Current AI wants a public AI stack, starting with language inclusion

Lead

Most AI stories revolve around model size, product launches, or funding rounds. Current AI is taking a different path. The nonprofit is trying to create a public AI layer that feels closer to early web infrastructure: open, shared, and available to communities that have usually been left out.

Key points

• Founded in February 2025 by Martin Tisne, Current AI positions itself as a public-interest AI nonprofit.
• The French government seeded it with $100 million, and commitments from Ford Foundation, MacArthur Foundation, DeepMind, and Salesforce bring total pledged funding to $400 million.
• In partnership with Bhashini, India’s government language AI division, it helped launch Suno Sutra, an offline pocket device supporting 22 Indian languages.
• Its first grant cohort allocated $3.2 million to four organizations working in Kenya, Lebanon, and the Brazilian Amazon.
• In Geneva, it launched Alpha Chat, an open-source chatbot assembled in seven weeks by 10 organizations including Hugging Face, Mozilla, and MIT Media Lab.

What Current AI is really arguing

The nonprofit’s thesis is simple: if AI will shape nearly every part of life, there needs to be a public alternative to systems owned by a handful of private companies. Ayah Bdeir, who joined as CEO after leading Mozilla’s AI strategy, frames the goal as building something closer to the World Wide Web—an infrastructure that anyone can use without paying rent to a single platform.

That framing matters because the current AI market is built around concentration. The biggest systems come from companies with tight control over models, data pipelines, and product direction. Current AI is pushing in the opposite direction: local storage, community consent, open sourcing, and shared governance.

Why language is central

For Current AI, language is not just a user interface. It is the vessel for memory, tradition, and identity. That is why the organization starts with multilingual access, especially in places where English-centered AI systems fail to reflect local realities.

The Suno Sutra device captures that approach well. It works offline, supports 22 Indian languages, and is open source so developers can build on it. In settings with limited connectivity, that is not just a convenience feature; it is the difference between participation and exclusion.

Broader impact

The grant portfolio shows how that philosophy translates into practice. Masakhane is building datasets across more than 50 African languages for health, farming, and education. Lebanon’s Institute for Worldmaking is digitizing Arab cultural history in forms communities can control. Portal sem Porteiras is building offline AI tools with Indigenous Amazon communities. And the African Internet Rights Alliance is developing audit tools for accountability.

The scale is still modest, and many problems remain unresolved. But that is also the point. Current AI is not claiming to have solved ownership, consent, or representation. It is trying to make those questions part of the system from the start.

The bigger significance is strategic: AI’s next frontier may not be just smarter models, but public infrastructure that reflects more of the world than the dominant market currently serves.

TechCrunch AI