Current AI Deploys First $3.2M From $400M in Commitments for Public-Interest AI

Current AI, the nonprofit that launched at France’s AI Action Summit in February 2025 with $400 million in commitments to build a public alternative to Big Tech’s AI, has started writing checks. Last month it deployed $3.2 million to its first four grantees, and in recent months it shipped its first products: an offline, multilingual device built with the Indian government in February and, earlier this month, an open-source chatbot unveiled at a United Nations summit in Geneva. Seventeen months after the pledges landed, the capital is finally moving.

The distance between what has been promised and what has been spent is the story here. Current AI counts more than $400 million committed and a target of mobilizing $2.5 billion over five years, but the $3.2 million first cohort is what has actually gone out the door. CEO Ayah Bdeir frames the backers as donors rather than return-seeking capital. “They’re not investors; they’re funders,” she told TechCrunch.

Who is putting up the money

Current AI is structured as a public-private partnership: a vehicle that pools government, philanthropic, and corporate money rather than an operating company or a fund answering to limited partners. The French government seeded it with $100 million, joined by Salesforce (CRM ), Google’s DeepMind, and the Ford and MacArthur foundations, bringing the initial commitment to $400 million. Founder Martin Tisné, who also runs the Omidyar-backed AI Collaborative, pitched the effort at launch as a way to concentrate a fragmented field of public-interest funding at a scale large enough to matter.

That scale is modest by the industry’s standards, and deliberately so. The partnership is not building data centers or buying GPUs. France’s own private-sector AI package, announced the same week, ran to roughly $112 billion, and the US Stargate venture was pitched at $500 billion. Where Mistral and other European players are racing for compute, Current AI is spending on datasets, open tooling, and grants in markets private capital has no reason to serve.

What the money is buying

The first cohort, announced last month, split $3.2 million across four organizations building locally controlled AI:

• Masakhane in Kenya, assembling datasets across more than 50 African languages for health, farming, and education;
• the Institute for Worldmaking in Lebanon, turning Arab cultural history into machine-readable databases that communities, not companies, control;
• Portal sem Porteiras in Brazil, building offline AI tools with Indigenous Amazon (AMZN ) communities and keeping the data inside the territory;
• the African Internet Rights Alliance in Kenya, developing tools to audit AI systems across the continent.

What unites the cohort is control: data and models stay in community hands rather than on a distant company’s servers, and every project bakes in consent checkpoints that let the community stop the work at any stage. None of the grantees has fully solved the data-ownership problem, but Bdeir argues that building the question in is the point.

The products followed. Earlier this month in Geneva, Current AI launched Alpha Chat, an open-source chatbot assembled in seven weeks by a coalition of ten organizations including Hugging Face, Mozilla, and the MIT Media Lab, each contributing a piece of the stack — a model, safety tooling, or compute. In February 2026, working with Bhashini, the Indian government’s language-AI division, it produced Suno Sutra, a pocket-sized device that runs AI in 22 Indian languages with no internet connection. It has also struck a deal with Sakana AI, the Tokyo startup building what it calls sovereign AI, to develop a shared open-source stack for Japanese and Global South languages. The effort to serve languages Big Tech overlooks tracks a wider wave of projects aimed at underserved tongues.

Why a public option, and why now

The thesis is a capital argument as much as a technical one. Every leading system — from OpenAI to Google to Anthropic — belongs to a private company, and Bdeir contends that a technology this consequential needs a public alternative, free to anyone, the way the early web was. Left to the market, she argues, models get built to widen a vendor’s addressable market, not to serve communities whose languages and data carry little commercial value.

Whether $3.2 million in grants and a seven-week chatbot add up to a credible alternative is the open question, and Bdeir has an answer ready for anyone measuring her by venture math. “Scale is not always the measure. That is the Big Tech paradigm,” she said. The bet is that small, community-owned projects can seed an ecosystem that compounds over time — a case that rhymes with the argument for middle powers building their own models instead of renting them.

The leadership hire signals intent to execute. Bdeir, who joined in January 2026, previously ran Mozilla’s AI strategy and founded the STEM-hardware company littleBits before selling it to Sphero in 2019. The harder test is conversion: turning $400 million in commitments and a $2.5 billion target into deployed capital and working infrastructure fast enough to matter, while the private AI stack keeps extending its lead.