Between July 2024 and May 2026, a continental conversation on artificial intelligence consolidated across Africa. The African Union adopted a Continental AI Strategy. African states meeting at the Global AI Summit on Africa endorsed the Africa Declaration on Artificial Intelligence in Kigali. The AU High-Level Policy Dialogue convened in Addis Ababa with heads of state and continental commissioners. Researchers at the ILINA Program and the African Hub on AI Safety, Peace and Security at the University of Cape Town proposed an Africa-centric framework for AI safety evaluations. African technology leaders and regulators launched the REST-AI Governance Framework in Abuja. Three sovereignty vocabularies circulate through these texts: data, AI, digital. A fourth, narrative, has emerged in cultural and policy debate. Each is necessary. None is sufficient. This paper introduces a fifth, conceptually upstream layer: cognitive sovereignty. It is defined as the capacity of a political community to produce, validate, and circulate the categories from which it understands itself, particularly when AI becomes the dominant medium of categorial production. The paper articulates three operational implications (for public policy, universities, and creative industries) and proposes a single legible criterion for evaluating Africa-Europe AI cooperation: its capacity to expand, rather than narrow, the cognitive sovereignty of the African interlocutor. The argument advanced here is neither Chinese-style autarky nor American-style hegemony. It is a third option: a continent that produces the categories from which it negotiates its place. — v2.0 (22 May 2026) — English working paper. Adapted, not literally translated, from the French v1.2 (May 2026). Integrates the institutional and academic record of the African AI conversation: AU Continental AI Strategy (July 2024), Africa Declaration on AI (Kigali, 4 April 2025), AU High-Level Policy Dialogue (Addis Ababa, 17 May 2025), Ireri et al. (2026) on Africa-centric AI safety evaluations, and the REST-AI Governance Framework (Abuja, 30 April 2026). The francophone bibliographic lineage (Stiegler, Mhalla, Garno) is preserved in a dedicated note. This working paper is part of the research activity of CERHCA (Cercle d'Études et de Recherche des Humanités et Cultures Africaines, French law 1901 association, WALDEC W452019663), whose vocation is to explore African humanities and cultures, including the political and conceptual dimensions of artificial intelligence on the continent.
Béatrice ODUNLAMI (Fri,) studied this question.