On September 21 in New York, the Gates Foundation, Anthropic, Microsoft, Google, the OpenAI Foundation, the World Bank Group, and others signed a joint commitment. Within five years, they want an estimated 3.4 billion speakers of poorly served languages to be able to use AI in their own language and voice.
New to this? Read it in simple words
- On September 21, the Gates Foundation and 59 partners signed a pledge in New York. Signers include Anthropic, Microsoft, Google and the World Bank Group.
- Within five years, they want AI to work in people’s own language and voice. The goal covers an estimated 3.4 billion speakers of poorly served languages.
- Most AI tools work best in English and a few other large languages. Access in more languages depends on data, tests and local builders, not only promises.
- The pledge has no budget yet. The partners will decide over the next year how the group is run.
- Pledge
- A public promise to do something.
Who signed, and what they promise
The Gates Foundation and 59 other signatories published the joint commitment in New York on September 21, during the high-level week of the UN General Assembly. The document is titled “Ensuring AI speaks everyone’s language”.
Signatories include AI companies such as Anthropic, Microsoft, Google, Amazon, NVIDIA, and Mistral. The OpenAI Foundation signed, not the OpenAI company. Unicef, the World Bank Group, the UK Foreign Office, and Senegal’s digital ministry are also on the list.
Many signatories already build AI for their own languages. They include Masakhane, Lelapa AI, and Data Science Nigeria in Africa, and AI4Bharat, Karya, and AI Singapore in Asia.
The pledge aims at an estimated 3.4 billion speakers within five years. No money is attached yet.
Four areas of work, and a big number
The partners plan to work in four areas. The first is open infrastructure for language data. The second is benchmarks, tests that show how well AI works in each language.
The third is models and apps that any developer can use. The fourth is safe use that protects privacy, consent, and data sovereignty, which means that communities keep control of their own data.
The goal is that an estimated 3.4 billion speakers of poorly served languages can use AI in their own language and voice. The foundation has not published how it calculated this number or how it will measure progress.
No budget yet
No money is attached to the pledge. On September 14, the Gates Foundation separately promised at least $1 billion over two years for AI that improves health and expands opportunity. About 10 percent of that is for groundwork such as language datasets.
The partners will decide over the next year how the group is run. Each partner’s contribution is voluntary and not yet public.
At the foundation’s Goalkeepers event, Bill Gates called for the “smart use” of AI, AP reports. Anthropic’s Elizabeth Kelly told AP that its products lag in many African languages.
Some signatories add their own warnings. The Internet Society says communities should be able to opt out, and that many people first need internet access.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1Primary source · September 21, 2026Global Organizations Announce Five-Year Goal to Help More Than Three Billion People Use AI in Their Own Language and Voice Gates Foundation
- 2Primary source · September 2026Ensuring AI speaks everyone’s language: A joint commitment Gates Foundation
- 3Research · September 21, 2026Bill Gates pushes for “smart use” of AI as his foundation builds more representative language data AP via KOB
- 4
- 5Primary source · September 14, 2026Gates Foundation Commits US$1 Billion to Help Build and Deliver Equitable AI That Improves Health and Expands Opportunity Gates Foundation
- 6Primary source · September 21, 2026Closing AI’s Language Gap: Why the Internet Society Is Joining a Global Commitment Internet Society
We read the Gates Foundation’s announcement, the full commitment and its signatory list, and a signatory’s blog post. We compared them with AP and CBS News. The 3.4 billion figure is the foundation’s own estimate, and it has not published its method.