How a $400M Nonprofit Plans to Build AI's Public Alternative

A new nonprofit called Current AI isn’t waiting for Silicon Valley to make artificial intelligence speak every language. It’s building the public infrastructure itself—offline, open-source, and in 22 Indian languages to start. The Suno Sutra device, a pocket-sized translator co-developed with the Indian government’s Bhashini division, runs entirely without the internet. It’s the first tangible output from an organization that has moved with unusual speed since its February 2025 founding. CEO Ayah Bdeir, the former Mozilla AI lead and littleBits founder, frames the mission bluntly: every major AI system today is controlled by a private company, and the world needs a public alternative as fundamental as the World Wide Web.

That vision is backed by $400 million in committed funding, seeded with $100 million from the French government and bolstered by the Ford Foundation, MacArthur Foundation, Google DeepMind, and Salesforce. Bdeir draws a sharp distinction between these funders and traditional investors. The money is already flowing. Last month, Current AI deployed $3.2 million in its first grant round to four projects across Kenya, Lebanon, and the Brazilian Amazon. The work ranges from building AI datasets for over 50 African languages to creating offline tools for Indigenous communities that keep data strictly within their territories.

Bdeir is deeply skeptical of Big Tech’s multilingual push, arguing it’s market expansion masquerading as inclusion. She points to a grim reality: missionary Bible translations often become training data for Indigenous languages before the communities themselves have set any rules. For Current AI, language is not just a technical problem. It’s how knowledge, tradition, and identity move between generations. A technology that can’t speak your language can’t hold your culture either, she argues. The nonprofit’s approach is deliberately small-scale, embedding consent protocols into the pipeline so communities can halt projects at any stage, and bringing in local experts before a single line of code is written.

Bdeir rejects the Big Tech paradigm that scale is the only measure of success. She envisions a different kind of victory: an Indigenous elder in the Amazon using a tool built in Kenya to pass down ecological knowledge in their own language. None of the grantees have fully solved the data ownership question, but that ongoing negotiation is exactly the point. Complexity is not an excuse to let a government or a tech company decide for everyone. The question Current AI is testing is whether a modest budget, distributed locally and built in the open, can create an alternative that the centralized giants will never bother to build.