African Union

Most technology governance forums follow a familiar script: wealthy nations write the rules, then invite everyone else to endorse them. Success here would mean breaking that pattern decisively. With more than 100 countries, particularly from the Global South, expected to participate, the conditions for genuine inclusion exist on paper. African delegations should not merely be present; they should be co-authoring frameworks, not ratifying ones written elsewhere. The continent has a direct stake in this. AI systems are already shaping credit access, healthcare triage, and public security across Africa, often built on assumptions imported from elsewhere, and the costs of getting it wrong fall on people with the least recourse. The stated ambitions include promoting interoperability between governance regimes and anchoring AI in international law and human rights. Those goals only mean something if they produce specific, time-bound commitments rather than elegant declarations. Africa has seen enough summits end in communiques. What this Dialogue owes the continent, and the world, is a governance architecture that treats every nation as a rule-maker, not an afterthought.

• Safe, secure and trustworthy AI
• Protection and promotion of human rights
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Interoperability of governance approaches

These four priorities are deeply interconnected and reflect the realities of AI deployment across Africa today. Safe, secure and trustworthy AI is the foundation everything else rests on. AI systems that are unreliable, biased, or opaque cause the most damage in contexts where affected communities have limited means to challenge or seek redress. Getting safety right is not a luxury; it is a prerequisite. The social, economic, ethical, cultural, linguistic and technical implications of AI matter enormously for a continent as diverse as Africa. Many AI systems deployed here were not designed with African languages, cultural contexts, or economic realities in mind. Governance that ignores these dimensions risks entrenching existing inequalities rather than addressing them. Interoperability of governance approaches is a practical necessity. Africa cannot afford a fragmented landscape where different regulatory regimes create barriers to participation or force countries to simply adopt frameworks designed elsewhere. Building bridges between governance systems ensures African nations can engage on equal terms rather than being locked into standards set without them. Finally, the protection and promotion of human rights ties all three together. AI is already influencing access to credit, movement, employment, and public services across the continent. Without explicit human rights guardrails, efficiency gains from AI can come at the cost of dignity and fairness for the most vulnerable. This priority ensures that people, not systems, remain at the centre of any governance framework that emerges from this Dialogue.

One significant gap is "data sovereignty and ownership". The listed themes touch on implications and rights, but none directly address who owns the data that AI systems are trained on. Africa generates vast amounts of data that flows to external platforms and is used to build commercial AI products from which the continent sees little return. Without explicit governance around data ownership, consent, and benefit-sharing, African nations risk becoming raw material suppliers in an AI economy they do not control. This cuts across safety, rights, and interoperability but is distinct enough to deserve its own dedicated focus in the Dialogue.

West Africa's tech and innovation sector is growing rapidly, yet AI governance gaps are creating real friction. Startups and developers across the region are building on fragmented regulatory foundations, navigating between competing external frameworks like the EU AI Act and emerging African Union guidance, with little coherent regional direction. This creates uncertainty that slows investment and pushes promising ventures toward jurisdictions with clearer rules. The absence of safety and rights standards is particularly felt in healthtech and the creative industries, two sectors where AI adoption is accelerating fastest across West Africa. Tools are making consequential decisions about medical triage while AI-generated content is increasingly displacing local artists and creators, often with no transparency, rights protection, or accountability mechanisms in place. The linguistic gap remains a stubborn structural problem. West Africa's extraordinary language diversity is almost entirely absent from mainstream AI systems, meaning the most innovative local solutions struggle to serve local populations effectively. The opportunity is significant though. The region has the demographic energy, the entrepreneurial culture, and increasingly the technical talent to engage meaningfully in shaping AI development. Nigeria's 2023 Data Protection Act offers a building block. With stronger regional coordination through ECOWAS and genuine participation in forums like this Dialogue, West Africa has a real chance to contribute to governance frameworks that actually reflect its realities, rather than simply adapting to ones that do not.

Few multilateral spaces exist where countries at different stages of AI development can negotiate on genuinely equal terms. This is where the Dialogue can make a real difference. Most existing forums are shaped by nations with mature AI industries and established regulatory positions. A UN-convened process carries the legitimacy needed to bring everyone to the table, including those historically excluded from standard-setting conversations. Practically, it can build shared definitions and principles that make interoperability between regional governance frameworks more achievable, reducing the fragmentation that currently burdens developers and regulators across regions like West Africa. It can also serve as a knowledge-sharing mechanism, helping countries with less regulatory capacity learn from others without simply importing frameworks that do not fit their contexts. The measure of its success will be whether it produces durable, inclusive agreements rather than consensus documents that reflect only the priorities of the most powerful participants.

Several African-led initiatives provide a strong foundation. The AU's Continental AI Strategy, the AU Data Policy Framework, the African Digital Compact, and the Smart Africa Alliance collectively represent years of consultative work translating African priorities into actionable governance principles. At the regional level, ECOWAS digital economy frameworks add another relevant layer. The Dialogue's added value is convening power. It can elevate these existing frameworks into binding international conversations and create accountability structures that regional strategies alone cannot enforce. With Africa accounting for just 1% of global AI compute capacity, continental strategies need the political weight and resourcing that only a UN-convened process can mobilise. The goal should be coherence, ensuring African-led initiatives genuinely shape global standards rather than being absorbed into frameworks built without them.

Meaningful participation requires more than an open invitation. Governments, civil society, the private sector, and research institutions each bring perspectives that the Dialogue cannot afford to ignore. For African contexts specifically, grassroots digital rights organisations should have structured, not tokenistic, access to the process. On format, the Dialogue should move beyond plenary declarations toward working groups organised around specific thematic areas where real negotiation can happen. Hybrid participation is essential to ensure contributors from regions with limited travel budgets are not structurally excluded, and regional consultations should precede sessions so that positions arriving in Geneva reflect genuine constituencies. The structure should be designed so that expertise and preparation carry more weight than diplomatic influence, creating genuine space for smaller nations and non-state actors to shape outcomes rather than simply observe them.

African and Global South communities are the most consequentially underrepresented, particularly given that AI systems are being deployed at scale across these regions with minimal input from the people most affected by them. Beyond geography, women, persons with disabilities, indigenous communities, and informal sector workers rarely have structured access to these conversations despite bearing disproportionate risks from poorly governed AI. Inclusion requires more than inviting these groups to existing forums. It means funding their meaningful participation, holding regional consultations that feed directly into global deliberations, and designing processes where their input shapes outcomes rather than decorates them. The Dialogue should establish clear mechanisms for civil society and community representatives to engage at every stage, not just during open comment periods.

Beyond traditional panel discussions, asynchronous digital participation tools would allow contributors across time zones and resource constraints to engage substantively between sessions, broadening participation beyond those who can physically attend. Dedicated constituency caucuses, where civil society, the private sector, and technical communities develop joint positions before plenary sessions, would ensure diverse voices arrive with coherent, well-prepared input rather than being overwhelmed by the pace of formal proceedings. Embedding real-time translation and accessible formats across all engagement channels would further ensure that participation is genuinely inclusive rather than nominally so.

Several examples offer useful models, such as the EU AI Act, which, despite its limitations for non-European contexts, established a risk-tiered regulatory approach that has become a practical reference point for countries developing their own frameworks. Singapore's Model AI Governance Framework offers a more flexible, principles-based alternative that smaller economies have found easier to adapt. On the African continent, Nigeria's 2023 Data Protection Act demonstrates that domestic regulatory progress is achievable and provides a foundation for AI-specific legislation. Rwanda has moved further, actively integrating AI governance into its national digital strategy in a way that links regulation directly to development outcomes. At the multilateral level, UNESCO's Recommendation on the Ethics of AI remains one of the few globally agreed normative frameworks and should serve as a baseline rather than a ceiling for the Dialogue's ambitions. The NIST AI Risk Management Framework offers practical technical guidance that policymakers with limited regulatory capacity have found accessible. What these examples share is clarity, adaptability, and grounding in specific contexts. The Dialogue should draw on all of them as a baseline while being careful not to simply globalise any single model for the Dialogue's ambitions.