Billions of people risk being left behind by artificial intelligence because it does not understand their languages or the conditions they live in, the Gates Foundation has warned.

The rapid rise in the use of AI risks widening the gap between rich and poor unless it is deliberately built to serve people in the world’s poorest countries, the Foundation’s annual report on global progress found.

More than 90 per cent of the data used to train early large language models came from English-language sources, with leading AI speech-recognition systems makeing errors less than 6 per cent of the time in English.

In Yoruba, spoken widely in Nigeria and elsewhere in west Africa, the error rate is ten times higher.

AI tools also risk failing when they are transplanted from wealthy countries into poorer ones without local data, the foundation found. A system designed for a commercial farmer in Iowa, for example, may not understand the crops, pests, markets or constraints facing a smallholder in Ethiopia.

The Gates Foundation’s 2026 Goalkeepers Report, published Tuesday, also warned that, left to the market, the most capable AI tools are likely to be built first for the people and institutions able to pay for them, rather than those who could benefit from them most.

However, it also highlights how AI could also help close huge shortages of doctors, teachers and agricultural advisers in poorer countries, by putting specialist knowledge in the hands of frontline workers.

The foundation is committing at least US$1bn over the next two years to efforts to make AI more accessible, with funding going to healthcare, education, agriculture and towards digital infrastructure, such as local-language datasets.

“AI could be a great equalizer – or widen the gap. This is not a long-range prediction. It’s a present-tense choice,” Bill Gates said in the report.

“The decisions being made in the next 12 to 18 months – about how AI is built, funded, and deployed – will determine whether this technology primarily benefits the people who already have the most or reaches those who have the least,” he added.

Early projects include an AI system that has improved diagnostic accuracy in Kenyan clinics, education tools showing large learning gains in Sierra Leone and the US, and an agricultural service now used by more than 740,000 farmers in India.

This article has been produced as part of The Independent’s Rethinking Global Aid project