Case Study 2: AI Governance in the Global South — Africa's Draft Continental AI Strategy

Introduction: Who Sets the Terms?

In February 2024, the African Union Commission's Department of Infrastructure and Energy released the updated draft of the AU AI Continental Strategy for Africa. The document represented years of effort by African policymakers, researchers, and civil society organizations to articulate a distinctly African vision for AI governance — one that centers African development priorities, protects African data sovereignty, and ensures that the benefits of AI accrue to African populations rather than being extracted by foreign corporations and governments.

The strategy's development illustrated both the promise and the difficulty of genuine Global South participation in AI governance. Africa is home to 54 of the world's 195 countries and 1.4 billion of the world's 8 billion people. African economies are among the fastest-growing in the world. Africa has enormous data wealth — agricultural data, health data, financial transaction data, mobility data — that AI systems are increasingly hungry for. And Africa is increasingly the site of AI deployments, ranging from Chinese surveillance systems to Western-developed healthcare AI to homegrown fintech applications.

Yet Africa's voice in global AI governance has been systematically marginal. The OECD AI Principles were developed primarily by wealthy countries. The G7 Hiroshima Process explicitly excludes African nations. The Global Partnership on AI, with 29 members, includes only a handful of African countries and no mechanism to translate African research priorities into GPAI's working agenda. Even the UNESCO Recommendation, despite its universal membership basis, was shaped primarily by the priorities and perspectives of wealthier nations with more robust diplomatic and technical capacity. Understanding the AU AI Continental Strategy — what it proposes, what it reflects, and what it faces — illuminates both the stakes of inclusive AI governance and the structural barriers to achieving it.

The AU AI Continental Strategy: What It Proposes

The AU AI Continental Strategy is built around a vision of AI as a driver of Africa's development agenda, explicitly linked to the African Union's Agenda 2063, which sets development targets for the continent across economic, social, and governance dimensions. The strategy's framing is explicitly dual: AI represents transformative opportunity for African development, and AI also represents significant risk if it develops without adequate governance frameworks that center African interests.

On the opportunity side, the strategy identifies several high-priority application domains. Agriculture — the foundation of most African economies — offers enormous potential for AI applications including crop disease detection, precision irrigation, market price prediction, and supply chain optimization. Healthcare AI offers potential to extend the reach of limited healthcare workforces in settings where doctor-patient ratios are far below global averages. Financial inclusion AI — the kind of alternative credit scoring used by companies like Branch, M-Pesa-linked services, and Jumo — offers the potential to extend financial services to the hundreds of millions of Africans who remain unbanked. Educational AI could help address Africa's massive teacher shortage while supporting learning in local languages.

On the governance side, the strategy identifies several priority concerns. Data sovereignty — the principle that African data should be governed by African institutions and should primarily benefit African populations — is central. The strategy expresses concern about what it terms "data colonialism": the pattern by which AI companies from wealthy countries collect data from African users and markets, use that data to train AI systems, and then sell those AI systems back to African customers without adequate benefit-sharing or local capacity building.

The strategy also addresses the infrastructure gap. Most advanced AI applications require reliable internet connectivity, smartphone access, and electricity — none of which is universally available across Africa. AI governance that does not address these infrastructure prerequisites will serve only the most economically connected segments of African populations while leaving the most vulnerable communities unaffected.

On the regulatory framework, the strategy calls for African countries to develop national AI strategies and regulatory frameworks that are coordinated at the continental level through the African Union. It proposes a continental AI governance body, with AU commission oversight, that could develop harmonized standards, share regulatory intelligence, and represent Africa's interests in global AI governance processes.

Navigating Between the Blocs

One of the most analytically interesting aspects of Africa's AI governance challenge is the continent's position at the intersection of the three major AI governance blocs. African countries are simultaneously receiving EU-origin regulation (through GDPR-aligned data protection laws), US-origin AI systems (through the dominance of American AI companies), and Chinese-origin AI infrastructure (through Huawei networks, Chinese surveillance systems, and technology provided through Belt and Road Initiative relationships). Each of these influences carries governance implications that are not always compatible.

The EU's regulatory influence is strongest in trade-dependent African countries with significant economic ties to Europe — notably the francophone West African nations. Several African countries have enacted data protection laws explicitly modeled on GDPR: Ghana's Data Protection Act, Nigeria's Nigeria Data Protection Regulation (subsequently replaced by the Nigeria Data Protection Act), Kenya's Data Protection Act, and South Africa's Protection of Personal Information Act all show significant EU influence. As the EU AI Act begins to bite, African companies exporting to or partnering with European businesses will face pressure to meet EU AI Act standards for any AI systems used in those relationships.

The US influence is primarily technological and commercial rather than regulatory. American AI companies — Google, Meta, Microsoft, Amazon, and numerous smaller firms — have significant presences in African markets and significant influence over the AI tools, platforms, and infrastructure that African businesses use. American companies generally prefer the light-touch, innovation-friendly governance model that the US has historically pursued, and their commercial influence can translate into resistance to more stringent African regulatory frameworks.

The Chinese influence is primarily through infrastructure and surveillance technology. Chinese companies — Huawei, ZTE, Hikvision, Dahua — have become major suppliers of telecommunications infrastructure and surveillance systems to many African governments. This technology often comes with governance implications: governments that depend on Chinese-supplied surveillance infrastructure face pressure not to regulate in ways that conflict with Chinese commercial interests, and the surveillance systems themselves can be used in ways that raise serious human rights concerns. Several documented cases exist of Chinese-supplied surveillance technology being used against political opposition in African countries.

The tension between these influences creates a genuine governance dilemma for African policymakers. Aligning fully with EU regulatory standards may impede commercial relationships with Chinese and American AI providers. Accepting Chinese technology governance norms may compromise democratic governance. Adopting the US innovation-first approach may leave populations vulnerable to AI harms without adequate protection. The AU AI Continental Strategy attempts to chart a path that preserves African agency rather than simply adopting any of these external models — but doing so requires governance capacity that many African states are still building.

Building Local AI Governance Capacity

The most significant structural challenge facing African AI governance is the capacity gap. Building effective AI regulatory frameworks requires technical expertise that is in short supply across Africa — expertise in machine learning, data science, algorithmic auditing, and AI policy. African universities are producing rapidly growing numbers of computer scientists and data scientists, but competition from international companies (including American and Chinese AI companies) for that talent creates a brain drain that depletes the very human capital needed for domestic governance.

Several initiatives are attempting to address this capacity gap. The Smart Africa Alliance, an initiative of African heads of state established in 2013, has developed AI governance training programs for African civil servants and regulators. The African Union's Digital Transformation Strategy includes capacity building components. Individual countries — notably Rwanda, Kenya, South Africa, Nigeria, Egypt, and Morocco — have developed national AI strategies with varying degrees of implementation capacity behind them.

The research community is building important foundations for African AI governance. Masakhane NLP, a grassroots research organization focused on natural language processing for African languages, has demonstrated that community-driven, locally-relevant AI research is possible without relying on external funders' priorities. The Deep Learning Indaba, an annual gathering of African machine learning researchers, has built a community of practice that is developing the talent base for both African AI development and African AI governance.

Civil society organizations are playing an increasingly important role. The Initiative for a Just Society (Kenya), the Collaboration on International ICT Policy for East and Southern Africa (CIPESA), and Research ICT Africa have produced important research on AI governance challenges specific to African contexts, ranging from the use of AI in electoral processes to the governance of AI-enabled financial services.

What Genuine Inclusion Would Require

The story of African AI governance illustrates what genuine inclusion in global AI governance would require — and how far current arrangements fall short.

Genuine inclusion requires resources as well as seats at the table. Civil society organizations and government agencies in African countries that want to participate in GPAI, UNESCO, ITU, and other global AI governance processes face funding constraints that their wealthier country counterparts do not. Meaningful inclusion requires not just invitations to participate but financial support for that participation — covering travel, translation, and the organizational capacity needed to engage substantively in complex technical policy processes.

Genuine inclusion requires agenda-setting power, not just participation in processes others have defined. The OECD AI Principles, the G7 Hiroshima Code of Conduct, and even the UNESCO Recommendation were substantially developed before African voices were fully incorporated, and African participation has largely taken the form of endorsement or reaction rather than agenda-setting. Governance processes that genuinely include Africa would incorporate African research priorities, African governance challenges, and African perspectives on what AI ethics means from the outset.

Genuine inclusion requires that global AI governance address the harms most salient in African contexts. The most sophisticated AI governance discussions in Western capitals focus on issues like deepfakes, algorithmic bias in credit scoring for middle-class consumers, and job displacement from automation. These are real concerns. But they are not necessarily the most urgent AI governance concerns for most Africans. The use of AI-enabled surveillance to suppress political opposition, the extraction of African data for AI training without benefit-sharing, and the deployment of AI systems that don't work adequately for African populations and languages are, in many ways, more acute harms that receive less attention in global governance discussions.

Finally, genuine inclusion requires that the knowledge and perspectives of African communities be recognized as governance contributions rather than mere inputs to governance processes designed elsewhere. The experience of African farmers with agricultural AI, African patients with health AI, and African workers with algorithmic management is governance-relevant knowledge that current international AI governance processes are largely failing to incorporate.

Lessons for Global AI Governance

Africa's experience with AI governance offers several lessons for those working on global AI governance more broadly.

The diversity within the "Global South" is as important as the diversity between it and the Global North. Africa's 54 countries span enormous variation in governance capacity, digital infrastructure, economic development, and AI exposure. A continental AI strategy that works for Rwanda — with its sophisticated digital infrastructure, strong state capacity, and established AI regulatory framework — may not work for the Central African Republic or Somalia. Governance frameworks must be designed with genuine flexibility for local adaptation, not merely cosmetic acknowledgment of diversity.

The competitive dynamics between major AI powers can be turned to governance advantage in some cases. African governments that face pressure from Chinese infrastructure providers and American AI companies simultaneously can sometimes leverage that competition to extract better governance commitments from both. Some African governments have used the credible threat of choosing Chinese alternatives to negotiate better data governance terms with American companies, and vice versa. This leverage is limited but real.

Local capacity building is not just a development priority but a governance prerequisite. International AI governance frameworks that require regulatory capacity to implement will fail in jurisdictions that lack that capacity. Genuine global AI governance must therefore include meaningful commitments to technical assistance and capacity building, not as an afterthought but as a core governance function.

And the content of AI governance is inseparable from who governs. The AI governance frameworks that emerge from predominantly wealthy-country processes will reflect the priorities of wealthy-country populations. Making global AI governance genuinely global requires not just broadening participation but ensuring that broadened participation has genuine influence over substantive outcomes. The AU AI Continental Strategy represents a serious effort to assert African agency in AI governance — to say, in effect, that the terms of AI's deployment in Africa should be set by Africans and for Africans. Whether that assertion translates into effective governance depends on factors that extend well beyond the drafting of strategy documents.

Discussion Questions: (1) The AU AI Continental Strategy calls for African data sovereignty. What would that mean in practice for international AI companies operating in African markets? (2) Several African countries have adopted GDPR-aligned data protection laws. Does this represent genuine adoption of appropriate governance frameworks, or regulatory mimicry that may not serve African contexts? (3) What specific commitments would wealthy-country governments and major AI companies need to make for Global South inclusion in AI governance to be meaningful rather than performative?