AFRICA AI POLICY LAB (a Lawyers Hub Africa initiative)

Success in Geneva is structural, not performative. The first Dialogue should deliver four concrete outcomes. First, a public compendium of AI governance approaches drawn from every region, not only the European Union, the United States, and China. Africa AI Policy Lab already tracks national AI strategies across all 54 African countries; a comparable global mapping anchored at the UN would be a foundational reference for the Independent International Scientific Panel on AI and for Member States designing first-generation regulation. Second, a stakeholder accountability framework that names what genuine inclusion of the Global South looks like in operational terms: structured drafting roles for civil society, the technical community, academia, and legal practitioners from underrepresented regions, with measurable participation indicators for the May 2027 New York session. Third, an interoperability principles statement establishing how compatibility across governance approaches can be achieved through equivalence, not imposition. African states are repeatedly asked to align with rules drafted in jurisdictions whose enforcement capacity, language ecosystems, and infrastructure realities differ fundamentally. The Dialogue should commission ODET and the Scientific Panel to produce a baseline interoperability map covering the EU AI Act, the Council of Europe Framework Convention, the OECD Principles, the AU Continental AI Strategy, and major national instruments. Fourth, a Scientific Panel agenda that includes evidence streams from researchers and practitioners building AI for the contexts the majority of the world inhabits: informal economies, multilingual societies, infrastructure-constrained environments, and resource-limited regulators. The Dialogue will be judged not by the eloquence of its closing statement but by whether the people who shape it look like the people AI will affect. With less than one percent of the world's data centres in Africa and roughly three percent of global AI talent there, the Dialogue cannot succeed if it reproduces the asymmetries it was created to correct.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Interoperability of governance approaches
• Protection and promotion of human rights
• AI capacity-building

The seven thematic areas are operationally interdependent. AI Policy Africa, working across all 54 African jurisdictions, treats them as a single integrated agenda where rights and oversight are protected by capacity, context-fit, interoperable enforcement, and openness, not by aspiration. Capacity-building is the equity mechanism for the entire architecture. Africa holds roughly three percent of global AI talent. Togo's 50,000-trainee programme, Kenya's KAISA at 600,000 trainees, and Egypt's 5,000 annual professionals remain far below the scale a continent of 1.4 billion people requires. Without a critical mass of African judges, regulators, parliamentary drafters, litigators, and engineers, every other commitment becomes performative. The implications cluster (social, economic, ethical, cultural, linguistic, technical) matters because frontier models have negligible training-data coverage of the 1,000 to 2,000 African languages spoken by hundreds of millions. Frameworks that ignore linguistic, infrastructural, and economic context produce regulation that protects only those already online. Interoperability is the political pivot. Our research across Kenya, South Africa, Egypt, Morocco, Rwanda, and Nigeria shows that African regulators are repeatedly asked to align with frameworks they did not draft. Interoperability should be reciprocal and equivalence-based. Open-source, open data, and open AI models are the structural counter-weight to dependency. Masakhane, Egypt's Karnak LLM, and the JAZARI ROOT Institute show that sovereignty-respecting AI is feasible when openness is treated as infrastructure. Human rights protection is the frame within which AI governance must sit, anchored in the African Charter and the Malabo Convention. Transparency, accountability, and human oversight matter most where state capacity is uneven and public-sector AI deployment is scaling faster than the institutions that govern it. Safety, security, and trustworthiness must be defined in contexts the majority of the world inhabits, not assumed from frontier-firm conditions. The four priorities operationalize the other three.

Four cross-cutting issues are critical and inadequately captured by the listed themes. First, AI infrastructure sovereignty. Less than one percent of the world's data centres are in Africa. Recent investments such as Equinix's USD 438 million commitment in South Africa are welcome, but no mechanism yet ensures that private compute investment serves AI sovereignty rather than deepens dependency. The Dialogue should treat compute, data centres, and submarine cable governance as a first-order AI governance question. Second, the political economy of AI partnerships. The AUC-Google MoU, Microsoft's USD 50 billion commitment to emerging markets, and the documented USD 2 billion Chinese surveillance footprint across 11 African states are reshaping AI governance faster than any treaty. The Dialogue should establish principles for evaluating large bilateral and corporate AI deals, including data sovereignty audits, exit clauses, technology transfer obligations, and revenue transparency. Third, enforcement. AI Policy Africa's Africa AI Privacy Report 2025 documents that across the continent, laws are increasingly comprehensive on paper while data protection authorities operate with five to ten staff and no budget to audit multinationals. The Dialogue should commit to a workstream on enforcement cooperation: cross-border investigation protocols, shared enforcement funds, and mutual recognition so that a Kenyan fintech certified at home does not need to recertify from scratch in Ghana or Senegal. Fourth, AI in the public sector and the rule of law. AI is increasingly embedded in functions of the state itself: judicial case management, social protection eligibility, immigration triage, policing, and electoral administration. These deployments touch directly on due process, equality before the law, and the legitimacy of public authority. Synthetic media now accounts for roughly 40 percent of rejected biometric verifications in Southern Africa, an electoral and democratic-integrity issue that warrants a dedicated thematic brief from the Scientific Panel.

Across Africa the governance gaps in our priority areas translate into concrete daily losses and possibilities. On capacity, the brain-drain paradox is acute. Africa has the world's youngest population yet roughly three percent of global AI talent. Eighty-three percent of African AI startup funding in Q1 2025 flowed to four countries (Kenya, Nigeria, South Africa, Egypt), deepening internal concentration. The opportunity is real: Kenya's National AI Strategy 2025-2030, Egypt's National AI Strategy 2021-2026 with the Karnak sovereign LLM, Morocco's AI Made in Morocco roadmap, and Rwanda's AI ethics guidelines show that domestic capacity, when built, produces governance that fits. On implications, the linguistic dimension is consequential. Major LLMs have negligible coverage of 1,000 to 2,000 African languages. Citizens cannot exercise rights, access services, or scrutinize automated decisions in tools that do not understand them. Masakhane and the Huniki Federation are responses; they need international legitimacy and resourcing. On rights and oversight, the African Charter and the Malabo Convention offer continent-specific normative resources that international AI governance has under-utilized. Few African data protection authorities have AI-specific competencies; few national human rights institutions have technologists on staff; few courts have heard substantive AI-related matters. Public sector AI procurement rarely requires algorithmic impact assessments, and biometric systems are deployed faster than independent audit can keep pace. On interoperability, the GDPR-equivalence pattern is illustrative. Many African data laws were drafted in GDPR's image, not because GDPR fits, but because donors point states toward the model they know. Compliance costs designed for Google fall hardest on the 12-person teams that constitute Africa's actual AI ecosystem. On openness, approximately 70 percent of open-source research data lacks clear IP documentation. The AU Continental AI Strategy and Kigali Africa Declaration on AI provide political scaffolding; they need international partners that respect rather than displace them.

The Dialogue's added value is to be the only universal venue at the UN where AI governance is shaped by all Member States and stakeholder constituencies on equal footing. Four roles flow from that uniqueness. First, it can be a translation layer between fragmented governance regimes. Today, rules emerging from the EU AI Act, US executive orders and state laws, China's algorithmic and generative AI regulations, the G7 Hiroshima Process, the OECD AI Principles, the Council of Europe Framework Convention on AI and Human Rights, UNESCO's Recommendation on the Ethics of AI, and the AU Continental AI Strategy interact unevenly. The Dialogue can produce shared vocabularies, equivalence maps, and cross-walking tools that allow compatibility without coercion. Second, it can be a legitimacy multiplier for the Scientific Panel's findings. The Panel's annual report and thematic briefs require a political venue where Member States and stakeholders engage with them substantively. Without the Dialogue, evidence has no destination. Without evidence, the Dialogue has no anchor. Third, it can be a coordinated capacity-building accelerator. ITU's AI for Good summits, UNESCO's Recommendation implementation work, ODET's open-source and DPI tracks, and bilateral programmes are not currently coordinated. The Dialogue can convene a capacity-building compact aligned with the Secretary-General's report on innovative voluntary financing options for AI capacity-building, with the proposed Global Fund on AI as the longer-term destination. Fourth, it can rebalance the geography of agenda-setting. Anchored to the General Assembly, the Dialogue gives African, Asian, Latin American, Caribbean, and Pacific Member States structural standing in the agenda itself, alongside their stakeholder communities. A precondition: the Dialogue must speak with rather than past existing regional bodies. The African Union, ASEAN, CARICOM, and the GCC each have AI tracks; the Dialogue's role is to amplify and connect them, not duplicate them.

UN system: UNESCO's Recommendation on the Ethics of AI and its Readiness Assessment Methodology, piloted across multiple African states; the ITU's AI for Good Summit (the Geneva session is back-to-back with it for good reason); ODET's Universal DPI Safeguards Framework and Open Source Week; the Independent International Scientific Panel on AI; the High-level Advisory Body's Governing AI for Humanity recommendations; OHCHR's B-Tech project; the IGF and its Coalition on Data and AI Governance. Intergovernmental and regional: the African Union Continental AI Strategy and the Kigali Africa Declaration signed by 49 African countries with the USD 60 billion Africa AI Fund commitment; ECOWAS, EAC, and SADC digital strategies; the AU-EU Digital Partnership; the BRICS Leaders' Statement on the Global Governance of AI; the OECD AI Policy Observatory and Network of AI Experts; the GPAI; the Hiroshima Process; the Council of Europe Framework Convention on AI and Human Rights, Democracy, and the Rule of Law (the first binding international AI treaty open to non-European parties). Multi-stakeholder and technical: the Frontier Model Forum and AI Safety Institutes; civil-society networks including Article 19, Access Now, AlgorithmWatch, Pollicy, Research ICT Africa, and Datasphere; technical networks including Masakhane, Zindi, Huniki Federation, Lacuna Fund, the Equal Compute Network; standards bodies including ISO/IEC JTC 1 SC 42, IEEE, and ETSI; the Network of African Data Protection Authorities and the Global Network of National Human Rights Institutions. The Dialogue's added value is universality. None of the above is universal. The risk is reproducing existing asymmetries through procedural defaults. The mitigation is participation support, written-input parity, and standing liaison arrangements with the African Union and other regional bodies, with concrete deliverables timed to the 2027 Compact review. The Dialogue should also formally recognise and integrate regional convenings as structured input mechanisms. For example, the proposed Africa AI Governance Summit led by AI Policy Africa (Lawyers Hub) can serve as a dedicated platform to consolidate African stakeholder inputs, facilitate policy co-creation, and feed evidence-based recommendations into the Dialogue process. Such side events would strengthen regional coordination while ensuring continuous, grounded contributions between formal sessions.

Stakeholders should contribute through structured roles, not symbolic participation. We recommend the following format. Pre-Dialogue (now to July 2026): structured written-input cycles with all submissions published in full. Regional consultations co-hosted with regional bodies (the AU AI Working Group, ASEAN Digital Ministers, OAS digital ministerial track). Co-creation drafting sprints in which civil society, the technical community, academia, and legal practitioners are paired with Member State delegations to draft thematic position papers, not commentary on State drafts. Geneva Session (6 to 7 July 2026): a plenary anchored by the Scientific Panel's inaugural report, followed by working tracks corresponding to the four thematic clusters. Each track should include at minimum one government, one civil society, one technical community, one academic, and one private-sector lead, with paired developed and developing country chairs. Deliverable per track: a one-page outcome statement plus a six-month workplan. Between sessions (July 2026 to May 2027): permanent multi-stakeholder working groups (not ad hoc) on the four thematic clusters, meeting publicly online, with rotating co-chairs and published minutes. Open observer status for any registered stakeholder. A practitioner stream comprising regulators, judges, data protection authorities, ombudspersons, and technical auditors should be institutionalized alongside the policy and academic streams. New York Session (May 2027): outcomes feed directly into the high-level review of the Global Digital Compact at the 82nd session of the General Assembly. Working group reports become the substantive basis, not background documents. Four structural rules should govern format. First, no closed sessions on substantive AI policy questions. Second, speaking-time parity between developed and developing country interventions, monitored publicly. Third, all background documents in the six UN languages plus at minimum one African, one Asian, and one Latin American working language per session. Fourth, every working group at least 50 percent representation from Africa, Asia-Pacific, and Latin America combined.

Seven communities are systematically underrepresented in global AI governance and require deliberate inclusion mechanisms. First, African civil society and technical community voices. Despite tracking AI governance across all 54 African countries, the continent's research and advocacy organizations are routinely consulted late, briefly, and as box-checking exercises. Inclusion requires structured drafting roles in working groups, not panel slots. Second, African legal practitioners and judiciaries. Across 54 jurisdictions and over 1.4 billion people, African judges, magistrates, prosecutors, and litigators are rarely present in international AI governance fora. The rule-of-law dimensions of AI (due process, evidentiary standards, contestability, judicial review of automated decisions) cannot be addressed without them. Mechanism: structured engagement with the African Court, regional judicial training institutes, and national judiciaries through the AU. Third, indigenous peoples and language communities. The 1,000 to 2,000 languages of Africa, the 700+ of Asia-Pacific, and indigenous languages of the Americas are largely absent from frontier AI development. The Dialogue should embed indigenous data governance leadership (CARE principles, Te Mana Raraunga model) in each thematic cluster. Fourth, women in AI from the Global South. Despite the Scientific Panel's 19/21 gender balance, the broader Dialogue ecosystem still skews male and Northern. Women-in-AI networks across Africa, Asia-Pacific, and Latin America need dedicated working-group seats. Fifth, small-island developing States and least-developed countries, whose challenges (climate-AI intersections, sovereignty, micro-economies) differ structurally from larger developing economies. Sixth, frontline workers reshaped by AI: gig workers, content moderators, agricultural workers, and informal-sector workers. Their experience is the empirical ground truth for impact assessments. Seventh, children and youth. The AU Child Online Safety and Empowerment Policy and the Children Global AI Summit on Africa show youth-led contributions are substantive. Inclusion mechanisms: participation support, remote co-chair authority, interpretation in regional working languages, honoraria for civil-society and indigenous experts, and structured rapporteur roles.

The Dialogue should adopt formats that produce policy artefacts, not panel readouts. Africa AI Policy Lab proposes seven; Policy hackathons. Two-day cross-regional teams (government, civil society, technical, academic, private sector, legal practitioners) draft regulatory sandbox terms or model clauses on a defined problem, for example biometric AI in financial inclusion. Outputs become draft Dialogue annexes. Regulatory sandbox showcases. Each session features two to three live national sandboxes (Mauritius's RAEAL AI licence, Kenya's CMA fintech sandbox, Rwanda's regulatory framework, Singapore's AI Verify, the United Kingdom's ICO sandbox) with regulators presenting what worked, what failed, and what they would change. Practitioner evidence, not principle statements. Live evidence panels with the Scientific Panel. Researchers present preliminary findings; Member States and stakeholders pose questions in real time. The Panel commits to incorporating questions in its next thematic brief. Reverse hearings. Civil society, technical community, and legal practitioner representatives question Member State delegations on national implementation, rather than the customary inverse. Outputs are formally noted in session reports. Case clinics on AI incidents. Africa AI Policy Lab maintains an Africa AI Incidents Tracker. Each session takes one or two real incidents (Worldcoin Kenya, the Chinese surveillance footprint, biometric fraud cases, algorithmic benefits denial) and works through them as governance case studies, with affected communities, litigators, and regulators in the room. Cross-regional working pairings. Each working group pairs a developed and a developing country team to co-draft. The pairing is permanent for the year and produces joint outputs, not parallel statements. African Languages AI demo track. Live demonstrations of African, Asian, and Latin American language AI projects, paired with policy discussions on data governance, IP, and capacity. Demonstrations grounded in the realities the rest of the Dialogue is debating. The common principle: every format should produce a dated, attributable, citable artefact. Anything else is performance.

Africa AI Policy Lab offers concrete examples drawn from continental research and global comparison. National policy: Kenya's National AI Strategy 2025-2030 with its cross-sector implementation roadmap; Egypt's National AI Strategy 2021-2026 paired with the National Council for AI and the Karnak sovereign LLM; Morocco's AI Made in Morocco roadmap and the JAZARI ROOT Institute; Rwanda's Guidelines on Ethical Development and Implementation of AI; South Africa's National AI Policy Framework; Algeria's December 2024 AI strategy; Nigeria's National AI Strategy. These first-generation African frameworks merit comparative study. Continental architecture: the AU Continental AI Strategy (2024); the Kigali Africa Declaration on AI signed by 49 African countries with the USD 60 billion Africa AI Fund and the Africa AI Council; the AU Data Policy Framework; the AU Child Online Safety and Empowerment Policy; the Network of African Data Protection Authorities as a cross-border regulatory cooperation model. Regulatory innovation: Mauritius's RAEAL AI licence, the first dedicated AI regulatory licence in Africa for financial services; Kenya's Worldcoin enforcement action setting a global precedent on biometric data extraction; Morocco's CNDP facial-recognition enforcement; Nigeria's NDPC DPIA requirements. Soft-law and treaty templates: UNESCO's Recommendation on the Ethics of AI with its Readiness Assessment Methodology engaged by over fifty countries; the Council of Europe Framework Convention on AI, the first binding international AI treaty open to non-European parties; the OECD AI Principles and Policy Observatory. Multi-stakeholder convening models: the Africa Law Tech Festival convening 2,500 stakeholders annually; the Africa AI Policy Summer School with the University of Cape Town; the Africa AI Research Colloquium with Qhala, Carnegie Endowment for International Peace, Carnegie Mellon University, and Masakhane. Strategic litigation: cases in Kenya, South Africa, India, the Netherlands, and the United Kingdom show courts can interrogate algorithmic decision-making when claimants are equipped. Africa AI Policy Lab research outputs are publicly available at aipolicy.africa.