Africa Tech for Development Initiative

Outcomes That Would Make the First Global Dialogue on AI Governance a Success A successful Global Dialogue on AI Governance should produce concrete, actionable outcomes across the following priority areas: 1. International Legal Framework Advance broad global support for an International AI Treaty that enshrines binding commitments, including explicit prohibitions on AI systems that undermine human rights, democratic values, and fundamental freedoms. 2. Equitable Infrastructure Investment Encourage member nations and multilateral institutions to invest in foundational AI infrastructure including Internet access, computational capacity, localized data centers, green energy, and reliable electricity to ensure no nation is structurally excluded from the AI era. 3. Inclusive and Culturally Relevant AI Development Champion the development of AI systems that are interoperable, open-source, context-aware, and hyper-local, culturally and linguistically relevant, with robust capacity-building programmes supporting Global South nations in developing and deploying AI on their own terms. 4. Human Oversight Across the AI Lifecycle Charge all participating stakeholders with embedding meaningful human oversight at every stage of the AI lifecycle, from design and development through to deployment and decommissioning. 5. Implementation and Enforcement of Governance Frameworks Accelerate the harmonised implementation and enforcement of existing AI governance instruments, including the EU AI Act, the UNESCO Recommendation on the Ethics of AI, and the OECD AI Principles. 6. Algorithmic Transparency and Accountability Mandate algorithmic transparency throughout the AI lifecycle and establish accessible mechanisms for individuals and communities to contest harmful or questionable AI-generated outcomes. 7. Liability and Accountability Rules Establish clear, enforceable liability rules applicable to AI developers, deployers, and users, ensuring responsibility is allocated proportionately across the value chain. Collectively, these outcomes would transform the Dialogue from a forum of aspiration into a catalyst for governance with genuine global accountability.

• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI

"Safe, secure and trustworthy" AI is implicitly embedded by my recommendation for human rights and oversight priorities, but it is not explicitly emphasized in my submission. "Interoperability of governance approaches" and "Open-source software, open data and open AI models" are captured in my submission, but as enabling conditions rather than primary priorities, making them secondary to other four selected above.

Cross-Cutting and Emerging Issues Not Captured by the Listed Themes Several critical issues remain underexplored in conventional AI governance discourse, and this Dialogue presents a rare opportunity to spotlight them: The Governance Gap in AI Testing Environments One conspicuous absence is the global absence of structured AI sandbox frameworks. Sandboxes - regulatory, operational, and hybrid - provide controlled environments where AI systems can be tested, validated, and stress-tested before deployment at scale. Evidence from implementations in Thailand, India, Tanzania, and the Dominican Republic demonstrates that sandbox projects achieve a 60 to 70 percent success rate, while the remaining failures generate governance intelligence that strengthens both regulation and product design. Co-creation between public and private sectors from project inception is critical; it is the difference between AI systems that earn public trust and those that do not. This dialogue must advance a shared international framework for AI sandboxing, particularly to support Global South nations building sovereign AI capacity. A Global AI Incident Register Equally absent from mainstream governance conversations is the urgent need for a robust, multilateral AI incident register, a living repository documenting AI risks, harms, abuses, and rights violations as they occur in real time across jurisdictions. We cannot govern what we refuse to systematically record. Without this infrastructure, accountability will remain a mere hope rather than an operational reality. Digital Public Infrastructure as Governance Backbone The intersection of AI and Digital Public Infrastructure (DPI), encompassing digital identity, interoperable payments, and data exchange, demands its own governance lane. DPI is increasingly the delivery mechanism for AI-powered public services, yet its governance implications remain fragmented across siloed policy conversations. The Democratic Deficit in AI Standard-Setting Finally, who sets the standards matters as much as the standards themselves. The Global South remains structurally underrepresented in technical standard-setting bodies. Genuine inclusion is not a courtesy; it is a governance imperative.

How Governance Gaps Are Affecting Nigeria, Africa, and the AI Policy Sector Africa is not on the periphery of the AI revolution. It is at its pressure point. The African Union's Continental AI Strategy, endorsed in July 2024, is the continent's most ambitious attempt to govern AI as a tool for development rather than dependency. Yet for Nigeria, Africa's most populous nation and largest economy, the gap between policy ambition and ground reality remains stark and consequential. The Nigerian Reality Nigeria has made meaningful institutional progress through FMCIDE, NITDA, NCAIR, and the Nigeria Data Protection Commission. Yet no dedicated AI regulatory authority with binding enforcement powers currently exists. A proposed National AI Commission under the 2025 Establishment Bill remains a pipe dream, while AI deployments in fintech, healthcare, and law enforcement are already outpacing governance. The risks are documented. Research drawing on the Kano textile manufacturers' lawsuit and the Lagos "Buy Now Pay Later" scandal confirms tangible harm from exclusionary credit scoring practices. An independent audit of ten fintech models across Nigeria, Kenya, and South Africa found women-led businesses penalized with a 37 percent underfunding penalty. On surveillance, Nigeria is Africa's largest buyer of digital surveillance technology, having spent at least 2.7 billion US dollars on known contracts over a decade, with documented cases of citizens detained without trial following surveillance of their online speech. These are not prospective risks. They are present realities. The Continental Tension The AU Strategy is visionary but structurally fragile, lacking binding enforcement, cross-border coordination, and meaningful civil society inclusion. Africa's dependence on foreign foundational models and cloud infrastructure means the continent risks consuming AI shaped entirely by other people's values, identity and commercial interests. Where the Opportunities Live AI regulatory sandboxes, a continental incident register, expanded Digital Public Infrastructure, high-speed internet access, investment in computational energy, and open-source models in African languages could collectively reposition Africa from AI consumer to AI co-creator. I charge the global community to build the conditions that will make African leadership structurally possible in AI development and governance.

Its most consequential role is to serve as the connective tissue between fragmented governance efforts. The EU AI Act, the UNESCO Recommendation on the Ethics of AI, the OECD AI Principles, and the African Union Continental AI Strategy each represent serious governance thinking. Yet they operate in silos, producing regulatory inconsistency that multinational AI developers exploit and that smaller nations, particularly in the Global South, are least equipped to navigate. The Dialogue must advance genuine harmonization, not lowest-common-denominator consensus, but interoperable frameworks that respect jurisdictional sovereignty while establishing non-negotiable global floors on rights, transparency, and accountability. Second, the Dialogue must institutionalize what informal convenings cannot. This means championing a multilateral AI incident register, a living repository of documented AI harms, risks, and rights violations that transforms scattered civil society reports and court filings into binding governance intelligence. It also means advancing a shared international framework for AI regulatory sandboxes, enabling nations to test and validate AI systems in controlled, rights-respecting environments before deployment at scale. The dialogue most drive advocacy for nations across the globe to sign and enforce international AI treaties such as the EU AI Act, UNESCO Recommendation on AI Ethics, OECD AI Principles, Council of Europe AI Treaty and other relevant international instruments. Finally, and most critically, the Dialogue must structurally rebalance who governs. Global South nations, including African states with the AU Continental AI Strategy as their collective voice, must move from consultation objects to co-architects of international AI governance frameworks. The Dialogue can engender inclusion, meaningful engagement, and cooperation among global stakeholders.

Existing Initiatives the AI Dialogue Should Build Upon and the Added Value It Can Bring: Several foundational instruments demand explicit recognition and integration. The EU AI Act establishes the most comprehensive binding regulatory framework currently in force. The UNESCO Recommendation on the Ethics of AI provides a universal ethical compass grounded in human rights. The OECD AI Principles offer intergovernmental consensus on trustworthy AI. The African Union Continental AI Strategy, endorsed in July 2024, represents the Global South's most ambitious collective governance vision. The UN Secretary-General's Roadmap for Digital Cooperation and the Global Digital Compact provide the broader digital governance architecture within which AI governance must be embedded. The UN Independent International Scientific Panel on Artificial Intelligence provides evidence-based assessments of AI's societal, economic, and security impacts, offering the Dialogue a credible scientific foundation for informed policymaking. The AI for Developing Countries Forum, founded in 2023, is bridging AI accessibility gaps between developed and developing nations, and represents a natural partner for advancing equitable governance outcomes. The Center for AI and Digital Policy, a Washington DC-based nonprofit think tank, has been instrumental in shaping AI governance through democratic values, human rights, and the rule of law. Its Universal Guidelines for AI, alongside the OECD AI Principles and the UNESCO Ethics of AI Recommendation, collectively provide a rights-based governance scaffolding the Dialogue can actively champion and operationalize. Existing initiatives have laid foundations worth building on. The Dialogue's distinct added value lies in its capacity to convene these efforts under a unified multilateral mandate, generating research-driven innovations, cross-sector capacity building, inclusive brainstorming across geographies and disciplines, and actionable recommendations that translate governance ambition into measurable global impact.

The credibility of any global governance Dialogue is dependent on several factors. For the Dialogue to produce outcomes that are legitimate, durable and enforceable, it must include diverse stakeholders participation from different backgrounds. Beyond national interests, Governments must arrive as co-architects of binding commitments. They should bring draft regulatory frameworks, enforcement gap analyses, and concrete capacity building needs, particularly governments from the Global South whose lived governance realities must inform global standards rather than simply receive them. Civil society organisations and human rights bodies are the custodians of accountability. They carry documented evidence of AI harms, community-level impact assessments, and the institutional memory of what happens when governance fails. Their participation must be substantive and meaningful. The private sector must engage with genuine transparency obligations, including disclosure of known system risks, algorithmic audit findings, and incident reports. Voluntary commitments without accountability mechanisms have a poor track record. The Dialogue should establish clear expectations accordingly. Academia and independent research researchers should lead research-driven findings, bring rigorous, rights-based assessments free from commercial and political colouration. Youth and grassroots innovators, particularly from the Global South, represent both the primary users and the next generation of AI governance architects. Their contribution should is imperative to achieving the goals of the dialogue. On structure, the Dialogue should adopt a permanent secretariat, transparent working groups organized by thematic area, regional preparatory forums that feed into global sessions, and a publicly accessible AI incident register as a living accountability mechanism. Dialogue that does not structurally include those most affected by AI will not govern AI. It will merely describe it.

Africa has over 2,000 languages and 3,000 ethnic groups, each with distinct identities, traditions, and social cultures. Nigeria alone has over 550 languages and 250 ethnic groups. UNESCO describes culture as a complex whole of knowledge, beliefs, morals, and practices, further enriched by historical trade and colonial influences, with no single unified African culture. Yet the voices, identities, communities, and perspectives from Africa and other Global South regions, including Latin America and South East Asia, remain largely underrepresented and marginalized in most large language models built by major technology companies from western nations. This produces significant biases, discrimination, racism, and prejudice embedded within AI systems. In many cases, existing foundational models are not context-aware and therefore cannot adequately address the challenges faced by these nations and their communities. The voices and perspectives of older persons, who are among the most vulnerable and digitally disadvantaged, are not meaningfully involved in current AI development processes. Research from Accenture indicates that over one billion people globally live with disabilities today. Yet, existing AI systems rarely embed insights from persons with disabilities, making them inconsistent with their lived realities and needs. Women and gender-diverse communities face documented, systemic harm from biased AI systems, including unjust credit scoring penalties and facial recognition errors, yet remain structurally underrepresented in both technical development and policy governance spaces. These stakeholders must be included as co-creators of AI systems from inception, not as afterthoughts. Meaningful inclusion requires sustained capacity building initiatives, investment in open-source interoperable datasets particularly for low-resource languages, structured knowledge sharing, regional preparatory forums, dedicated seats in working groups, and governance processes intentionally designed to democratize accessibility from the outset.

Innovative Engagement Formats for a Meaningful and Dynamic AI Dialogue For the AI Dialogue to produce outcomes that are legitimate and impactful, it must itself be designed with the same principles it seeks to advance: inclusion, transparency, accountability, and contextual relevance. Deliberative Democracy Panels Rather than the traditional conference model of prepared statements delivered to passive audiences, the Dialogue should convene structured deliberative panels that bring together citizens, policymakers, technologists, and affected communities to reason through governance challenges collectively. Evidence from deliberative democracy practice consistently demonstrates that informed, diverse groups produce more nuanced and publicly legitimate policy recommendations than expert-only processes. AI Sandbox Demonstrations Live, facilitated demonstrations of AI regulatory sandboxes would transform abstract governance concepts into tangible, observable realities. Showcasing real sandbox implementations from Thailand, India, Tanzania, and the Dominican Republic would provide concrete models that delegations, particularly from the Global South, can adapt and adopt. Regional Preparatory Forums Substantive global dialogue requires substantive local preparation. Dedicated regional forums held prior to the global session, conducted in local languages and designed around regional realities, would ensure that Global South perspectives arrive at the global table already fully formed, evidenced, and impossible to marginalize. Youth and Civil Society Innovation Labs Structured innovation labs giving youth delegates, grassroots technologists, and civil society organizations dedicated time and resources to develop and present governance proposals would institutionalize bottom-up innovation rather than simply celebrate it. A Live AI Incident Registry Launching a publicly accessible, real-time AI incident register during the Dialogue itself would signal unmistakably that this is a governance process built around accountability, not aspiration.

The EU AI Act The EU AI Act represents the most comprehensive binding AI regulatory framework currently in force, establishing a risk-based classification system that imposes proportionate obligations on developers and deployers, with explicit prohibitions on AI applications that threaten fundamental rights. It sets a legislative benchmark every jurisdiction can learn from. UNESCO Recommendation on the Ethics of AI Adopted by 193 member states in 2021, this remains the broadest multilateral ethical consensus ever achieved on AI governance, providing a rights-based framework that is particularly instructive for nations building sovereign governance capacity. AI Regulatory Sandboxes As documented by Datasphere Initiative, Sandbox programmes in Thailand, India, Tanzania, and the Dominican Republic demonstrate that controlled, co-created testing environments produce safer, more contextually relevant AI systems. With a documented success rate of 60 to 70 percent, sandboxes generate governance intelligence even from projects that do not succeed. India's Digital Public Infrastructure India's DPI stack, encompassing digital identity, interoperable payments, and data exchange, demonstrates how open-source, rights-respecting infrastructure can democratise AI-powered public service delivery at population scale. Beyond DPI, the Indian AI Plan is a fantastic blueprint for building an enabling ecosystem for innovation whole ensuring that trust, safety, accountability, responsibility, transparency and sustainability are not eroded in the pursuit of efficiency. The OECD AI Principles Adopted by 46 countries, these principles provide an intergovernmental consensus framework on trustworthy AI that has meaningfully shaped national AI strategies across diverse political and economic contexts. The Center for AI and Digital Policy Universal Guidelines for AI These guidelines offer a practical, rights-based accountability framework that governments, civil society, and institutions can operationalize independently of commercial AI interests. CAIDP promotes frameworks like its Universal Guidelines for AI (2018), the OECD/G20 AI Principles (2019), and UNESCO's Ethics of AI Recommendation (2021) to ensure AI advances social inclusion and accountability. It critiques national AI policies via tools like the annual AI and Democratic Values Index published annually. It participates in high-level public voice opportunities, advocates for adherence to democratic principles, rule of law and encourages the implementation of international treaties by governments and ministries.