Generative Artificial Intelligence and Linguistic Epistemic Justice for isiZulu and African Languages in Multilingual Education

Abstract Generative artificial intelligence is rapidly entering educational practice, yet its implications for African languages remain unevenly understood. This article presents a critical integrative review and conceptual synthesis of 44 substantive scholarly and policy sources spanning generative AI, African-language technologies, multilingual education, language policy, decolonial AI and epistemic justice. Using South Africa and isiZulu as a critical case, the review asks how AI-mediated education may recognise, marginalise or reconfigure historically minoritised languages as vehicles of knowledge. The synthesis connects evidence from AI in education, African-language natural language processing, translanguaging, data justice, teacher agency and responsible AI governance. It argues that the central issue is not whether ChatGPT can generate isiZulu text, but whether AI systems, institutional policies and pedagogical practices can sustain African languages as infrastructures of reasoning, identity, pedagogy, cultural memory and intellectual authority. The article develops linguistic-epistemic AI justice as a concept that links language equity to data representation, model performance, teacher mediation, and institutional accountability. It then proposes a process-based framework for language-equitable AI governance in multilingual education. By linking digital infrastructure, pedagogy and language policy, the review reframes low-resource language support as a core problem of educational justice and AI governance. It offers a governance agenda for more inclusive and equitable AI-supported education.