Current AI’s offline language device with Bhashini puts Indian AI inclusion in focus

Current AI, a non-profit working on public-interest AI infrastructure, has reportedly teamed up with Bhashini on Suno Sutra, a pocket-sized offline device designed to run AI in 22 Indian languages without an internet connection.

The development, reported by TechCrunch, gives Indian readers a concrete example of a broader debate: whether AI systems should be built mainly as private products for connected, high-value users, or also as public infrastructure for communities that need local-language access, offline tools and stronger control over data.

For India, the example is especially relevant because language coverage remains one of the biggest barriers to AI adoption outside English-speaking urban users. The source frames Suno Sutra around a practical use case: a farmer in rural India trying to identify a failing plant but unable to rely on English-language search or always-on connectivity.

Datos clave

Item	Detail
Organisation	Current AI, a non-profit founded in February 2025
India link	Collaboration with Bhashini on Suno Sutra
Device claim	Offline AI device supporting 22 Indian languages
Wider funding	TechCrunch reports $400 million in committed funding for Current AI

Why Suno Sutra matters for Indian users

According to TechCrunch, Current AI and Bhashini worked together after the India AI Summit in February on Suno Sutra, whose name is described as Hindi for “listening chronicles”. The device is said to be open-sourced so developer communities can build on it.

The stated goal is not only translation. Current AI chief executive Ayah Bdeir told TechCrunch that India has hundreds of languages and dialects, and that current AI systems do not adequately represent them. That distinction matters for India’s AI ecosystem because a model that can produce text in a language may still fail on local agricultural vocabulary, dialects, oral knowledge, cultural references or safety-sensitive advice.

Offline access is another important part of the claim. Many AI services assume constant internet connectivity, cloud processing and a smartphone-first workflow. If a device can run locally, it could be more useful for rural education, farming support, public services and community-led language projects. The report does not provide device specifications, deployment numbers or independent testing results, so those details remain to be checked through official releases or open-source repositories.

Current AI’s public infrastructure pitch

Current AI was founded in February 2025 by Martin Tisne and is positioning itself as a public-private partnership for AI. TechCrunch reports that the French government seeded it with $100 million, with support from the Ford Foundation, MacArthur Foundation, DeepMind and Salesforce bringing committed funding to $400 million.

Bdeir, who joined as chief executive in January after leading Mozilla’s AI strategy, described the backers as funders rather than investors. The organisation’s argument is that transformative AI should have a public alternative, not only private systems owned by large technology companies.

That framing is likely to resonate with policy and developer communities in India, where language AI, digital public infrastructure and sovereign technology choices are already active areas of debate. However, Current AI’s model is still early. The available report highlights ambition and partnerships, but it does not yet show long-term adoption metrics, governance outcomes or whether public-interest AI tools can match the reliability and support expectations of production systems.

Grants focus on community-controlled AI

Beyond India, Current AI has announced a first grant round of $3.2 million across four organisations, according to the TechCrunch report. The projects span Kenya, Lebanon and the Brazilian Amazon, with a focus on community-controlled datasets, offline tools and AI accountability.

One project with Masakhane in Kenya is aimed at datasets across more than 50 African languages for areas such as health, farming and education. Lebanon’s Institute for Worldmaking is working on machine-readable cultural databases intended to remain under community control. Portal sem Porteiras is building offline AI tools with Indigenous Amazon communities, while the African Internet Rights Alliance is developing audit tools for AI systems.

The relevance for India is the design principle: language data is not just raw material for model training. It can include memory, oral history, farming practices, local governance terms and community knowledge. If such data is collected without consent or context, AI systems may scale extraction rather than inclusion.

Open-source chatbot and Sakana AI deal

Current AI has also launched Alpha Chat, described by TechCrunch as an open-source chatbot assembled in seven weeks by a coalition of ten organisations, including Hugging Face, Mozilla and MIT Media Lab. The report says contributors brought different parts of the stack, including model components, safety tooling and computing power.

The non-profit has also struck a deal with Tokyo-based Sakana AI to work on a shared open-source AI stack. The collaboration is described as supporting Japanese language and culture, while also being relevant to communities across the Global South that are under-served by dominant AI systems.

For Indian developers, the open-source angle is worth watching. If code, models, datasets and safety tools are released with usable documentation and clear licences, such projects could become building blocks for regional-language applications. If they remain mostly announcements, their practical value will be harder to assess.

What to watch next

The main unanswered questions are practical ones. Current AI and Bhashini’s Suno Sutra claims need more detail on hardware availability, open-source repositories, licensing, supported language quality, privacy safeguards, offline model performance and whether Indian community developers can actually test and extend the device.

For Indian startups, civil society groups, edtech teams and public-sector technologists, the story is not only about one device. It is about whether AI infrastructure can be designed for users who are not English-first, always online or already well served by commercial platforms.

Source: TechCrunch AI, https://techcrunch.com/2026/07/19/nonprofit-current-ai-is-racing-to-build-the-world-wide-web-of-ai-free-for-all/