Artificial intelligence is rapidly becoming a foundational technology, comparable to electricity or the internet. Like previous general-purpose technologies, it promises productivity gains, new services, and economic growth. But history shows that such technologies rarely spread evenly. They tend to reward those with capital, infrastructure, and institutional capacity first, often widening gaps long before any convergence occurs.

A classic illustration of this pattern is the adoption of electricity in the late 19th century. Electricity’s major productivity gains emerged decades after its invention, initially benefiting large urban factories that could invest in new machinery and organizational redesign, while smaller firms and rural areas fell behind. This uneven diffusion initially widened economic disparities, as detailed by Paul David in "The Dynamo and the Computer." Only later, as supporting infrastructure and skills spread, did broader convergence in productivity occur.

This is the central warning of the United Nations Development Programme’s report, "The Next Great Divergence." Examining AI through a human development lens, the report argues that without deliberate policy choices, AI is likely to intensify inequality between and within countries rather than reduce it. AI, the report insists, is not destiny. Its impact will depend on choices made now, about investment, governance, skills, and who gets to participate in shaping these systems.

For Africa, this warning is not abstract. It describes conditions that are already visible.

Unequal starting points, unequal outcomes

The UNDP identifies three channels through which AI may shape inequality: people, the economy, and governance. Across all three, starting points matter. Countries with reliable electricity, high-quality connectivity, compute infrastructure, skilled workforces, and strong institutions are positioned to capture early gains. Others risk slower uptake, weaker system performance, and growing dependence on externally developed technologies.

Although the report focuses on Asia and the Pacific, its diagnosis resonates strongly in African contexts. Many African countries face gaps in digital infrastructure, limited access to compute, shortages of specialized skills, and constrained regulatory capacity. Where these foundations are weak, AI adoption often takes the form of donor-funded pilots, vendor-controlled platforms, or imported “black-box” systems.

Natural language processing Systems deployed or piloted in Africa, including translation and conversational tools, have struggled because the underlying models were trained on data that doesn’t reflect local languages, dialects, or cultural context. They produce inaccurate outputs and rely on data and expertise that reside outside the countries where they are used, illustrating the risks of vendor-controlled, black-box AI adoption when local foundations are weak. As the UNDP warns, such patterns can deepen dependency rather than build long-term capability.

For people, AI offers the promise of improved health diagnostics, personalized education, and more accessible public services. Yet when African communities are missing from datasets – or are represented only through proxies – systems trained elsewhere misclassify, exclude, or distort lived realities. The report highlights risks of biased or opaque systems denying benefits, undermining rights, and eroding trust, particularly for women, rural populations, and marginalized groups.

For the economy, AI may raise productivity and create new forms of work, but gains are likely to concentrate where skills, capital, and innovation ecosystems already exist. Countries unable to invest in infrastructure, research, and local enterprise risk being locked into low-value roles in global AI value chains, as data sources or end-users rather than creators.

For governance, AI can strengthen public administration and decision-making, but only where institutions can audit systems, enforce accountability, and provide meaningful avenues for redress. In lower-capacity settings, the report warns, governments may become dependent on systems they cannot fully understand, adapt to, or contest, weakening both sovereignty and public trust.

Taken together, these dynamics describe the conditions for what the UNDP calls a “next great divergence”: overall technological progress accompanied by widening gaps in agency, capability, and power.

Choice, not inevitability

Crucially, the UNDP does not present this outcome as inevitable. Whether AI narrows or widens inequality depends on how governments sequence action, invest in both solid foundations – power, connectivity, compute – and soft capacity, such as skills, institutions, and governance, along with design frameworks that put people at the center of technological change.

What remains underdeveloped in global AI debates, however, is how countries with structurally unequal starting points can enter AI ecosystems on terms that build long-term capability rather than reinforce dependency. Many prevailing governance models assume mature infrastructure, strong regulatory institutions, and domestic AI industries. Simply importing these frameworks risks reproducing misalignment between rules and reality.

For Africa, this gap matters. Governance cannot be reduced to risk checklists or ethical principles abstracted from context. It must grapple directly with questions of power, participation, and development.

Ubuntu as an Africa-centred response

Building on a key insight from Gwagwa et al., I argue for a governance framework that treats AI as relational and social rather than purely technical or proprietary. Gwagwa and colleagues highlight how Ubuntu can ground AI ethics in African values, emphasizing inclusivity and collective well-being in shaping who benefits from and who is accountable for AI systems.

A governance framework that treats artificial intelligence as relational and social, rather than purely technical or proprietary. A framework that recognizes that AI systems are socio-technical arrangements that shape relationships between states, firms, institutions, and communities, and that these relationships determine who benefits, who bears risk, and who builds lasting capability.

In a forthcoming set of publications, “The Ubuntu AI Framework” and “Ubuntu AI Scorecard,” I respond to this challenge. Rooted in the African ethic of Ubuntu, “I am because we are” the Ubuntu AI Framework and Ubuntu AI Scorecard are products of collective reflection rather than top-down design. They grew out of a gathering where researchers, artists, technologists, and civil society actors came together to wrestle with questions of power, inclusion, and accountability in the age of AI.They are not a rejection of global AI development, but a way to engage with it deliberately and on clearer terms.

The framework begins from a simple recognition: most AI systems used in Africa today are developed elsewhere. Engagement with global firms, standards, and platforms is therefore unavoidable. The policy question is not whether to participate, but how to do so in ways that strengthen domestic institutions rather than bypass them.

The framework treats intelligence as relational and social. This reframing shifts governance away from narrow performance metrics toward questions of responsibility, benefit-sharing, accountability, and long-term capability-building.

Importantly, the Ubuntu AI Framework does not downplay the need for African-owned technologies. On the contrary, it affirms that building local models, infrastructure, skills, and enterprises is a strategic objective. What it rejects is the false choice between immediate participation and future sovereignty. Instead, it offers a pathway to pursue both in parallel.

Operationalizing capability through the Ubuntu AI Scorecard

Its companion Ubuntu AI Scorecard translates these values into measurable standards, assessing AI partnerships on ownership, skills, data sovereignty, governance, and socio-economic impact to ensure Africa is a co-creator, not just a data source. It provides a structured way to evaluate AI initiatives, partnerships, and procurements, not only on technical or cost criteria, but also whether they contribute to developmental outcomes.

The Scorecard asks whether AI deployments support skills transfer and institutional strengthening, invest in local infrastructure and innovation ecosystems, uphold data governance and accountability, and generate shared socio-economic value. In doing so, it complements regulation by embedding development considerations directly into decision-making.

This approach aligns closely with the UNDP’s emphasis on sequencing and starting points. Rather than assuming equal capacity, it treats governance as a tool for building capability over time, ensuring that AI adoption strengthens, rather than bypasses, domestic foundations.

Partnerships on deliberate terms

Given Africa’s current position in global AI ecosystems, partnerships are unavoidable. The Ubuntu AI Framework does not deny this reality. What it insists on is that partnerships must be structured to accelerate domestic capability, not substitute for it.

This includes explicit attention to skills development, co-creation, infrastructure investment, data stewardship, and support for African enterprises. Ubuntu thus functions as a negotiation framework, equipping public institutions to engage global AI actors with clearer expectations and enforceable conditions. Rather than passive adoption, it promotes intentional participation.

From warning to response

The "Next Great Divergence" is ultimately a warning about what happens when transformative technologies diffuse along existing fault lines. Africa’s history offers painful lessons about such moments, but also reminders of agency. Convergence has occurred before when technology was aligned with investment in people, institutions, and public value.

The Ubuntu AI Framework and Scorecard represent an attempt to translate warnings into response. They are explicitly Africa-centred, grounded in lived realities, and designed to complement global governance efforts rather than replicate them uncritically.

If AI’s future will be shaped by choice, as the UNDP argues, then Africa’s task is clear: to make those choices explicit, enforceable, and development-oriented, before the next divergence becomes entrenched.