Current AI: The 'Public Option' for Open Artificial Intelligence

What happened?

In July 2025, the creation of Current AI was announced, a nonprofit organization that aspires to become the 'public option' for artificial intelligence, similar to what public libraries represent for knowledge. The initiative has initial funding of $400 million, backed by the French government and Google's AI division, DeepMind, among other partners. According to The Next Web, the idea stems from the realization that current AI models are mostly in English and do not serve the needs of rural and non-English-speaking communities. The announcement was made at the AI for Good summit in Paris, with representatives from governments, academia, and tech companies.

Why is it important?

Until now, AI development has been dominated by large tech corporations like OpenAI, Google, and Meta, whose models are often proprietary and require costly subscriptions or licenses. Current AI aims to change this paradigm by offering open language models accessible to anyone, regardless of language or geographic location. This is especially relevant for regions like Africa, Asia, and Latin America, where language and economic barriers limit access to these technologies. For example, a farmer in India who speaks Hindi or Telugu cannot fully leverage ChatGPT or Gemini, which work best in English. Current AI plans to train models in dozens of languages, including Swahili, Quechua, and Bengali, aiming to cover over 4 billion people currently underrepresented in AI.

Historical context

The initiative recalls open-source projects like Linux or Apache, which democratized access to software. However, in AI, open models like Meta's Llama or BLOOM have had limited impact due to lack of sustained funding and institutional support. BLOOM, launched in 2022 by the BigScience initiative, was a collaborative effort to create a multilingual model but lacked long-term financial backing and saw modest adoption. Current AI aims to overcome these limitations with transparent governance and public-private funding. It also differs from previous initiatives by focusing on infrastructure: it will not only develop models but also curated datasets, evaluation tools, and open-access platforms, similar to what Hugging Face has done with its model library.

Consequences and projections

If Current AI achieves its goals, it could accelerate AI adoption in sectors like education, agriculture, and healthcare in developing countries. For instance, a farmer in India could use an AI assistant in their local language to diagnose crop diseases, or a rural doctor in Kenya could access up-to-date medical information in Swahili. However, doubts remain about the quality of open models compared to proprietary ones, as well as long-term financial sustainability. The initial $400 million covers approximately three years of operation, according to organizers' estimates. By then, the project is expected to have demonstrated its value and attracted more funding from governments and foundations. Additionally, DeepMind's involvement raises questions about potential conflicts of interest: to what extent will Google allow an open model to compete with its own products? DeepMind has stated that its contribution is technical and will not involve control over governance, but skeptics point out that the tech giant could influence the project's direction.

"Current AI represents a bold step toward democratizing artificial intelligence, but its success will depend on global collaboration and transparency in its governance," noted an analyst from TheVortiq.

An additional risk is the potential digital divide: if open models are inferior to proprietary ones, they could create a "two-speed" AI, where the rich access cutting-edge models and the poor get limited versions. To avoid this, Current AI plans to launch models competitive with the best on the market, though it acknowledges this is an ambitious goal. In comparison, the open-source software movement saw Linux become the dominant operating system on servers, but it remains a minority on desktops. In AI, the gap could be even larger due to training costs and the need for massive data.

What should readers know?

• Current AI is a nonprofit with $400 million in initial funding, coming from France ($100 million), DeepMind ($50 million), and other partners like the Rockefeller Foundation and the World Bank.
• It aims to create open and free AI models accessible in multiple languages, with a focus on underrepresented languages.
• The backing of France and DeepMind lends credibility but also generates skepticism about its independence. France has promised political and regulatory support, while DeepMind will contribute research expertise.
• The project is still in its early stages; the first models are expected to be available in 2026, with an initial focus on health and agriculture.
• It could transform AI access in underserved regions but faces technical challenges (model quality) and governance issues (potential corporate influence).
• Compared to previous initiatives like BLOOM or EleutherAI, Current AI has much larger funding and unprecedented government backing, increasing its chances of success but also raising expectations.

In summary, Current AI is an ambitious experiment that could redefine the balance of power in AI, shifting from a centralized, proprietary model to a decentralized, open one. However, the path is fraught with technical, political, and financial obstacles. The coming months will be crucial to see if it can consolidate its team, define its roadmap, and start building the first prototypes. If it succeeds, it could be a milestone comparable to the launch of Linux in 1991 or the development of the World Wide Web.