United Nations

A successful Global Dialogue on AI Governance must produce outcomes that go beyond symbolic declarations. It should create mechanisms that acknowledge the structural inequities, particularly the entrenched digital divide, the uneven global distribution of AI research capacity, and the geopolitical fragmentation that shapes how AI norms emerge. Success must therefore include a shared recognition that the current governance landscape is marked by deep interconnectedness yet persistent disconnection. Many powerful states have matured their regulatory frameworks while developing countries struggle to build even foundational digital infrastructure. An effective dialogue must not ignore this imbalance. It should instead treat it as a central governance concern. A second metric of success involves concrete pathways for capacity building. The Dialogue must commit to multiyear programs that strengthen institutional capabilities in the Global South through training, technical assistance, and resource mobilization. Without such interventions, the AI divide will widen. Yet the Dialogue must also remain realistic. Capacity building efforts often fail when they are designed without attention to local governance environments, infrastructure, political constraints, and scalability challenges in highly populous countries such as Ethiopia or Nigeria. A third outcome relates to global coherence. Success requires an honest confrontation with the geopolitical competition that currently shapes AI governance. The Dialogue should establish processes that reduce fragmentation while avoiding over centralization. Attempting to impose uniform governance models would risk marginalizing states with limited negotiating power. Instead, the Dialogue should construct interoperable principles that can accommodate diverse regulatory philosophies while protecting human rights, promoting transparency, and safeguarding global security. Finally, success demands implementation tracking. The Dialogue must establish measurable follow up mechanisms that ensure commitments evolve into tangible reforms rather than rhetorical alignment.

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

I have selected the following four thematic areas for urgent action because they align with both my field experience and the systemic risks identified across my research work: 1. Safe, secure, trustworthy AI 2. Transparency, accountability, human oversight (Protection and promotion of human rights is included here as well) 3. Social, economic, ethical, cultural, linguistic, technical implications 4. AI capacity-building My involvement in AI governance research highlights that safe, secure, trustworthy AI is not merely a technical priority. It is a foundational requirement for any society that intends to deploy AI in essential sectors. Across Sub-Saharan Africa, where infrastructure gaps and uneven institutional readiness remain persistent, safety failures carry disproportionate consequences. My work on the digital divide stresses that trustworthiness depends not only on security protocols but on broader societal conditions that determine how AI systems are adopted, questioned, or resisted. Transparency, accountability, and human oversight are equally essential. In environments with limited regulatory maturity, opaque AI systems can reinforce inequalities and weaken public confidence. However, I remain cautious about assuming that transparency alone will fix structural governance gaps. Oversight mechanisms must be supported by institutions that actually have the authority, technical skill, and political stability to enforce them. This directly and indirectly support protection and promotion of human rights) The social, economic, ethical, cultural, linguistic, and technical implications of AI require urgent attention because they shape long term development trajectories. Research illustrates that these implications differ significantly across regions. Treating them as universal would risk imposing external governance models that might not align with local realities. AI capacity-building is indispensable. It must be understood as a sustained, multi layer effort rather than a quick intervention. Part of my research work repeatedly shows that without strong human capital, no governance commitment will translate into effective implementation.

Cross-Cutting or Emerging Issues That Should Be Explicitly Mentioned or Categorized as Thematic Areas 1. AI implications for weapons of mass destruction, chemical agents, and high-risk dual-use science This issue aligns with safe, secure, trustworthy AI because current thematic areas do not explicitly address the ways AI accelerates modelling, experimentation support, or logistical optimisation that can influence chemical, biological, or nuclear threats. Existing non-proliferation regimes do not yet account for AI as an enabling factor, which creates a governance gap that must be integrated into global safety discussions. 2. Concentration of computational power, data resources, and strategic dependency Researches suggest including my work that current themes do not capture the structural risks created by extreme concentration of compute, frontier model training resources, or cloud infrastructure. These monopolized assets determine participation, technical baselines, and foreign supply chain dependency, which deepens strategic inequality and geopolitical asymmetry. AI procurement practices intensify this risk when opaque contracts or exclusive cloud arrangements restrict auditability, limit oversight, and weaken local adaptation. Governance frameworks must categorize compute concentration, data resources, strategic dependency, and procurement practices as a single thematic area to protect sovereignty, accountability, and long term resilience. 3. Environmental sustainability and resource-intensive AI ecosystems This belongs within the broader social, economic, ethical, cultural, linguistic, and technical implications of AI. Large-scale AI models impose significant energy demands, hardware extraction pressures, and water usage for cooling. These burdens fall disproportionately on countries with limited environmental protections. Governance frameworks must incorporate environmental thresholds and require environmental reporting for AI system development and deployment.

Governance gaps in the selected thematic areas affect my country Ethiopia and my region Sub‑Saharan Africa in ways that are structural, systemic, and persistent. The region operates in an environment characterized by limited digital infrastructure, uneven connectivity, and substantial disparities in institutional capacity. These conditions amplify the consequences of weak governance in safe, secure, trustworthy AI. Without strong regulatory institutions, rapidly advancing AI systems enter sectors such as agriculture, health, and public administration with minimal safeguards. This increases the risk of misaligned deployment, inaccurate decision support, and heightened vulnerability to external security threats. Transparency, accountability, and human oversight are also constrained by the region's limited technical maturity. Many public institutions lack the tools needed to assess algorithmic behaviour or verify system integrity. Imported AI systems often arrive as closed products that cannot be audited or modified. This shapes public dependence on external providers and weakens national oversight capabilities. The absence of context‑specific transparency frameworks further complicates accountability, especially where legal systems are overstretched. The social, economic, ethical, cultural, linguistic, and technical implications of AI are particularly pronounced. AI systems trained on non‑African data often fail in local environments, generating biased outputs that disadvantage rural communities, minority language groups, or low‑income populations. These biases interact with existing inequalities, intensifying the digital divide documented across the region. Ethical implementation becomes difficult because many countries lack national AI strategies, strong data protection laws, or human rights enforcement mechanisms. AI capacity‑building remains the most significant opportunity and the most significant challenge. The region's young population provides a strong foundation for developing AI skills, local innovation, and indigenous AI research hubs. However, the scale of human capital gaps, especially in populous countries, requires long‑term investment. Without this investment, governance reforms will struggle to materialize. Despite these challenges, the region has clear opportunities to define its own governance path through targeted capacity‑building, localized AI ecosystems, and regional cooperation driven by the African Union or the UN but tailored to the environment for a result with positive impact.

The AI Dialogue can play a critical role in advancing international cooperation by creating a structured space where fragmented global efforts can be aligned, where governance asymmetries can be addressed, and where developing regions can participate as equal stakeholders. Current governance arrangements are dispersed across national frameworks, regional strategies, and sector‑specific initiatives. This fragmentation creates uneven standards, duplication of effort, and regulatory gaps that allow harmful practices to persist. The Dialogue can act as a coordination mechanism that links existing frameworks, reduces overlap, and promotes interoperable principles that respect regional diversity while maintaining global coherence. The Dialogue can also enable meaningful participation for regions such as Sub‑Saharan Africa, where limited institutional capacity and digital infrastructure restrict engagement. My papers highlight that without targeted support, the region risks a widening AI divide. The Dialogue can therefore prioritize capacity‑building partnerships, technical assistance programs, and resource‑sharing models that elevate underrepresented regions and reduce global asymmetry. Another essential role involves promoting transparent procurement practices, which are often neglected in current governance discussions. International cooperation must include shared procurement standards that ensure auditability, safeguard public accountability, and prevent dependency on opaque systems controlled by a small number of global actors. Collaborative procurement guidance can strengthen national oversight capacities and reduce structural vulnerabilities. Finally, the Dialogue can support joint research on emerging risks such as dual‑use science, environmental impacts, and compute concentration. These risks transcend borders and require collective intelligence rather than isolated national responses. By fostering shared research agendas and cross‑regional knowledge networks, the Dialogue can accelerate responsible global governance and ensure that AI development contributes to security, equity, and sustainable development.

The AI Dialogue can play a unifying role by connecting with established global initiatives that already provide credible foundations for ethics, governance, and capacity building. UNESCO's Recommendation on the Ethics of Artificial Intelligence and its RAM exercise offer a common ethical baseline and a structured readiness assessment method that supports national self evaluation and gap identification. The ITU's AI for Good capacity building programs complement this by strengthening institutional skills, digital literacy, and technical competence in regions with limited readiness. The Dialogue can expand these initiatives by promoting coordinated adoption, shared indicators, and regional implementation strategies. International technology companies also operate AI for Good labs that contribute early warning systems, climate tools, and inclusive digital applications. Their technical capabilities are valuable, although concerns about transparency and dependency persist. The Dialogue can provide governance guidance that ensures such collaborations follow accountable and human centered principles. To enhance global coherence, the AI Dialogue should integrate outputs from the earlier United Nations AI for Humanity scientific advisory report and expand engagement with credible non government actors including the Center for AI and Digital Policy, the Alan Turing Institute, the Ada Lovelace Institute, the OECD AI Observatory, NEPAD, and regional African and Asian think tanks. The Dialogue should explicitly link with critical global governance processes such as the G7 Hiroshima Process, the G20 AI Principles, the AI Action Summit in Paris, the AI Seoul Summit, China's Generative AI Services Law, the African Union's emerging AI strategies, the ISO and IEC standards, the IEEE ethics initiatives, and the NIST AI Risk Management Framework. It should also align with the EU AI Act updates, the UN General Assembly AI Resolution, the Global Partnership on AI, the WEF AI Governance Alliance, the Rome Call for AI Ethics, the Bletchley Declaration, the Doha Declaration, the BRICS statement, and the Montevideo and Asilomar principles. The added value of the AI Dialogue lies in its ability to unify these dispersed efforts into a coherent coordination mechanism that promotes shared standards, equitable capacity building, and global trust.

Different stakeholders can contribute meaningfully to the AI Dialogue when its structure supports continuous, transparent, and inclusive engagement rather than one‑time consultations. Governments can provide regulatory perspectives, national risk assessments, and policy priorities that reflect domestic capabilities. International Organisations can supply normative frameworks, technical standards, and comparative benchmarking. Academia can contribute independent evidence, interdisciplinary research, and foresight analysis. Civil society can highlight rights based concerns, social impacts, and community level risks. Private sector actors can share technical insights, safety practices, and implementation challenges. Regional institutions can bring contextual knowledge that ensures global governance remains relevant across diverse environments. To enable high quality participation, the Dialogue should adopt a multi stage consultation structure that invites contributions throughout the process, not only during the initial call. This includes structured online submissions, in person and virtual workshops, regional roundtables, and focused expert groups that address specific thematic areas. Mid process engagement should become a formal requirement. The Dialogue should request concise input at key milestones so stakeholders can react to emerging drafts and evolving proposals. A realistic public voice mechanism is essential. Stakeholders should not be limited to three minute video messages or one time surveys. Instead, they should receive periodic invitations to comment on draft outputs through online platforms, moderated forums, and virtual hearings. Focus groups composed of marginalized communities, youth networks, and practitioners from low resource regions should be integrated into the process. Input must also be acknowledged. When the previous scientific advisory board received submissions from more than one hundred contributors, ALL individuals were not visibly recognized. A simple mechanism such as a stakeholder contribution list or a public thank you notice in the final publication builds trust and signals respect. Email acknowledgments should be standard. A structure built on continuous participation, multi format engagement, and visible recognition will increase legitimacy and encourage more detailed and meaningful contributions across the AI ecosystem.

Global discussions on AI governance consistently underrepresent communities that experience the deepest structural disadvantages. Research on the digital divide, inclusiveness, and uneven AI readiness across Sub‑Saharan Africa shows that rural populations, low-income communities, and regions with limited digital infrastructure rarely influence global norm setting processes. Their exclusion affects how global standards are framed because many governance debates assume baseline levels of connectivity, institutional stability, and digital literacy that do not exist in large parts of the world. Technical governance processes frequently lack input from minority language groups as well. AI systems trained on dominant global languages do not reflect diverse linguistic realities, which reinforces cultural and knowledge imbalances. Public sector institutions in developing regions also remain underrepresented. Many countries without national AI strategies or robust regulatory bodies struggle to participate in global forums due to resource constraints and limited technical capacity. Without their involvement, global governance frameworks risk becoming misaligned with the operational realities of low resource environments. Civil society organizations from the Global South, especially those focused on human rights, gender equity, disability inclusion, and indigenous knowledge systems, are often present but not meaningfully integrated into decision making. To address these gaps, the AI Dialogue should adopt an inclusion architecture that goes beyond symbolic participation. This includes establishing regional consultation hubs across Africa, Latin America, the Caribbean, and South Asia where stakeholders can contribute in their own contexts. Multi language engagement platforms should be used to ensure that linguistic diversity is not treated as an afterthought. Dedicated funding streams should support participation by organizations and experts from low resource countries. Indigenous communities and traditional knowledge holders should be invited to co design thematic inputs related to ethics, cultural preservation, and environmental impacts. Finally, stakeholder contributions must be acknowledged through transparent listing mechanisms, summary reports, and follow up engagement so communities can see how their input shapes outcomes.

1. Continual engagement cycles: The Dialogue should operate through regular consultation rounds rather than single‑phase input collection. This creates predictable opportunities for stakeholders to contribute throughout the process. 2. Open virtual collaboration tools: Utilizing platforms such as Teams or Slack enables ongoing participation. These tools allow stakeholders to submit ideas, review drafts, raise concerns, and exchange resources without requiring formal meetings. 3. Multi channel thematic spaces: Creating channels within Teams or Slack for thematic areas and regional clusters allows focused engagement. Participants can join channels aligned with their expertise or interests, which increases relevance and reduces noise. 4. Clear requests for concise contributions: Each engagement stage should provide a precise action request such as a short-written input, targeted feedback on draft recommendations, or a structured poll. This improves quality and reduces fatigue. 5. Rotating stakeholder panels: Panels composed of civil society groups, academia, industry, and governments should rotate quarterly. This ensures representation shifts across regions and sectors and prevents dominance by a single group. 6. Regional community observatories: Local hubs hosted by universities or policy institutes can coordinate community feedback, track impacts, and collect insights from rural or underserved areas. 7. Scenario simulation workshops: Interactive workshops can allow stakeholders to test governance scenarios, debate tradeoffs, and identify blind spots. 8. Open method of coordination forums: Borrowed from EU governance practice, these forums allow participants to compare national approaches, share best practices, and co-develop voluntary guidance. 9. Public annotation of draft documents: Using collaborative document tools, stakeholders can annotate drafts, offer line by line suggestions, and identify omissions. 10. Formal recognition of contributions: The Dialogue should acknowledge all stakeholder inputs through public contributor lists, email acknowledgments, and inclusion in annexes. This builds trust and encourages meaningful participation.

Several existing policies and governance approaches offer practical solutions for strengthening responsible AI development and can guide the next phase of international cooperation. UNESCO's Recommendation on the Ethics of Artificial Intelligence provides a globally endorsed ethical framework supported by its RAM assessment tool, which enables countries to evaluate readiness, identify governance gaps, and plan capacity building in a structured way. The ITU's AI for Good ecosystem offers another effective model because it links capacity building, digital literacy, sector specific use cases, and global technical expertise through an accessible and collaborative platform. The OECD AI Principles, the EU AI Act, the NIST AI Risk Management Framework, and the ISO and IEC technical standards provide practical regulatory and technical baselines that can be locally adapted. These frameworks introduce clear requirements for transparency, risk management, documentation, and safety assurance. They also support interoperability across jurisdictions, which reduces fragmentation and strengthens accountability. At the regional level, the African Union's digital transformation strategies and emerging continental AI initiatives illustrate how regions with limited resources can develop governance systems aligned with local development priorities. These approaches emphasize capacity building, infrastructure investment, and inclusive policy making. Non government actors such as the Center for AI and Digital Policy, the Ada Lovelace Institute, the Alan Turing Institute, and the Global Partnership on AI provide research, evaluation tools, and policy resources that strengthen evidence based decision making. Technology companies also contribute through AI for Good labs that support innovation in climate adaptation, health, and accessibility, although governance guidance is needed to ensure transparency and avoid dependency. Together, these policies and platforms offer realistic pathways for safe, inclusive, and accountable AI governance.