AI Must Speak With Africa, Not Merely About Africa

The East African Community has placed a question at the centre of artificial-intelligence governance that technology debates too often leave at the edge: how well does an AI system actually understand, speak and serve the language of the people expected to trust it? In a press release published on 17 July, the East African Kiswahili Commission reported the Bujumbura Resolutions on Kiswahili, Multilingualism and AI and announced the beginning of work on a Kiswahili Trust Mark—Alama ya Uthibitisho wa Kiswahili—for AI systems.

The proposed certification framework is intended to assess linguistic quality, terminological accuracy, cultural context and safety, with a public register of certified systems. Consultations over the coming year will shape the final framework, so this is a commitment under development rather than an operating certification scheme. Its governing idea is nevertheless powerful: African language performance should be independently tested, publicly visible and relevant to decisions made by governments, schools, businesses and citizens.

Digital projects often treat language as a translation layer added after a platform has been designed. That approach mistakes words for the whole of meaning. Language carries categories, relationships, humour, authority, emotion, memory, theology and expectations about how knowledge is communicated. An answer may be grammatically correct while remaining culturally tone-deaf, pastorally unsafe or educationally confusing.

The EAC Partner States have therefore committed to treating Kiswahili as strategic digital infrastructure. Their…