A $400 Million Nonprofit Wants to Build the Public Internet of AI
Current AI is handing out grants, shipping offline devices, and fighting to keep AI from becoming a private club. Here is what it is building, who is paying for it, and why a farmer in rural India is the whole point.
Key points
- Current AI, a nonprofit founded in February 2025, has secured $400 million in committed funding from governments, foundations, and companies including DeepMind and Salesforce.
- The French government seeded the organisation with $100 million as the first major contribution.
- Current AI allocated $3.2 million in grants last month to four organisations working in Kenya, Lebanon, and the Brazilian Amazon.
- Its Suno Sutra device runs AI across 22 Indian languages without an internet connection.
- Alpha Chat, an open-source chatbot (a conversational AI tool anyone can inspect and modify), was built in seven weeks by ten organisations and launched last week in Geneva.
Picture a smallholder farmer in rural India. Her crops are dying. She photographs the sick plant, hoping to look it up online, but every useful search result comes back in English, a language she does not speak.
That single image is the founding argument of Current AI, a nonprofit that wants to do for artificial intelligence what Tim Berners-Lee did for the web: make it free, open, and available to everyone on earth.
The organisation, first reported by TechCrunch, was founded in February 2025 by Martin Tisne. Its CEO is Ayah Bdeir, who previously led Mozilla's AI strategy and founded littleBits, a STEM education company that sold to Sphero in 2019 after reaching millions of children.
The money behind it is serious. The French government committed $100 million to seed the fund. The Ford Foundation, MacArthur Foundation, Google DeepMind, and Salesforce followed. Total committed funding now stands at $400 million. "They're not investors; they're funders," Bdeir said. The distinction matters: no one is expecting a financial return.
So what is Current AI actually building?
At the India AI Summit in February, it partnered with Bhashini, the Indian government's AI language division, to create Suno Sutra (Hindi for "listening chronicles"). It is a small, pocket-sized hardware device that runs AI models, meaning the software that powers voice and text tools, across 22 Indian languages. No internet required.
Last month it awarded $3.2 million across four grantees. Masakhane in Kenya is building AI training datasets in more than 50 African languages for health, farming, and education. Lebanon's Institute for Worldmaking is digitising Arab cultural history into databases controlled by communities, not tech companies. Brazil's Portal sem Porteiras is building offline AI tools with Indigenous Amazon communities, keeping all data inside those territories. Kenya's African Internet Rights Alliance is developing audit tools to hold AI systems accountable across the continent.
Who actually controls the data?
Bdeir is direct: not a Silicon Valley company. Current AI stores models and data locally, brings community experts in before anything is designed, and writes consent protocols into every project so communities can stop the process at any point.
None of the four grantees has fully solved data ownership yet. Bdeir calls that the point.
The concern is real. She notes that for Indigenous languages, missionary Bible translations often become AI training data before communities have agreed to anything.
Big Tech has multilingual AI products too, but Bdeir draws a clear line. "Big tech builds multilingual models to expand their market," she said, "regardless of consent or context."
The practical takeaway here is simple. If you work in a field, a classroom, or a clinic that serves people in minority or regional languages, the tools you eventually get access to will be shaped by who funded the underlying models. Public funding pushes that in a different direction than private profit does. Watch which way your sector's AI tools are going.
