Makerere University College of Health Sciences

A successful first Global Dialogue on AI Governance should deliver outcomes that move beyond discussion into practical, inclusive, and actionable progress. First, it should establish a shared baseline understanding of AI governance priorities across regions. Countries are currently at different stages of AI adoption and regulation, so aligning on core principles such as transparency, accountability, safety, and human rights would be a critical foundation. This alignment should also recognize the specific realities of low- and middle-income countries, ensuring that governance frameworks are not only globally relevant but also locally applicable. Second, the dialogue should produce a clear roadmap for collaboration. This includes commitments to knowledge sharing, capacity building, and technical support, particularly for countries with limited regulatory or technical infrastructure. Mechanisms for ongoing cooperation such as regional hubs, working groups, or joint research initiatives would help sustain momentum beyond the event itself. Third, success would mean elevating the voices of underrepresented stakeholders. Governments alone cannot shape effective AI governance. Participation from academia, civil society, the private sector, and communities from the Global South is essential to ensure that policies are inclusive, equitable, and responsive to real-world needs. Finally, the dialogue should result in tangible outputs. These could include a draft framework or set of guiding principles, pilot initiatives to test governance approaches in different contexts, and a commitment to monitor progress through measurable indicators. In essence, the dialogue would be successful if it transitions AI governance from fragmented, high-level conversations into coordinated, inclusive, and action-oriented efforts that can guide responsible AI development globally.

AI capacity-building;Interoperability of governance approaches;Open-source software, open data and open AI models;Transparency, accountability, and human oversight;

From my perspective, the four priority areas for urgent action and active engagement are: 1. AI capacity-building This is the most immediate need, especially in low- and middle-income settings. Strengthening technical, regulatory, and institutional capacity ensures that countries can not only adopt AI responsibly but also actively shape its governance. Without this, global AI governance risks becoming uneven and exclusionary. 2. Transparency, accountability, and human oversight Trust in AI systems depends on clear mechanisms for explaining decisions, assigning responsibility, and ensuring meaningful human control. Prioritizing this area helps safeguard ethical use, particularly in sensitive sectors like health, where decisions directly affect lives. 3. Interoperability of governance approaches Given the global nature of AI, fragmented regulatory frameworks can create inefficiencies and risks. Promoting interoperability allows different countries and systems to align on core standards while maintaining contextual flexibility, enabling cross-border collaboration and innovation. 4. Open-source software, open data, and open AI models Open ecosystems can accelerate innovation, reduce costs, and democratize access to AI technologies. For many institutions, especially in resource-constrained environments, open approaches are essential for experimentation, localization, and scaling impactful solutions. Together, these priorities balance immediate capacity needs with long-term governance structures, ensuring that AI development is inclusive, coordinated, and grounded in accountability.

Yes. While the listed themes are critical, several cross-cutting and emerging issues deserve explicit attention. 1. Contextual equity and localization AI systems are often developed using data and assumptions from high-income settings, which can limit their relevance and fairness elsewhere. There is a need to prioritize localization of data, language, and use cases so that AI solutions reflect local realities, especially in sectors like health, agriculture, and education. 2. Data governance and sovereignty Beyond openness, questions of who owns, controls, and benefits from data are increasingly important. Countries and communities need clear frameworks for data stewardship, cross-border data flows, and protection against extractive data practices, particularly where sensitive public data is involved. 3. Implementation and real-world integration Much of the global conversation focuses on principles, but less attention is given to how AI is deployed within existing systems. Integrating AI into workflows, ensuring interoperability with legacy systems, and supporting frontline users (such as health workers) are critical for meaningful impact. 4. Sustainability and long-term maintenance AI systems require continuous updates, monitoring, and resources. There is a risk of pilot-driven innovation without long-term support. Governance discussions should include sustainable financing models, local ownership, and lifecycle management. 5. Evaluation and evidence generation There is still limited rigorous evidence on the effectiveness, cost-effectiveness, and unintended consequences of AI in many real-world settings. Strengthening evaluation frameworks, including ethical and impact assessments, is essential for informed decision-making. Addressing these issues would ensure that AI governance is not only principled but also practical, equitable, and sustainable across diverse contexts.

Governance gaps in AI are already shaping both the risks and opportunities across Uganda, Africa, and academia. Challenges A major challenge is limited AI capacity. Many institutions lack the technical expertise, infrastructure, and regulatory readiness to design, deploy, and oversee AI systems effectively. This creates dependence on external technologies, often without sufficient local adaptation. There are also gaps in transparency and accountability. AI tools are increasingly being introduced in sectors like health and education, yet there are few standardized mechanisms for auditing algorithms, explaining outputs, or assigning responsibility when systems fail. This raises concerns about trust, safety, and ethical use. Fragmented governance approaches across countries in Africa further complicate matters. Without harmonized frameworks, cross-border collaboration, data sharing, and scaling of AI solutions remain difficult. This fragmentation slows innovation and creates regulatory uncertainty. Additionally, while open-source and open data present opportunities, weak data governance frameworks expose risks around privacy, misuse, and inequitable data extraction, especially where safeguards are still evolving. Opportunities Despite these gaps, there is strong potential for leapfrogging. With the right investments in capacity-building, African countries can adopt context-specific AI solutions that directly address local challenges, particularly in health systems, agriculture, and education. The growing interest in open AI ecosystems provides an opportunity for academia and local innovators to experiment, adapt, and co-create solutions at lower cost. This is especially relevant for universities and research institutions seeking to build locally relevant evidence. There is also momentum for regional collaboration, which can support the development of interoperable governance frameworks, shared standards, and joint research initiatives. Overall, addressing these governance gaps can position Uganda and the broader African region not just as adopters, but as active contributors to responsible and inclusive AI development.

The AI Dialogue can play a practical convening and coordination role, helping translate global ambition into shared action. First, it can build a common foundation by aligning countries on core principles for responsible AI. Many nations are developing policies in parallel, but without coordination. The Dialogue can help bridge these efforts, creating a baseline that supports trust, reduces fragmentation, and enables cross-border collaboration. Second, it can facilitate structured cooperation mechanisms. This includes establishing working groups, regional hubs, and technical partnerships that focus on priority areas such as capacity-building, data governance, and standards development. For regions like Africa, this is especially important to ensure that cooperation is not one-sided but supports local ownership and capability. Third, the Dialogue can serve as a platform for knowledge exchange and peer learning. Countries and institutions can share practical experiences, lessons learned, and evidence from real-world AI deployments. This helps move discussions beyond theory and supports more informed, context-sensitive policymaking. Fourth, it can mobilize resources and partnerships. By bringing together governments, development partners, academia, and the private sector, the Dialogue can catalyze investments in infrastructure, skills development, and research, particularly for under-resourced regions. Finally, the Dialogue can promote accountability and continuity. Through agreed follow-up mechanisms, progress tracking, and periodic engagements, it can ensure that commitments made are implemented and refined over time. In essence, the AI Dialogue can act as a bridge between global coordination and local implementation, enabling inclusive, sustained, and action-oriented international cooperation on AI governance.

The AI Dialogue should build on existing global, regional, and sectoral initiatives to avoid duplication and strengthen coherence. At the global level, frameworks such as the UNESCO Recommendation on the Ethics of AI and the OECD AI Principles provide widely recognized normative guidance. Multi-stakeholder platforms like the Global Partnership on AI (GPAI) and standards bodies such as the International Telecommunication Union (ITU) are already advancing technical standards, policy dialogue, and applied research. Regionally, Africa is making progress through efforts led by the African Union, including emerging AI strategies and data governance discussions. Academic and research networks, as well as partnerships between universities and implementation organizations, are also contributing to locally grounded innovation and evidence generation. However, these efforts often operate in parallel, with limited coordination across regions and sectors. The added value of the AI Dialogue lies in its ability to connect these initiatives into a more coherent ecosystem. It can serve as a neutral platform to align global principles with regional priorities, ensuring that frameworks are not only adopted but meaningfully implemented in diverse contexts. The Dialogue can also bridge gaps between policy and practice by linking standard-setting bodies with implementers on the ground, including governments, academia, and civil society. This would support the translation of high-level principles into actionable tools, guidelines, and pilot programs. Additionally, it can amplify underrepresented voices, particularly from the Global South, ensuring that international AI governance reflects a broader range of experiences and priorities. Ultimately, the AI Dialogue can move the ecosystem from fragmented initiatives toward coordinated, inclusive, and action-oriented collaboration.

Stakeholder contributions Governments can provide policy direction, regulatory frameworks, and public sector priorities. Academia can contribute research, evidence generation, and independent evaluation of AI systems. The private sector brings technical expertise, innovation capacity, and scalable solutions. Civil society and community organizations play a critical role in highlighting ethical concerns, societal impacts, and user perspectives, especially for vulnerable populations. Development partners can support financing, coordination, and capacity-building efforts. Recommended format and structure First, the Dialogue should combine high-level plenary sessions with focused technical working groups. Plenaries can align on principles and priorities, while smaller groups can address specific themes such as capacity-building, data governance, or transparency in more depth. Second, it should include regional and sector-specific tracks. This allows participants from regions like Africa, or sectors such as health and education, to engage in context-driven discussions and share practical experiences. Third, the Dialogue should adopt a hybrid and accessible format. Virtual participation, multilingual support, and inclusive scheduling can help ensure broader participation, particularly from low-resource settings. Fourth, there should be structured outputs and follow-up mechanisms. Each session or working group should aim to produce clear recommendations, action plans, or pilot ideas, with defined responsibilities and timelines. Finally, the Dialogue should integrate voices from the ground through case studies, demonstrations, and user experiences. This ensures that discussions remain practical and grounded in real-world needs.

1. Communities from low- and middle-income countries (LMICs) Much of the AI governance agenda is shaped by high-income countries, yet LMICs are increasingly affected by AI deployments. Their policy priorities, infrastructure realities, and development needs are often insufficiently reflected. 2. Frontline practitioners Health workers, teachers, and community-based actors who interact directly with AI-enabled tools are rarely included. Their practical insights on usability, risks, and real-world impact are critical for effective governance. 3. Local researchers and academic institutions While global research institutions are well represented, many local universities and researchers, particularly in Africa, have limited visibility despite generating context-specific evidence. 4. Marginalized and vulnerable populations Groups such as rural communities, women, and persons with limited digital access are often excluded, yet they are among the most affected by biased or poorly designed AI systems. How to include them First, provide targeted support for participation, including travel funding, stipends, and connectivity support for virtual engagement. Second, establish regional consultation platforms that feed into global processes, ensuring that local perspectives are gathered systematically rather than incidentally. Third, integrate practitioner and community voices through case studies, user panels, and field-based evidence sessions, not just policy discussions. Fourth, promote equitable research partnerships, where local institutions are co-creators rather than data providers, with shared ownership of outputs. Finally, ensure multilingual and accessible formats to lower participation barriers.

1. Policy-to-practice labs Small, facilitated groups work on real governance challenges (e.g., regulating AI in health systems or managing cross-border data flows). Participants co-develop practical solutions, draft guidelines, or prototype policies within the session. 2. Live case demonstrations Stakeholders present real-world AI deployments, followed by structured critique from policymakers, researchers, and end-users. This grounds discussions in reality and surfaces implementation challenges early. 3. Scenario-based simulations Participants engage in role-play exercises around emerging risks (e.g., AI system failure, bias in decision-making). This helps test governance responses, clarify roles, and identify gaps in current frameworks. 4. Regional and sector roundtables Focused discussions for regions like Africa or sectors such as health and education allow participants to explore context-specific issues and generate tailored recommendations. 5. Multi-stakeholder design sprints Short, intensive sessions where mixed groups (government, academia, private sector, civil society) co-create solutions such as model policies, ethical guidelines, or capacity-building plans. 6. Open innovation and solution showcases Platforms where innovators, especially from underrepresented regions, present tools and approaches. Participants can explore partnerships, scaling opportunities, and adaptation pathways. 7. Structured networking and matchmaking Curated sessions that connect stakeholders with shared interests, enabling partnerships beyond the Dialogue. 8. Continuous engagement platforms Digital workspaces (before and after the Dialogue) to share resources, track progress, and sustain collaboration.

1. Risk-based regulatory approaches The European Union AI Act provides a structured model that classifies AI systems by risk level and applies proportionate obligations. This approach can be adapted by African countries to balance innovation with safeguards, especially in sensitive sectors like health and finance. 2. Continental policy frameworks The African Union is advancing an AI strategy that emphasizes inclusive development, data governance, and capacity-building. It offers a regional foundation for harmonizing policies and promoting collaboration across countries. 3. Data protection as a foundation Uganda's Data Protection and Privacy Act provides a legal basis for responsible data use. Such frameworks are essential for AI governance, as they address consent, privacy, and accountability in data-driven systems. 4. Open and collaborative platforms Platforms like Hugging Face support transparency and accessibility by providing open-source models, datasets, and evaluation tools. These enable researchers and developers, including those in Africa, to build and adapt AI systems responsibly. 5. Standard-setting and global coordination The International Telecommunication Union (ITU) develops technical standards and convenes global initiatives such as AI for Good, helping translate governance principles into implementable practices. 6. Responsible AI practices in implementation Tools such as algorithmic impact assessments, model documentation, and independent audits are increasingly used to ensure transparency, detect bias, and improve accountability during AI deployment.