Nonprofit Current AI is racing to build the World Wide Web of AI, free for all

A $400 million nonprofit named Current AI, founded in February 2025, shipped its first hardware product this month and announced a $3.2 million grant round — concrete early steps in what its founders describe as a bid to build public AI infrastructure with the same open, community-controlled character as the early web.

What's Converging

The dominant AI story of the past two years has been consolidation. A small number of companies have captured the frontier model market, and the data, infrastructure, and economic incentives embedded in their systems reflect that. Models trained primarily on English-language internet text tend to perform best in English; multilingual extensions are typically retrofits aimed at expanding commercial reach rather than building genuine cultural fidelity. The result is a growing gap between what AI can do for a developer in San Francisco and what it can do for a community archivist in Beirut or a rural health worker in sub-Saharan Africa.

That gap has started drawing serious institutional attention. The Paris AI Action Summit in early 2025 made AI governance a diplomatic priority, and France committed $100 million as the seed round for Current AI at the event. UN Secretary-General António Guterres backed the launch explicitly, with Current AI's own site quoting him on the need for AI to bridge rather than widen the development gap. European Commission President Ursula von der Leyen offered public support as well. Political endorsements at that level — sitting heads of two major institutions backing a nonprofit at launch — are unusual and signal that the "public option for AI" framing has found real traction beyond the tech industry.

Meanwhile, the open-source AI ecosystem has been quietly maturing. Tools and model repositories from organizations like Hugging Face, Mozilla, and MIT have created a commons that barely existed five years ago. What has been missing is a coordination layer — an entity with enough funding and neutrality to assemble those pieces into something communities in low-resource language environments can actually deploy, own, and control. That is the specific gap Current AI is attempting to fill.

The Specific Development

Current AI's progress this month is more concrete than most nonprofit announcements. The organization revealed Suno Sutra, a pocket-sized device that runs AI inference locally across 22 Indian languages with no internet connection required. Built in partnership with Bhashini, India's government-backed language AI division, the underlying code has been open-sourced for developers to extend. The offline-first design is deliberate: in many of the communities Current AI targets, intermittent connectivity is the norm, and any tool requiring a live cloud API call will fail at the worst moment.

The second delivery is Alpha Chat, an open-source chatbot that ten organizations assembled in seven weeks. The coalition included Hugging Face contributing model infrastructure, Mozilla providing safety tooling, and MIT Media Lab on the research layer — each organization owning one distinct piece of the stack. That modular design makes individual components auditable and replaceable, a meaningful contrast to the monolithic systems commercial providers ship. The seven-week timeline is confirmed across multiple reports, and none dispute it.

On the funding side, Current AI deployed $3.2 million in grants to four projects. Masakhane, a Kenya-based research network, received support for dataset work spanning more than fifty African languages — the kind of foundational data infrastructure commercial builders have little economic incentive to create. The Lebanon Institute for Worldmaking is digitizing Arab cultural history into community-owned archives. Brazil's Portal sem Porteiras is building offline AI tools for Indigenous communities with an explicit data-sovereignty constraint: information stays within the territory where it was gathered. The African Internet Rights Alliance received funding for AI audit tools designed to catch when AI systems behave in ways that harm local populations.

Beyond the $400 million already committed, the organization's own site states it is working toward mobilizing $2.5 billion over five years — a figure the TechCrunch primary reporting does not foreground but which the currentai.org source adds directly. A partnership with Tokyo-based Sakana AI extends the shared open-source stack to Japanese-language and Global South applications, adding a commercially oriented but mission-aligned partner to a coalition that otherwise runs on grant and government funding.

Our read is that the Alpha Chat coalition model is the strategically interesting piece here. Ten organizations each owning one layer of an AI pipeline — model, safety, compute, interface — is a working proof-of-concept for something the industry has mostly theorized: a modular AI stack where no single entity controls the whole. If it holds at scale, it becomes a template others can fork. The seven-week build speed suggests that the coordination infrastructure supporting it is functional, not merely aspirational.

What's Likely Next

The immediate test is whether the coalition model holds as the stakes get higher. Safety tooling for a small open-source chatbot is a different engineering problem than safety tooling for systems serving millions of users in languages where evaluation benchmarks barely exist. Current AI has committed to building the "full stack of public interest AI, from hardware to applications" — the gap between a prototype offline device and a genuinely deployable infrastructure layer is where most open-source AI projects stall or fragment.

Watch the $2.5 billion mobilization timeline over the next 90 days. If France's seed commitment was designed to catalyze follow-on pledges from other G7 governments — which the Paris Summit framing suggested — more announcements should follow. The Sakana partnership also deserves scrutiny: Sakana is a for-profit research company, and how intellectual property flows between a commercial partner and a nonprofit infrastructure project will determine whether that alliance stays clean or introduces the same capture dynamics Current AI was built to counter. For developers who want to engage now, both Suno Sutra and the Alpha Chat stack are open-source and available to inspect, independent of how the organizational story resolves.

Sources