Generative AI and the rewriting of historical narratives in post-colonial Africa

This article examines the influence of generative artificial intelligence on the formation and dissemination of historical memory in postcolonial Africa. Drawing on field research conducted in Senegal, Kenya and Nigeria (45 semi-structured interviews, a corpus of 12,500 digital verbatim transcripts and 300 comparative algorithmic tests), the study analyses how global language models, although perceived as neutral technological tools, actually operate as mechanisms of semantic extraction and standardization. First, we highlight how the materiality of algorithms and the geopolitics of data establish a form of ‘digital coloniality’, confining African historical narratives within probabilistic frameworks defined by the Global North. We then examine the tactics of resistance deployed by local users. Through decolonial prompt engineering and practices of semantic ‘poaching’ (illustrated by concrete cases such as ‘memory jailbreaking’ of the Mau Mau uprising or the generation of images of postcolonial architecture), these actors manage to subvert computational architectures in order to reintroduce subjectivities and truths marginalized by the digitized ‘colonial library’. The article concludes with the paradox of digital sovereignty – which, whilst resorting to subterfuge, remains structurally dependent on external infrastructure – and calls for a material break through initiatives such as Masakhane or Lelapa AI, to achieve an effective decolonization of thought.