Bill Gates calls for “smart use of AI” as his foundation builds more inclusive language data

At the Gates Foundation’s annual Goalkeepers gathering in New York, Microsoft co‑founder Bill Gates urged wealthy countries to pair “increased generosity” in foreign aid with the “smart use of AI” to tackle global inequality. A central plank of that vision is a new push by the foundation to build language datasets that better represent communities in Africa, South Asia and other regions historically ignored by big tech.

Most large language models today are trained primarily on English and a handful of dominant languages. That can make AI tools awkward or even harmful when deployed in low‑income countries, where local idioms, health practices and social norms differ sharply. Gates argued that using AI for “life‑saving innovations” — from digital health advice to crop extension services — requires models that truly understand local languages and respect data privacy.

The foundation plans to work more closely with local researchers and NGOs to co‑design applications such as chatbots for maternal health, automated summaries of public‑health alerts and tools to help overstretched clinic workers. But the initiative also raises uncomfortable questions about power: who owns the resulting datasets, how consent is obtained, and how to avoid a new form of digital colonialism where communities provide data without gaining real control or long‑term benefits. For AI to narrow global gaps rather than widen them, governance will matter as much as model performance.