Government of Kenya

The first Global Dialogue on AI Governance will succeed if it shifts the conversation from principles to practical, shared progress. It should acknowledge that governance involves not only managing risks but also building capability across the entire AI stack — data, compute, talent, models, and use cases. In the Age of Intelligence, countries need to develop the ability to participate meaningfully in shaping and deploying AI while maintaining interoperability in an interconnected world. It will also be important for the dialogue to place the Global South at the center of the conversation. Success would mean governance frameworks that reflect different national starting points, development realities, and public-interest priorities. Inclusion must mean more than just representation; it also needs to ensure real agency in shaping norms, priorities, and institutional arrangements. For instance, it also means that conversations such as AI safety that have been focused on technical risks must now start to also encompass social technical risks that face global south countries, rather than frontier risks. A meaningful outcome would also connect governance to the delivery of AI benefits. For many countries, the greatest constraints are infrastructure, energy, financing, skills, and evaluation capacity. The Dialogue should therefore elevate capacity-building as a core governance issue and support innovative financing, stronger partnerships, and implementation pathways that enable countries to participate as builders. Finally, the Dialogue should help reduce fragmentation across the growing ecosystem of AI initiatives and institutions. Its value should be in creating greater coherence, complementarity, and coordination — a genuine dialogue of dialogues rather than another silo. Above all, success would mean advancing AI governance that is safe, secure, trustworthy, and inclusive, while ensuring that AI contributes meaningfully to sustainable development. If countries leave better heard, better aligned, and better equipped to act, the first Dialogue will have achieved its objectives.

• AI capacity-building
• Safe, secure and trustworthy AI
• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI

My priorities for urgent action and active engagement are: (1) Safe, secure and trustworthy AI; (2) AI capacity-building; (3) Social, economic, ethical, cultural, linguistic and technical implications of AI; and (4) Transparency, accountability, and human oversight. Taken together, these areas reflect the need for AI governance that is not only normatively sound, but also inclusive, development-oriented, and implementable. Safe, secure, and trustworthy AI remains fundamental. As AI systems are adopted more broadly across public institutions, markets, and societies, governance must ensure these systems are reliable, responsible, and aligned with human well-being, public interest, and sustainable development. Trust cannot be an afterthought; it is essential for both legitimacy and adoption. AI capacity-building is just as urgent. For many countries, especially in the Global South, the main challenge is not only how to regulate AI but also how to participate meaningfully in shaping, deploying, and governing it. Capacity-building must be understood broadly: including talent, institutions, data systems, compute infrastructure, evaluation capabilities, and the ability to turn policy into action. This also requires strong attention to financing. Without sufficient and innovative funding-including public investment, development finance, blended finance, and catalytic partnerships-capacity-building risks staying aspirational. For inclusion to be truly meaningful, it must be supported by resources that help countries build, adapt, and maintain their own capabilities. The social, economic, ethical, cultural, linguistic, and technical implications of AI are also key, as countries start from very different points in AI. Governance frameworks must account for these differences and ensure that development realities, language diversity, local contexts, and public-interest priorities are fully considered. Finally, transparency, accountability, and human oversight are essential to ensuring that AI systems remain understandable, contestable, and ultimately answerable to people. These key elements support an approach to AI governance that is safe and trustworthy, yet also practical, fair, and responsive to the realities of an interconnected world.

While the listed themes are comprehensive, a few cross-cutting issues I would suggest are First is Financing for AI governance and capacity-building. For many developing countries, the challenge is not only policy design, but the ability to resource talent, institutions, data systems, evaluation capacity, and implementation. Without sustainable and innovative financing - including public investment, development finance, blended finance, and catalytic partnerships - capacity-building risks remaining aspirational, and meaningful participation will remain uneven. Second is environmental sustainability across the AI lifecycle. As AI scales, governance should address both how AI can support sustainable development and environmental action, and how to manage the environmental footprint of AI systems themselves, including energy use, water consumption, minerals, and e-waste across the lifecycle. Third is the need to strengthen the UN and the wider multilateral system in AI governance. As the number of initiatives, institutions, and governance processes continues to grow, there is a clear need for greater coherence, complementarity, and legitimacy. AI governance should reinforce multilateralism, reduce fragmentation, and help ensure that global norms are shaped through inclusive processes that reflect the realities and priorities of all countries.

While Kenya has made important progress — including an AI Strategy, AI Principles, and ongoing work toward an AI Policy — the central governance gap remains that AI is advancing faster than the institutions, infrastructure, and financing required to guide it in the public interest. This is not unique to Kenya; it is a wider challenge across Africa and much of the Global South. The risk is that countries may adopt and consume AI systems developed elsewhere without having sufficient influence over the rules, standards, and value chains that increasingly shape economic competitiveness, social outcomes, and strategic autonomy. The most significant challenge is therefore not only regulatory readiness, but capability readiness. Even where policy frameworks are emerging, many countries still face major constraints in compute, data systems, energy infrastructure, talent, evaluation capacity, and institutional coordination. Financing cuts across all of these gaps. Without adequate and innovative financing, capacity-building remains aspirational, and participation in AI governance risks being formal rather than substantive. There are also growing concerns around transparency, accountability, and human oversight, especially as AI begins to shape public services, labor markets, information ecosystems, and democratic trust. At the same time, the opportunities are substantial. For Kenya and many African countries, AI can help unlock progress in agriculture, health, education, public administration, climate resilience, and financial inclusion. It also creates an opportunity to move beyond being users of imported systems toward building local capability, strengthening innovation ecosystems, and developing context-relevant applications, including in local languages and for local realities. These governance gaps are therefore double-edged. If left unaddressed, they could deepen dependency, inequality, and fragmentation. If addressed well, they could enable more inclusive growth, stronger public institutions, and a more balanced and representative global AI governance architecture.

The AI Dialogue can play an important role by helping to reduce fragmentation in the fast-growing AI governance landscape. Today, discussions are taking place across multiple forums, institutions, and coalitions, but many countries remain outside the spaces where priorities are being shaped. For that reason, the Dialogue should be designed not as another standalone process, but as a **dialogue of dialogues**: a platform that connects existing initiatives, identifies gaps, and builds greater coherence and complementarity across the wider ecosystem. It can also help ensure greater inclusivity, especially for countries in the Global South that are not part of forums such as the G7, G20, GPAI, the AI Safety or AI Impact Summits, or the OECD. If it is to have real value, the Dialogue must widen participation beyond the usual actors and create space for countries whose realities, constraints, and priorities are often underrepresented in global debates. Inclusion must mean not only presence, but influence. A further contribution would be to surface issues that are often insufficiently addressed in mainstream AI governance discussions. These include linguistic diversity, sovereignty, infrastructure gaps, financing, and the practical question of how countries can build capability without each having to replicate expensive systems on their own. In many cases, more cooperative, shared, and regionally anchored approaches to infrastructure, compute, and capacity-building may be more realistic and sustainable than isolated national efforts. The Dialogue should also strengthen multilateralism. At a time when AI governance risks being shaped by a limited number of actors and institutions, the UN offers a more universal and legitimate platform for building shared understanding. Its value lies not only in convening, but in ensuring that AI governance reflects the voices, interests, and development realities of all countries. If shaped as a genuine dialogue of dialogues, it can help foster coherence, inclusion, and practical cooperation in global AI governance.

The AI Dialogue should build on the Global Digital Compact architecture, including the Global Dialogue itself and the Independent International Scientific Panel on AI, as well as the 2024 General Assembly resolutions on safe, secure and trustworthy AI for sustainable development and on international cooperation on AI capacity-building. Together, these provide the most universal intergovernmental foundation currently available and already establish AI governance, scientific understanding, and capacity-building as shared multilateral priorities. It should also connect with existing normative and policy processes such as the UNESCO Recommendation on the Ethics of AI, the WSIS process, the Internet Governance Forum, and relevant work on interoperability emerging from the OECD AI Principles. These initiatives already contain valuable experience on ethics, human rights, accountability, multistakeholder participation, and practical policy cooperation. In addition, the Dialogue should engage the ecosystem emerging around the AI Safety and AI Impact Summits, including the growing body of work on AI safety institutions, evaluation, measurement, and standards. These processes have helped advance technical and policy thinking on frontier risks, testing, and governance tools, but they are not universal in participation. Connecting them to a broader multilateral process would help ensure that their insights are shared more widely and shaped by a more inclusive set of countries and perspectives. From an African perspective, the Dialogue should also build on the African Union Continental AI Strategy and the AU Data Policy Framework. These are important because they bring development realities, digital sovereignty, linguistic diversity, infrastructure constraints, and regional approaches to implementation into the conversation. The added value of the AI Dialogue is that it can serve as a true dialogue of dialogues: reducing fragmentation, connecting parallel efforts, and strengthening a more coherent, inclusive, and legitimate multilateral architecture for AI governance.

Different stakeholders should contribute in ways that reflect their comparative strengths, while keeping governments at the centre of norm-shaping and multilateral legitimacy. Member States should articulate national and regional priorities, identify governance gaps, and help ensure that outcomes reflect diverse development realities. The UN system can provide convening legitimacy, policy coherence, and links to existing multilateral processes. Technical experts and scientists should help clarify emerging risks, measurement challenges, standards, and evidence gaps. The private sector should contribute practical insight on deployment, infrastructure, safety practices, and investment realities, while remaining accountable to public-interest objectives. Civil society, academia, and affected communities are essential to ensuring that human rights, inclusion, linguistic diversity, and real societal impacts are not treated as secondary concerns. In terms of format, the AI Dialogue should be structured as a dialogue of dialogues, not as another isolated forum. Its purpose should be to connect existing processes, surface gaps, and strengthen complementarity across the wider ecosystem. It should combine high-level political engagement with technical and implementation-oriented tracks, so that discussions do not remain purely declaratory. A useful structure could include: a plenary segment for strategic direction; thematic roundtables on priority issues such as safety, capacity-building, financing, infrastructure, interoperability, and sustainable development; and dedicated multi-stakeholder sessions that engage technical institutions, industry, civil society, and regional bodies. Regional consultations should feed into the global meeting so that participation is broader and more grounded in local realities.

Several voices and perspectives remain underrepresented in global discussions on AI governance. These include countries from the Global South that are not part of smaller decision-shaping forums; linguistically and culturally underrepresented communities; and groups most likely to be excluded from the benefits of AI or disproportionately exposed to its risks, including women and girls, young people, persons with disabilities, rural communities, indigenous communities, and low-income or otherwise marginalized populations. Also underrepresented are those closest to practical implementation, such as educators, public servants, local innovators, researchers in low-resource environments, workers affected by technological change, and small enterprises. Too often, global discussions are dominated by a relatively narrow set of governments, firms, and institutions. Their inclusion requires intentional design. This means regional consultations that feed into global processes, financial support for participation by low-capacity countries and communities, and multilingual documentation, interpretation, and local-language engagement. It also means ensuring that participation is fully accessible to persons with disabilities, including through accessible venues, digital platforms, documents, and communication formats. Inclusion should also be reflected in the structure of the Dialogue itself: balanced representation in panels, advisory groups, and preparatory processes; dedicated space for civil society, academia, youth, and affected communities; and consultation methods that value lived experience and implementation realities alongside technical and policy expertise.

Given that the Dialogue is only two days, the format should be disciplined, interactive, and designed to maximize exchange rather than speeches. A useful approach would be to structure it around three layers. First, a high-level opening and closing plenary to set the political direction, frame the main questions, and capture key takeaways. Second, a limited number of thematic roundtables focused on priority issues, including safety, capacity building, interoperability, financing, and inclusion. These should be moderated as working sessions rather than formal panel discussions, and should bring together governments, technical experts, private-sector actors, and civil society in balanced groups. Third, there should be cross-cutting, multi-stakeholder sessions designed to surface often-missing perspectives, including those of Global South countries, youth, persons with disabilities, local innovators, and practitioners working in low-resource environments. To make the engagement more dynamic, the Dialogue could include short, scenario-based sessions focused on practical governance dilemmas. These would help move the discussion from general principles to concrete cooperation. It would also be useful to include a "dialogue of dialogues" segment in which existing initiatives and institutions briefly identify overlaps, gaps, and opportunities for coordination. That would help reduce fragmentation and clarify the added value of the UN process. Finally, the most effective format for a two-day Dialogue is one that is focused, inclusive, and action-oriented: high-level enough to build political legitimacy, but practical enough to generate real interaction and pathways for follow-up.

Useful examples include the African Union Continental AI Strategy, which offers a development-focused and Africa-centered approach to AI governance by linking ethics, inclusion, human rights, capacity-building, and innovation to local realities and public-interest priorities. Another important example is the Global Network of Artificial Intelligence Capacity-Building Centers, launched in 2025 by Saudi Arabia and Kenya and endorsed under the Global Digital Compact. It is valuable because it focuses on practical international cooperation to expand AI education, skills development, and institutional readiness, with particular relevance for Global South participation. The NIST AI Risk Management Framework is also useful because it translates broad governance goals into practical organizational steps, helping institutions operationalize risk management, accountability, and oversight. Another promising area is the growing ecosystem around AI safety institutions, measurement, evaluation, and standards, including the International Network for Advanced AI Measurement, Evaluation and Science, which supports shared scientific understanding and more recognized approaches to assessing advanced AI systems. Finally, the UNESCO Recommendation on the Ethics of AI remains an important global normative framework grounded in human rights, transparency, accountability, and environmental sustainability.