University Côte d'Azur

Building on my research on the AI divide and the twin digital–ecological transition, and informed by my experience supporting multilateral climate negotiations from COP23 to COP30, I respectfully suggest that the success of the first Global Dialogue could be measured against five complementary outcomes. First, a shared recognition that the AI divide warrants timely, cooperative action. Evidence increasingly suggests that disparities between the Global North and Global South in AI access, compute, and data may grow faster than earlier digital divides, with thresholds beyond which gaps become difficult to reverse. A common understanding of these dynamics, accompanied by indicative milestones for capacity-building, would represent a meaningful first step. Second, a multidimensional understanding of inclusion, encompassing not only access, but also meaningful use, performance, and developmental outcomes. Such a framing would help Member States and stakeholders design governance that is genuinely equitable. Third, exploratory pathways toward greater computational autonomy in developing countries, including possibilities for pooled compute access, openly available and representative datasets, and Global-South-anchored research partnerships. These structural enablers complement, and reinforce, regulatory dialogue. Fourth, coherence between AI governance and the climate and Sustainable Development agendas. Encouraging transparent reporting on the energy footprint, rebound effects, and sustainability contributions of AI initiatives would help align this Dialogue with commitments already undertaken under the UNFCCC and the 2030 Agenda. Fifth, early reflection on financing. Experience from climate diplomacy suggests that durable governance requires predictable resources. Initiating a conversation on blended finance and cooperative mechanisms for AI capacity in low-income countries would significantly strengthen the Dialogue's credibility.

• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Safe, secure and trustworthy AI
• Protection and promotion of human rights

1. Safe, secure and trustworthy AI The credibility of AI as a global public good depends on user, regulator, and citizen trust. Cross-border deployment raises shared challenges-algorithmic bias, opacity, energy footprints, and rebound effects-that no single jurisdiction can address alone. Coordinated international approaches to safety standards, independent evaluation, and transparent reporting would strengthen public confidence and help prevent the fragmentation of governance frameworks. 2. AI capacity-building Capacity-building is arguably the most consequential lever for an inclusive global AI ecosystem. Persistent asymmetries in compute, data, talent, and institutional readiness risk widening the gap between technology-producing and technology-receiving countries. Sustained cooperation on training, infrastructure-sharing, research partnerships, and South-South collaboration would help ensure that all Member States participate meaningfully in shaping and benefiting from AI, in line with the 2030 Agenda. 3. Social, economic, ethical, cultural, linguistic and technical implications of AI AI systems carry the imprint of the data and contexts in which they are developed, with significant consequences for under-represented languages, cultures, and labour markets. A shared international reflection on these implications-drawing on diverse disciplinary and geographic perspectives-would support governance that is culturally respectful, linguistically inclusive, and socioeconomically balanced across both developed and developing economies. 4. Protection and promotion of human rights Anchoring AI governance in the existing international human-rights framework provides a universally legitimate foundation for cooperation. Particular attention is warranted for non-discrimination, privacy, freedom of expression, and the rights of women, youth, and vulnerable communities, who may be disproportionately affected by automated decision-making. Alignment with established human-rights instruments would reinforce coherence across the multilateral system and help ensure that AI serves human dignity.

Beyond the themes identified, several cross-cutting issues would benefit from explicit consideration in the work of the Dialogue. The environmental footprint of AI and its alignment with climate commitments. AI's rapidly growing demand for energy, water, and critical materials, together with potential rebound effects, places it at the heart of the twin digital-ecological transition. Greater coherence between the AI Dialogue and the UNFCCC and 2030 Agenda processes would help ensure that AI supports, rather than complicates, existing climate and sustainability commitments. Compute and infrastructure governance. The concentration of AI compute, foundation models, and cloud infrastructure in a limited number of providers and jurisdictions has structural implications for sovereignty, competition, and resilience. International reflection on equitable, secure, and environmentally sound access to compute is increasingly necessary. Financing for inclusive AI development. Governance ambitions require predictable resources. Lessons from climate finance-on new and additional resources, blended-finance vehicles, and concessional support for low-income countries-could usefully inform a future financing architecture for AI capacity, public-interest AI research, and digital public infrastructure. Anticipatory governance and the pace of technological change. The rapid evolution of AI capabilities raises challenges of regulatory lag and policy obsolescence. Foresight mechanisms, agile review cycles, and shared horizon-scanning would enhance the resilience and longevity of governance instruments emerging from this process. Data governance as a cross-cutting layer. Cross-border data flows, the representativeness of training datasets, and the protection of personal and community data underpin all four thematic priorities and merit dedicated attention as a connective tissue across the Dialogue's work.

From the perspective of the Mediterranean and African region, where I have engaged through research, advisory work, and climate diplomacy, current governance gaps in the four selected areas present both significant challenges and emerging opportunities. Significant challenges Capacity-building gaps remain the most pressing. Limited access to compute, fragmented research infrastructure, persistent talent emigration, and uneven institutional readiness constrain meaningful participation in AI development. National AI strategies have multiplied across the region, but many still lack the financing, technical depth, and cross-border coordination needed for effective implementation. The under-representation of regional languages—Arabic and its varieties, Amazigh languages, and African languages—in AI training data creates measurable performance gaps, with downstream effects on education, public services, and cultural expression. Trustworthy AI is further challenged by reliance on imported systems whose evaluation, accountability, and redress mechanisms remain difficult to verify locally. Human-rights protections are tested by uneven data-protection frameworks and persistent digital divides along gender, age, and rural–urban lines. Finally, the convergence of AI's rising energy and water demand with the region's existing climate vulnerability adds a further layer of complexity, requiring deployment strategies that are coherent with national climate commitments. Emerging opportunities The region also presents real opportunities. A young, increasingly connected population offers considerable human-capital potential. South–South and Mediterranean cooperation frameworks, including continental and sub-regional initiatives, provide platforms for shared standards, pooled compute, and joint training programmes. The combination of expanding renewable-energy capacity and emerging AI ecosystems offers a credible foundation for a regionally adapted twin transition. A coordinated multilateral response, supported by the Dialogue, would help translate these opportunities into measurable progress on capacity, inclusion, and rights.

The AI Dialogue is uniquely positioned to play a convening, bridging, and norm-shaping role in an increasingly fragmented governance landscape. A convening role. As the only AI governance process anchored in universal Member State participation, the Dialogue can offer a legitimate space where governments, the scientific community, the private sector, civil society, and regional organisations engage on equal footing. This inclusive character is particularly valuable for ensuring that the perspectives of Global South countries, smaller economies, and under-represented stakeholders inform emerging norms. A bridging role. Numerous regional and plurilateral initiatives have advanced rapidly but remain uneven in scope and binding force. The Dialogue can help connect these efforts—identifying convergences, mapping divergences, and supporting interoperability—while reducing duplication and avoiding a fragmented landscape that disadvantages capacity-constrained countries. A norm-shaping and learning role. Through structured exchanges of experience, the Dialogue can facilitate peer learning, share lessons from early implementation, and gradually build shared understandings on contested issues. Its value lies in incremental, evidence-informed convergence rather than imposed solutions. A coherence role across the UN system. AI cuts across climate action, human rights, sustainable development, labour, health, and education. The Dialogue can foster coherence with existing multilateral processes, including the UNFCCC, the 2030 Agenda, and human-rights mechanisms, ensuring that AI governance reinforces broader international commitments. A foresight role. Given the speed of technological change, the Dialogue can host horizon-scanning exercises and structured scientific input, helping Member States anticipate emerging risks and opportunities and reducing regulatory lag. In playing these roles, the AI Dialogue can complement existing frameworks while offering what no single regional process can: universal legitimacy, sustained continuity, and connection to the wider multilateral system.

The AI Dialogue does not begin on a blank slate. A rich ecosystem of complementary initiatives has emerged, on which it can usefully build. Within the UN system. The Global Digital Compact, the work of the Secretary-General's High-Level Advisory Body on AI, the UNESCO Recommendation on the Ethics of Artificial Intelligence, the ITU AI for Good process, and the contributions of WHO, ILO, UNDP, UNCTAD, and WIPO offer well-established normative and operational foundations. Connections with the UNFCCC and the 2030 Agenda are also essential, given AI's environmental footprint and its potential contribution to the SDGs. Plurilateral and regional initiatives. The OECD AI Principles, the G7 Hiroshima AI Process, the G20 AI work, the Council of Europe Framework Convention on Artificial Intelligence, the European Union's AI Act, the African Union Continental AI Strategy, and the network of AI Safety Institutes all provide important reference points and lessons learned. Scientific and multi-stakeholder networks. The Global Partnership on AI, regional research networks, and the broader scientific community offer technical depth that the Dialogue can draw on, alongside contributions from civil society, indigenous communities, and the private sector. Added value of the AI Dialogue. What the Dialogue can uniquely bring is universal legitimacy, the genuine inclusion of all Member States—particularly those under-represented in current frameworks—and structural connection to the rest of the multilateral system. By mapping existing efforts, identifying common ground, supporting interoperability, and providing a continuous platform for exchange, it can help transform a patchwork of initiatives into a more coherent, equitable, and forward-looking global governance architecture, while respecting the autonomy and complementarity of existing instruments.

The Dialogue's effectiveness will rest on its ability to mobilise the complementary expertise of all stakeholder groups while preserving the central role of Member States.Governments can anchor the political legitimacy of the process, share regulatory experience, and align national strategies with emerging shared understandings. Intergovernmental and regional organisations can ensure coherence with existing mandates. The scientific and research community can provide independent analysis through a structured advisory function. The private sector, including AI developers, deployers, and SMEs, can contribute technical insight on real-world implementation. Civil society, workers, youth, women, indigenous peoples, and persons with disabilities bring perspectives essential to legitimacy and equity.Recommendations on format and structureA two-tier structure could combine an annual plenary—open to all Member States and accredited stakeholders—with thematic working groups operating between sessions on the priorities identified by Resolution 79/325. Regional preparatory consultations in Africa, Asia, the Americas, and Europe would strengthen geographic balance and substantive grounding.A permanent multidisciplinary scientific advisory function, co-chaired across regions, could provide structured evidence and horizon-scanning, complementing the political dialogue.Stakeholder participation could be organised through transparent constituencies, drawing on the experience of the UNFCCC observer system, with clear channels for input and synthesis. Hybrid participation, accessible documentation, and full multilingualism would help ensure that capacity is not a barrier to meaningful engagement.Outputs could take the form of voluntary recommendations, shared good practices, and progress reviews, prioritising incremental convergence and trust-building over premature binding instruments.

Despite considerable progress, several voices remain structurally under-represented in global AI governance discussions. Global South countries, particularly Least Developed Countries, Small Island Developing States, and Landlocked Developing Countries, often participate with limited delegation capacity, despite being significantly affected by AI's diffusion and concentration dynamics. Indigenous peoples and local communities are seldom systematically engaged, although questions of data sovereignty, language, and cultural representation directly affect them. Women, particularly from the Global South, remain under-represented in technical and decision-making spaces. Youth are rarely engaged as full participants, although they will live with the long-term consequences of present decisions. Workers, trade unions, and SMEs are not consistently consulted, despite AI's profound effect on labour markets and value chains. Local and subnational governments are largely absent, although they sit at the front line of implementation. Researchers from under-resourced institutions in Africa, Latin America, and parts of Asia face structural barriers to participation in agenda-setting fora. Suggested inclusion measures Sustained inclusion requires more than open invitations. The Dialogue could combine: dedicated travel and participation support; full multilingual interpretation including widely spoken non-UN languages; balanced regional and gender representation in working groups and advisory bodies; structured pre-Dialogue capacity-building; accessible hybrid formats; and explicit channels for community-led submissions. Establishing standing constituencies for youth, women, indigenous peoples, workers, and SMEs—each with clear roles in the working groups—would help institutionalise inclusion. Partnerships with regional networks, research consortia, and civil society platforms could further extend reach. Inclusion is not only an ethical requirement but a practical condition for the Dialogue's legitimacy and effectiveness.

Building on emerging multilateral and scientific practice, several innovative engagement formats could strengthen the Dialogue's relevance and dynamism. Structured deliberative formats. Citizens' assemblies and expert deliberation panels, drawn from diverse regions and backgrounds, could examine complex governance trade-offs and provide reasoned input on contested questions. Living documents and continuous synthesis. Rather than fixed annual outputs, the Dialogue could maintain regularly updated synthesis documents reflecting the state of the discussion, areas of common ground, and remaining divergences—supporting transparency and momentum between sessions. Regional and thematic field demonstrators. Practical case studies and pilots—on AI in education, agriculture, health, public services, climate adaptation, or smart manufacturing—could be presented and learned from collectively, grounding the Dialogue in implementation realities. Open virtual consultations. Building on the format of this present consultation, accessible online platforms with multilingual interfaces, structured submission templates, and transparent synthesis would help extend participation well beyond Geneva and New York. Scientific exchanges. Structured interactions with the scientific community—through joint reports, evidence sessions, and horizon-scanning exercises—could ensure that the Dialogue is informed by independent, peer-reviewed insight. Cross-stakeholder peer-learning networks. South–South, North–South, and triangular learning networks could share experience on national strategies, regulatory sandboxes, and capacity-building initiatives. Use of AI as a supporting tool. With appropriate safeguards on transparency, bias, and human oversight, AI itself could assist with translation, accessibility, synthesis of submissions, and visualisation of contributions—provided that human deliberation remains central. Public engagement and visibility. Open-access dashboards tracking commitments and progress, combined with periodic stocktaking, would enhance accountability and public trust. Together, these formats would help ensure that the Dialogue is participatory, evidence-based, and continuously responsive to a rapidly evolving field.

Several existing policies, practices, and platforms offer useful illustrations of effective approaches to AI governance. Normative and ethical frameworks. The UNESCO Recommendation on the Ethics of Artificial Intelligence, supported by its Readiness Assessment Methodology, demonstrates how a global ethical instrument can be operationalised at country level through structured self-assessment and capacity-building. Risk-based regulatory approaches. The European Union's AI Act and the Council of Europe Framework Convention on Artificial Intelligence illustrate how risk-tiered, rights-anchored regulation can address concrete harms while preserving innovation. Their experience offers transferable lessons on classification, conformity assessment, and oversight. Regional strategies. The African Union Continental AI Strategy and several national strategies in the Mediterranean and African region show how regional coordination can articulate shared priorities while respecting national specificities. Safety institutes and scientific cooperation. The emerging network of AI Safety Institutes and the international scientific report on advanced AI demonstrate the value of coordinated, independent technical evaluation. Sectoral guidance. WHO guidance on the ethics and governance of AI for health, ILO work on AI and the future of work, and UNESCO guidance on AI in education provide sector-specific blueprints aligned with existing mandates. Implementation tools. Regulatory sandboxes in finance, mobility, and health; AI impact assessments; public registries of high-risk systems; and procurement standards for trustworthy AI offer practical instruments adaptable to different contexts. Multi-stakeholder platforms. The ITU AI for Good Global Summit and the Global Partnership on AI illustrate the value of recurring, inclusive convening that bridges technical, policy, and societal communities. Sustainability-oriented practice. Initiatives integrating AI governance with climate and sustainability commitments-such as transparency on data-centre energy use and twin-transition strategies-offer early models for coherence across the UN agenda.