The Gates Foundation is convening a coalition of 60 artificial-intelligence laboratories, companies, philanthropies and other organizations to expand usable language data for communities poorly represented in current systems. The initiative aims to reach more than 3 billion people over five years by coordinating work that many partners were already pursuing separately.
Bill Gates argued that increased international generosity and the smart use of AI could accelerate progress against inequality. He also said developers had failed to embed the right human values and keep a moral code in control of AI, joining practical optimism with a warning about how the technology is governed.
The project arrives as 26 of the world’s 34 wealthiest countries reduced development aid to poorer nations in the previous year. Progress toward the United Nations Sustainable Development Goals remains slow and uneven, making the coalition’s technology promise inseparable from funding and institutional capacity.
The foundation had already committed $1 billion to AI-focused work in health, education and support for small farmers. At its annual Goalkeepers gathering, it highlighted teachers identifying learning gaps, farmers detecting pests and health workers retrieving patient information, but those examples do not yet prove impact at the coalition’s intended scale.
Language quality is central because mistranslation can have material consequences. One example showed how a pregnant woman in Malawi saying that her water had broken could be translated literally as having thrown away water, illustrating the risk when models lack local linguistic and cultural context.
Mozilla Data Collective emphasized that internet data is not representative and that communities should be able to contribute cultural and language sets on their own terms. Google’s Project Vaani is collecting more than 150,000 hours of speech across every district in India, using local partners to capture dialect differences inside languages.
Anthropic, Google and the OpenAI Foundation are among the participants. Anthropic acknowledged that its products lag in many African languages, while coalition governance is still being defined. A secretariat is expected to track commitments and identify gaps, but individual obligations have not all been finalized.
Reaching 3 billion people is therefore an objective, not a measure of present adoption. The evidence to watch will include consent and licensing for data, coverage of dialects, independent quality tests, documented local benefits and whether communities retain influence over how their language resources are used.
