designed in a fragmented way and if the citizen remains the one who personally carries their data from one administration to another. True transformation instead requires organizing around the citizen and making coherent use of data. Prompted again on the risk that those who control infrastructure and content will also dominate the making of the rules, Konaté broadens the debate further to the geoeconomics of AI. He argues that Africa currently accounts for only a very small share of the data mobilized in AI infrastructures, a situation he contrasts with much stronger concentration in other regions of the world. He concludes that African states must urgently govern their data, because AI feeds on this resource and the future is already being decided there. He goes further by arguing that Africa also has underestimated material assets, particularly in renewable energy. He states that the continent has around 3,000 hours of sunshine per year and suggests that this resource could support the establishment of computing centers and AI servers. In this perspective, he reverses the usual narrative of delay: in his view, Africa is not a peripheral “Global South,” but a potential “new center,” provided that data, energy, and computational infrastructure are understood as strategic issues in their own right. Digital sovereignty then emerges as a structuring theme, though with different emphases depending on the speakers. When Monceau asks Léna Dargham about the role of standards in this sovereignty - which he prefers to call strategic autonomy - she replies that the standard is “the language of the market” and the technical foundation of regulation. Those who draft standards therefore, in practice, write the rules of the market. Consequently, the low participation of Francophone countries, especially developing countries, in international standard-setting work constitutes, in her view, a major strategic handicap. Dargham identifies several causes for this under-participation, but particularly stresses the lack of access to standardization documents in French. Many Francophone countries, especially African ones, are not truly bilingual, which prevents their experts from participating usefully in consultations, votes, and drafting. She recalls that at the international level, one country counts for one vote, so not participating amounts to ceding the decision to others. She then presents a concrete response: the Réseau Normalisation et Francophonie is developing a machine-translation platform for draft standards, notably from ISO and IEC, in order to enable its members to review, discuss, and actively contribute to the texts. The linguistic and cultural dimension of AI is then explored further through the Quebec example presented by Anne-Marie Jean. She begins by recalling that the defense and promotion of French lie at the heart of Quebec’s political identity, but explains that the shift of cultural practices toward digital platforms and connected devices has made a new issue central: no longer only the production of content, but its discoverability. In her view, major platforms now play a decisive role in what citizens do or do not see, and one can observe a very worrying decline in certain cultural practices in French, especially among young people. She then presents Quebec’s Bill 109 as a strong policy response. Anne-Marie Jean states that this law was passed by the National Assembly of Quebec in December 2025. It is not intended to regulate content directly, but rather the conditions of its discoverability. The goal is for original French-language cultural content to be found, suggested, and chosen by citizens. Concretely, it imposes requirements on digital platforms and manufacturers of connected devices, with the aim that national platforms and original French-language cultural content be more easily visible and accessible. For Jean, the issue is to prevent digital transformation from taking place at the cost of cultural erasure. She then explicitly links this policy to AI. Since AI feeds on data, it is necessary to ensure that Francophone data are present, structured, and usable within systems so that content can be found and promoted. She also mentions a project led by Bibliothèque et Archives nationales du Québec to build, in compliance with copyright, a database capable of feeding AIs that respect diversity, promote Francophone and Indigenous content, and strengthen sovereignty as well as data traceability. Henri Verdier then takes up the linguistic issue again at a more philosophical, technical, and geopolitical level. Responding to a question about the change in scale or nature of regulation, he clearly states that it is a change in nature. He recalls that 92% of training data are said to be in English and argues that the problem lies not only in reduced access to non-English knowledge. More profoundly, models trained in one language also internalize representations of the world specific to that language. Using the literary example of “solitude” and “soledad,” he shows that apparently equivalent words in fact open onto different imaginaries; consequently, an AI educated mainly in English not only understands certain forms of knowledge less well, it also “does not hear the same question.” Verdier reinforces this point by linking it to different legal and political conceptions. He refers, for example, to the constitutional protection of freedom of expression granted to legal persons in the United States, to suggest that models trained in that linguistic and institutional universe implicitly incorporate visions of law different from those prevailing in French, where the Constitution protects humans first. His argument is therefore that the language of training structures mental categories, hierarchies of values, and responses. This gives the question of linguistic diversity a much broader significance than that of a simple translation policy. The same speaker then highlights two other major transformations linked to AI. Henri Verdier emphasizes that, unlike the Internet revolution centered on relationships between people, AI penetrates the sphere of intimacy, to the point of transforming personal practices such as writing love letters or breakup letters. He also notes that the Internet was born in an academic world structured by open standards, digital commons, transparency, and cooperation, whereas the AI revolution is emerging at the heart of private companies worth a trillion dollars. Consequently, the contemporary challenge is to “impose the commons” on something that was not born as a common good. This difference in origin explains, in his view, the intensification of power dynamics and geopolitical tensions around AI. From this perspective, Verdier formulates one of the strongest theses of the morning: regulation will not happen if creation does not happen as well. For him, good governance is not limited to legal regulation; it also requires open standards, public infrastructure, voluntary norms, codes of conduct, common resources, and real scientific and economic power. He explicitly links this regulatory capacity to the production of commons: open standards, public infrastructure, shared resources, and foundation models designed as commons or global public goods. Writing good texts without having infrastructure, open databases, computing centers, and innovation capacities amounts to producing standards with no real grip on the world.

When Henri Monceau then asks the panelists which indicators would make it possible to know, in five years, whether the Francophonie has taken the right direction, the answers outline a shared but non-homogeneous strategic agenda. Anne-Marie Jean first hopes for strong Francophone convergence in international forums, particularly around UNESCO’s work on an additional protocol to the 2005 Convention on the Diversity of Cultural Expressions. She also mentions the recent announcement of a special envoy on the discoverability of Francophone cultural content, who should mobilize states around the infrastructure needed for their cultural and digital sovereignty. Lacina Konaté responds in terms of continental and national AI sovereignty in Africa. He hopes to see, within a five-year horizon, homegrown African computing capacities emerge, local servers, adapted sectoral uses - agriculture, critical infrastructure, fintech, administration - and cross-border mechanisms for shared access to computing power. In this context, he mentions Smart Africa’s work on a “Compute Embassy Law,” intended to democratize access to computing capacity and promote more sovereign AI at the African scale. Léna Dargham, for her part, proposes as indicators Francophone leadership in the development of international standards, not only through participation in drafting, but also by taking on responsibilities in the governance of technical committees. She also stresses the accessibility of standardization documents in French and their national adoption by companies, innovators, and private-sector actors, because certification and compliance with standards condition access to international markets. Henri Verdier finally formulates a triptych of success: making foundation models a common good of humanity, accessible without permission and genuinely multilingual; encouraging a great diversity of innovators so that solutions adapted to local contexts can emerge; and treating AI literacy as a priority as fundamental as basic literacy. He rejects the idea of a humanity reduced to passively consuming automated systems designed elsewhere, and calls for mass education in the reasoned and dignified use of AI. The sequence of exchanges with the audience confirms and complicates these main lines. A first speaker emphasizes the negative spirals that AI could produce, particularly around mining extraction, conflict, pollution, and the growing distance between decision-making models and lived reality. This intervention is a reminder that AI governance must also integrate the material and environmental externalities of technological infrastructures. Oumar Ndiaye then reframes the morning’s central issue in terms of legitimacy: who regulates, for whom, and for what purpose? He warns against the tendency to copy external frameworks, such as the GDPR, without sufficient reflection on specific contexts, and defends the idea of regulation oriented toward the common good. Other interventions refer to the scientific and infrastructural capacities existing in Europe, cooperation among Swiss, Finnish, and African researchers around computing for AI, the possible value of regulatory “sandboxes,” the transferability of Bill 109, the role of universities in adapting curricula, and the possibility of a genuine “win-win” model between high-income Francophone countries and middle- or low-income countries.

In response to these interventions, Henri Verdier refocuses the debate on the effectiveness of regulations in a more conflictual international context. He states that the multilateral order based on the rule of law is now weakened, which means that it is not enough to write good texts: it is also necessary to create the conditions that allow them to exist. In his view, those conditions depend on infrastructure, open databases, computing centers, standards, and industrial policies. He speaks of “gently imposing” the desired regulations, that is, not through coercion alone, but through the construction of resources and ecosystems that make possible an AI grounded in law and reason. Léna Dargham then takes up the question of regulatory harmonization and suggests that the OIF take stock of existing regulations in the Francophone space in order to explore the relevant levels of harmonization - national, regional, international - on the basis of good practices, rather than through simple copy-and-paste. Lacina Konaté, for his part, returns to African demographic and institutional issues. He emphasizes the urgency of providing health, banking, and administrative services to a rapidly growing population, which requires thinking in terms of services rather than physical institutions: he says, in substance, that populations need health services or banking services more than the simple institutional reproduction of the hospital or the bank. He links this approach to the issue of public digital infrastructure, particularly unique digital identity, payments, and data sharing. He also mentions electronic signatures and PKI, to show that the obstacles do not primarily stem from the absence of technology, but from an insufficient understanding of the issues. He finally states that the digital economy and transformation do not constitute one vertical sector among others, but rather a cross-cutting dimension that affects all public action and university education. Anne-Marie Jean complements these responses by recalling that, within the Francophonie, Ministers of Culture have already asked the Parliamentary Assembly of the Francophonie to develop a shared legislative corpus. She also indicates that Quebec and Côte d’Ivoire have jointly developed a framework identifying the legislative, regulatory, and support elements needed for a discoverability ecosystem. In her view, these tools should help each state diagnose its situation and plan its implementation, particularly with a view to the next meeting of Ministers of Culture in Yerevan. Henri Monceau concludes this part by responding to a question about the role of the OIF Resource Centre for digital and AI governance. He explains that this center, launched at Switzerland’s initiative during the Djerba Summit, is intended to support Francophone states, in particular through AI-based tools capable of querying regulatory and standardization texts in a precise way. He also announces a workshop in Alexandria on native language models, which directly extends the morning’s debates on the need for corpora, data, and models rooted in Francophone languages and contexts. The session finally closes with an act of concrete cooperation: the signing of a memorandum of understanding between AFNIC and Smart Africa. Pierre Bonis explains that the objective is to pool the networks of African national registries, in particular through a college bringing together around twenty top-level domain name registries, and to connect them with Smart Africa’s actions. He presents these registries as building blocks of sovereign infrastructure, concentrating expertise, autonomy, and technical dialogue with governments. Lacina Konaté presents this signing as the formalization of a partnership already long underway. He stresses that issues of Internet governance remain fundamental, because there can be no viable AI without good Internet governance and without preserving essential technical resources. The practical conclusion of the morning is therefore consistent with all the exchanges: beyond principles, digital sovereignty also requires institutional and technical partnerships on the Internet’s critical infrastructures. Overall, this opening sequence and first roundtable reveal several clear convergences, despite differences in emphasis on action priorities. First, the Francophonie is seeking to establish itself as a space for structured dialogue, connected to global arenas, in order to weigh collectively on digital and AI governance. Second, linguistic and cultural diversity is conceived here as a condition of the universality of AI, not as a peripheral issue. Third, the opposition between regulation and innovation is largely rejected in favor of an approach based on trust, standardization, and proportionate frameworks. Finally, the debates show that the main line of discussion does not concern the abstract alternative of whether or not to regulate, but rather the levers to prioritize: rules, infrastructure, participation in standardization, digital commons, computational sovereignty, data governance, or training.

The morning’s implicit conclusion is that none of these dimensions can suffice on its own. The speakers converge on the idea that credible Francophone governance of AI will have to combine regulation, standardization, production of linguistic resources, computing infrastructure, data governance, technical cooperation, mass training, and a stronger presence in international bodies. In this sense, the forum does not merely open a reflection: it seeks to sketch out a Francophone action agenda that is at once normative, technical, cultural, and geopolitical, aligned with the major international debates on the concentration of digital power, unequal access, multilingual inclusion, and development.