Africa's AI Gap: Archives, Not Just Algorithms
Akintunde Babatunde argues that Africa's engagement with AI is structurally disadvantaged because the continent's knowledge is largely absent from digital archives and training data. Externally, AI systems train on datasets where major African languages like Hausa represent less than 0.004% of Common Crawl, and Africa supplies ~70% of global cobalt but captures only ~10% of its revenue. Internally, decades of underfunded universities and unarchived knowledge—from theses to oral traditions—means there was little digital content to scrape in the first place. Even locally built tools (e.g., CJID's WhatsApp fact-checker) face crowding from global platforms and rely on Northern cloud infrastructure. The solution isn't just building more tools but creating structured co-creation: developers and journalists collaborating from the start, continental compute infrastructure, and procurement policies favoring local innovation. AI governance funding must be independent, not from regulated companies. With Africa's population projected to hit 2.5 billion by 2050, the window is closing for the continent to move from being a source of raw materials and data to a site of decision-making and value capture. Will African states act collectively to treat knowledge infrastructure as a strategic priority, or accept permanent dependency?