Decolonising Generative AI in South African Teacher Education: Wayfinding for Multilingual, Indigenous, and Publicly Accountable Futures

This viewpoint article argues that South African teacher education requires a distinctly decolonial agenda for generative artificial intelligence (AI), developed in critical dialogue with scholarship from Aotearoa New Zealand and wider international contexts. AI is not a neutral efficiency tool: its infrastructures, training data, evaluation standards, and platform logics can intensify linguistic hierarchy, data extraction, epistemic marginalisation, and pedagogical standardisation. Drawing on epistemic justice, decoloniality, settler-colonial analysis, and wayfinding, the article synthesises policy and scholarship to develop a framework for multilingual, Indigenous, and publicly accountable AI in teacher education. The comparison is deliberately relational rather than transfer-oriented. South Africa and Aotearoa have distinct colonial histories, constitutional arrangements, and Indigenous political formations; nevertheless, Te Tiriti o Waitangi-informed educational thinking, Māori data and algorithmic sovereignty, and Pacific-led teacher education provide productive tests of whether AI governance redistributes authority or merely adds cultural content to externally controlled systems. The article advances six interdependent propositions: language justice by design; Indigenous Knowledge Systems stewardship; critical AI literacy for teacher agency; assessment redesign around process and judgment; public-interest data governance; and infrastructure-aware inclusion. Decolonising AI, it concludes, is not metaphorical branding or the insertion of local examples into imported systems. It requires material changes to curriculum purposes, decision rights, data relations, evaluation criteria, and institutional accountability so that AI strengthens multilingual participation, community authority, teacher judgment, and public benefit.