As artificial intelligence (AI) becomes critical infrastructure for global health, it reproduces colonial patterns of extraction, mining data from the Global South to train models owned by the Global North. While international bodies like the WHO emphasize “ethical AI,” they often overlook the structural violence of this digital colonialism. This perspective argues that true health equity requires more than bias mitigation; it demands cognitive sovereignty: the right of communities to govern not just their data but also the epistemic logic, interpretive frameworks, and algorithmic reasoning of the systems that analyze it. Drawing from Indigenous data governance principles (OCAP/CARE) and concrete implementation cases from Kenya, Nigeria, Rwanda, and Latin America, we demonstrate how cognitive sovereignty extends beyond data sovereignty to encompass control over knowledge production itself. By anchoring this political vision in specific technical architectures, federated learning, and community-led surveillance, we can move from extractive “AI for good” to a decolonial future of autonomous health intelligence. Recent cases from Kenya's AI health deployments and pathogen genomics illustrate both the urgency and feasibility of this transformation.
Kakraba et al. (Thu,) studied this question.