Leroy Abiguime speaks on the place of African languages in tomorrow's models, alongside other pan-African voices.
A summit almost never exists to announce a technology. It exists to get a problem recognised. The one carried to Geneva on 12 July 2026 fits in a sentence: a citizen who does not speak the language of their administration stays locked out of the digital services that administration puts online.
What was at stake in Geneva
The AI for Good global summit ran from 7 to 10 July 2026. Togo presented its project to build AI models for its national languages, led by Togo AI Lab with the pan-African platform Zindi and Umbaji. The announcement came on 12 July.
This was not a product demo. It was the framing of a public undertaking, with quantified targets and a schedule: fifty languages, fifty hours of validated speech and six thousand translated sentence pairs for each.
Infrastructure, not a showcase
Language models are essential public infrastructure.Cina Lawson, Minister for Public Service Efficiency and Digital Transformation
The word matters, and it commits. A showcase is judged by the effect it produces on stage; infrastructure is judged by what it carries for years. Treating a language model like a road means accepting that it is funded over time, measured in public, and corrected when the measurement disappoints.
Too many African languages remain under-represented.Celina Lee, CEO of Zindi
What Umbaji brings
Our role is collecting and validating the data. A speech model is worth exactly what its training corpus is worth, and a low-resource corpus cannot be scraped from anywhere: it is built, one recording at a time, by native speakers who review each other's work.
The detail of the partnership — the ten languages already named, the Zindi competitions, each partner's role — is in the article devoted to it.
