African AI is leaving out 100 million people with disabilities
Kenyan Sign Language, Kinyarwanda speech affected by speech disorders: the data needed for AI to serve people with disabilities is missing. A network of researchers wants to fill the gap.
- More than 100 million Africans have a disability, but they are absent from the data used to train AI.
- Kenyan Sign Language, Kinyarwanda speech: the data is missing.
- A network of researchers is preparing three datasets and will fund start-ups.
More than 100 million Africans live with a disability, according to the World Health Organization. In the 47 countries of its Africa region, more than one person in eight has a significant disability. Yet these people are almost absent from the data used to train artificial intelligence systems.
Data that does not exist
Examples abound. Data in Kenyan Sign Language is very limited compared with American Sign Language. Voice recordings in Kinyarwanda of people with speech disorders are virtually absent from large datasets. Live captioning tools struggle with accents, fast speech and background noise.
A study commissioned in 2025 by the AI4D programme, carried out in Ghana, Kenya and Rwanda with more than 585 participants, confirmed that people with disabilities are under-represented in training data.
What is being put in place
The Hub for AI and Disability Inclusion (HAIDI), hosted by the responsible AI lab at KNUST university in Ghana, and the African Disability Data Network are preparing three datasets: sign languages, mental and reproductive health, and inclusive employment. An 18-month programme will support start-ups, with grants of up to 5 million Kenyan shillings (about $38,600). These initiatives are backed notably by Canada’s IDRC and the UK Foreign Office.
Spotted an error? Write to infos@novotech.blog. Corrections are flagged at the end of the article.




