In rural India, a farmer captures an image of a wilting plant. She wishes to investigate it online, but English is not her language. There should be no barriers.

This is precisely the challenge that the nonprofit organization Current AI aims to address by creating open, public AI infrastructure. In February at the India AI Summit, it collaborated with Bhashini, the AI language division of the Indian government. This partnership led to the creation of Suno Sutra, which translates to “listening chronicles” in Hindi, a compact, offline device that operates AI in 22 Indian languages without needing an internet connection. “With hundreds of languages and dialects in India, AI currently does not encompass them all,” Current AI CEO Ayah Bdeir remarked in a TechCrunch interview. The device is open-source, inviting developer communities to engage with it.

Founded in February 2025 by Martin Tisne, the nonprofit is making rapid progress. Last month, it allocated $3.2 million in grants to various initiatives within four organizations; most recently (last week), it introduced an open-source AI chatbot at the AI for Good Summit in Geneva.    

Bdeir, who joined in January after heading Mozilla’s AI strategy, previously established littleBits, the STEM education platform that impacted millions of children before its sale to Sphero in 2019.

Current AI operates as a “public-private partnership,” uniting governments, corporations, and philanthropic entities to finance technology that serves the public interest, she informed TechCrunch.   The French government initially funded Current AI with $100 million, alongside contributions from the Ford Foundation, MacArthur Foundation, DeepMind, and Salesforce — leading to a total funding commitment of $400 million. “They are not investors; they are funders,” Bdeir emphasized.

The objective it seeks to tackle is clear: all prominent AI systems today, from OpenAI to Google to Anthropic, are owned by private corporations. “If AI is indeed a groundbreaking technology that will alter every facet of our lives, a public alternative must exist,” Bdeir stated. “Similar to the World Wide Web, accessible to everyone, free of charge.”

Half of the world’s spoken languages are at risk of disappearing. “With English dominating the largest language models and AI systems, many of the world’s languages, and by extension, their cultures and communities, are being neglected,” Bdeir noted.

In response to inquiries about Big Tech’s efforts in multiple languages, Bdeir sharply differentiated their motives. “Big tech constructs multilingual models to broaden their market,” she asserted, “without regard for consent or context.” The outcomes are tangible. “For Indigenous languages, translations of religious texts end up being used as training data before communities have established any guidelines,” she explained.

Beyond just language

An AI’s capability to communicate in a language is merely one aspect it must master. “Language is the vessel through which knowledge, tradition, memory, and identity are transmitted through generations. Thus, when technology cannot speak your language, it cannot preserve your culture,” she pointed out.

Her vision for Current AI is an open system inspired by the early web, where advancements benefit all, no one is excluded, and communities maintain control over their own data.

The first grant round by Current, announced last month, involved distributing $3.2 million to four organizations spanning Kenya, Lebanon, and the Brazilian Amazon.

The initiative in Masakhane, Kenya, focuses on creating AI datasets in over 50 African languages for sectors such as health, agriculture, and education; Lebanon’s Institute for Worldmaking is digitizing Arab cultural heritage and modern practices into machine-readable databases controlled by communities (not tech firms). Brazil’s Portal sem Porteiras is developing offline AI resources in collaboration with Indigenous Amazon communities, ensuring data remains within their territory. Meanwhile, Kenya’s African Internet Rights Alliance is creating auditing tools to hold AI systems accountable throughout the continent.

Data ownership matters

Regarding the issue of data ownership, Bdeir expressed her thoughts candidly. “Several models and proposals exist about data ownership within different communities, but one thing is crystal clear: it should not be in the hands of a Silicon Valley company aiming to enrich a select few,” she told TechCrunch.

The approach of the nonprofit is to store models and data on a local level, engaging community experts prior to any development, or incorporating consent protocols into the process so that communities can halt it whenever they choose.

So far, none of Current AI’s grant recipients have fully resolved the issue. However, Bdeir considers this the goal. “Each of them has integrated the question into their projects,” she stated, “rather than settling for the typical default, where complexity becomes an excuse for a government or a tech company to make decisions for everyone.”

As for the impact that can arise from a $3.2 million budget split among four organizations, Bdeir remarked, “Scale is not always the benchmark. That’s the paradigm of Big Tech,” she noted. “This could manifest as an Indigenous elder in the Brazilian Amazon utilizing a tool developed in Kenya to convey ecological wisdom in their own language.”  

Constructing the stack

Earlier this month in Geneva, Current introduced Alpha Chat, an open-source chatbot created in just seven weeks by a coalition of ten organizations, including Hugging Face, Mozilla, and MIT Media Lab. Each participant contributed a component of the stack, including a language model, safety tools, and computing capabilities.

Current AI also established a partnership with Sakana AI, a Tokyo-based startup recognized for its endeavors in what it labels Sovereign AI. The two organizations intend to develop a shared open-source AI stack, designed to cater not only to the Japanese language and culture but also to communities in the Global South that have largely been overlooked by dominant AI systems.