A series of glass-fronted offices that were nonexistent ten years ago are passed by the taxi on its way from Jomo Kenyatta International to Nairobi’s Westlands neighborhood. Swahili, Kikuyu, and Luo—languages that the majority of foundation models still treat as afterthoughts—are being used by a small team inside one of them to train a language model. Outside, vendors use their phones to pay for fruit, just as they did long before the term “fintech” became popular in San Francisco. The contrast between the new urgency of artificial intelligence and the old habit of leapfrogging reveals something about the true nature of this narrative.
Beijing, Washington, Brussels, and a section of California highway have dominated the AI discourse for the majority of the last five years. That is beginning to change, and not in a good way. The ITU’s Fred Werner made a persistent point when he spoke at the India AI Impact Summit earlier this year: the global AI debate has swung from existential fear in Bletchley Park to industrial competitiveness in Paris, and now it’s moving toward something quieter and more useful, impact. It remains genuinely unclear whether that pivot will hold or if it will be drawn back into the orbit of the large labs.
Kenya is the obvious case study, in part because its economists have been discussing “AI sovereignty” in a straightforward way that is uncommon elsewhere. In a piece for Brookings last fall, Bitange Ndemo posed the question of who owns the training data, the fiber, the data centers, and the foundational layers. Compared to the courteous “capacity-building” language you hear at UN side events, this framing is sharper. Speaking with members of the ecosystem gives the impression that they have already witnessed one wave of platform capitalism overtake their economies, and they have no intention of remaining passive this time.
Human capital is the more difficult question. Engineers are actually and continuously leaving Lagos, Bangalore, Manila, and Nairobi. Some scholars, such as the analysts at New America, contend that diasporas can be activated as bridges rather than lamented as departures, so this isn’t always a loss. Perhaps. It’s a neat theory. It depends on nations genuinely establishing the institutions that make going back worthwhile and on the North’s visa policies remaining open, neither of which seems safe to assume in 2026.
The labor underside is the part of the keynote slides that no one includes. Working through subcontractors for the most valuable AI companies in the world, Kenyan data annotators have spent years labeling training data for pay that hardly covers rent in the cities where they are contributing to the development of their products. To refer to this as the “first rung on the value chain” would be an overstatement. It is more accurate to refer to it as digital piecework. Talk about empowerment will seem a little hollow to some people until that aspect of the discussion becomes more prominent.
However, something is actually moving. Teachers in Manila and policy advisors in Bridgetown are taking courses offered by the AI Skills Coalition. Nairobi’s small startups are developing diagnostic tools for clinics without an MRI machine. This year, UN Women highlighted Bioniks, a Pakistani company that uses AI to create child prosthetics—a project that would not have been possible ten years ago. The leapfrog is not guaranteed by any of this. However, it’s difficult to ignore the fact that the individuals working on the most fascinating projects at the moment are frequently the ones that the initial AI maps completely ignored.
