Startup builds public AI to rival Big Tech

A farmer in rural India holds up her phone to a dying plant, hoping to identify the problem. She doesn’t speak English, and most AI systems barely understand her native tongue. A nonprofit called Current AI sees this gap not as a market oversight, but as a fundamental failure worth fixing.

The organization’s ambition is sweeping: create a public option for AI that is open and free to everyone, modeled after the early internet. “If it’s going to change every aspect of everyone’s life, there has to be a public alternative,” said chief executive Ayah Bdeir in an interview. “Like the World Wide Web, available to anyone, for free.”

Who is behind Current AI

Current AI launched at France’s AI Action Summit in early 2025. The French government contributed €100m, with additional backing from the Ford and MacArthur foundations, DeepMind, and Salesforce. Total commitments now exceed $400m, and the group aims to mobilize $2.5bn over five years.

Crucially, these backers aren’t looking for a financial return. “They’re not investors, they’re funders,” Bdeir explained. She previously led Mozilla’s AI strategy and founded the education-hardware company littleBits.

The initiative taps into a European preference for public alternatives to American platforms. It operates in a similar space as state-led public AI infrastructure projects, but as a global movement rather than a single nation’s agenda.

Starting with language

The first priority is preserving and supporting the world’s languages. Half of all languages spoken today are at risk of extinction, and English dominates the largest models. That, Bdeir argues, excludes most of the world’s cultures from AI entirely.

She is direct about how Big Tech fills the void. Companies build multilingual models “to expand their market, regardless of consent or context,” she said. For some Indigenous languages, she noted, missionary Bible translations become training data before communities have agreed to anything.

Current AI’s solution is to build with communities, not just for them, and to keep data local. In India, it partnered with the government’s language unit on Suno Sutra, a pocket-sized offline device that runs AI in 22 Indian languages with no internet connection. The device is open-sourced for others to adapt.

Money on the table

Spending has begun, though modestly. Last month, the group distributed $3.2m in first grants to four organizations in Kenya, Lebanon, and the Brazilian Amazon.

Projects range from datasets for over 50 African languages to offline tools that keep Indigenous Amazon data within the territory, plus audit tools to hold AI systems accountable. No grantee has fully solved the question of data ownership, Bdeir admits, but she counts embedding that question into the process as the point.

On whether $3.2m can move the needle, she challenges the premise. “Scale is not always the measure. That is the Big Tech paradigm,” she said.

Assembling a stack

The bigger bet is a full open stack, from chips to applications. This month in Geneva, Current launched Alpha Chat, an open-source chatbot assembled in seven weeks by ten organizations, including Hugging Face, Mozilla, and MIT Media Lab. Each contributed a piece: a model, safety tooling, or computing power.

It has also struck a deal with Japan’s Sakana AI to build a shared open stack for the Japanese language and for Global South communities that big models overlook. Whether a nonprofit can out-build companies giving models away for market share is the open question. Current AI is betting that a public option, however small it starts, is worth building anyway.