Current AI Foundation Wants to Build a Public Alternative to Big Tech's AI Models

Non-profit Current AI, founded by Martin Tisné in February 2025, is building open artificial intelligence infrastructure as an alternative to systems controlled by major tech companies. The organization has raised $400 million so far out of a planned $2.5 billion over five years, and has already distributed its first grants for AI projects in languages that large models don't support.

Since January, Current AI has been led by Ayah Bdeir, previously responsible for AI strategy at Mozilla and founder of littleBits, an education company sold to Sphero in 2019. Bdeir describes the organization's mission as building a public equivalent of the World Wide Web for artificial intelligence, available to everyone for free.

If AI is really going to change every aspect of someone's life, there has to be a public alternative. Like the World Wide Web, available to everyone, for free - Ayah Bdeir, CEO of Current AI

Where the money comes from

Current AI's funding follows a mixed public-private model. The French government has committed $100 million, joined by the Ford Foundation, the MacArthur Foundation, and two commercial companies with a direct stake in AI development, DeepMind and Salesforce. In total, $400 million has been raised toward a $2.5 billion target for the next five years.

Bdeir stresses that the relationship with donors differs from a typical Silicon Valley investor arrangement. The funds are meant to support projects that the commercial market overlooks because they don't generate enough user scale or return on investment, rather than to maximize market share.

Languages big models overlook

The organization's flagship project is Suno Sutra, a pocket-sized device that works without internet access and supports 22 Indian languages. The name, loosely translated from Hindi, means roughly "the story that is heard." The project was developed in February in partnership with Bhashini, the Indian government's language technology unit, and was presented at the India AI Summit.

Bdeir points out that the large language models built by major tech companies cover only a fraction of the world's languages, and decisions about which languages make it into training data are made without consulting the communities that speak them.

Big tech builds multilingual models to expand its market, regardless of consent or context. For indigenous languages, missionary translations of the Bible become training data before communities get a chance to set any rules - Ayah Bdeir, CEO of Current AI

The first grant round

In its first round, Current AI distributed $3.2 million last month among four organizations. Masakhane, from Kenya, is building AI datasets for more than 50 African languages, used in healthcare, agriculture and education. The Lebanese Institute for Worldmaking is digitizing Arab history and cultural practices. Portal sem Porteiras, from the Brazilian Amazon, is creating offline tools for indigenous communities, and the African Internet Rights Alliance, also from Kenya, is developing tools to audit AI systems across the African continent.

There's also Alpha Chat, an open-source chatbot unveiled in Geneva, built in seven weeks by a coalition of ten organizations, including Hugging Face, Mozilla and MIT Media Lab. Current AI has also signed an agreement with Tokyo-based startup Sakana AI to jointly build an open technology stack supporting the Japanese language and culture, as well as communities across the Global South.

What it means for smaller languages

Current AI's operating model rests on three pillars: funding projects overlooked by the commercial market, building a complete AI technology stack from hardware to applications, and connecting the global movement for public-interest AI through shared infrastructure. Bdeir stresses that the scale major tech companies chase doesn't have to be the only measure of success.

Scale isn't always the measure of success. That's big tech's paradigm - Ayah Bdeir, CEO of Current AI

For languages like Polish, present in large models but still treated as secondary to English, Current AI's approach points to an alternative path for developing AI tools, one grounded in local data ownership and community oversight of how a given language and culture are represented in models. Poland's SpeakLeash Foundation, creator of the Bielik model, has taken a similar direction, though on a smaller scale and without billion-dollar international funding.

The organization has announced further grant rounds and continued development of its technology stack, aiming to cover more regions and languages overlooked by commercial AI labs. Current AI's ultimate goal is to raise $2.5 billion over five years, which would make it one of the largest philanthropic initiatives focused exclusively on open AI infrastructure.