Current AI’s ‘Public Web’ Vision: Challenging the Gravity of Private Power

The Price of Access in a Polyglot World

The vision of a “World Wide Web of AI,” freely available and culturally resonant for everyone, stands in stark contrast to the closed, proprietary systems that currently dominate the artificial intelligence landscape. Current AI, a nonprofit founded by Martin Tisne in February 2025, champions this ambition, aiming to forge genuinely public AI infrastructure. With committed funding reaching $400 million, including significant seed money from the French government and contributions from entities like the Ford Foundation, MacArthur Foundation, DeepMind, and Salesforce, the organization is moving with impressive velocity.

Ayah Bdeir, Current AI’s CEO, articulate that the very notion of a transformative technology like AI necessitates a public alternative. The issue is immediate and profound: half the world’s spoken languages face extinction, and as Bdeir points out, English-centric large language models actively accelerate this cultural erosion. Projects like Suno Sutra, an offline device supporting 22 Indian languages developed in partnership with India’s Bhashini, demonstrate a tangible commitment to linguistic diversity. Similarly, grants to organizations in Kenya, Lebanon, and Brazil aim to build culturally specific AI datasets and tools for over 50 African languages and indigenous Amazonian communities.

This is not merely about translation; it’s about digital colonialism. Big Tech’s multilingual pushes, Bdeir argues, are driven by market expansion “regardless of consent or context,” leading to scenarios where sacred texts become unwitting training data for algorithms before communities establish their own rules for engagement. Current AI’s approach—prioritizing local data storage, community experts, and explicit consent protocols—is a direct challenge to the extractive models of Silicon Valley. Yet, the notion that a “public-private partnership” will inherently insulate Current AI from the gravitational pull of its corporate funders, like DeepMind and Salesforce, is a convenient fiction, overlooking the subtle ways established power can shape even altruistic endeavors.

The Architecture of an Open Future

Building an open-source AI stack from disparate components, as Current AI has done with Alpha Chat in just seven weeks alongside a coalition including Hugging Face and MIT Media Lab, is a testament to the power of collaborative development. This rapid assembly of a functional chatbot and the partnership with Sakana AI for a sovereign Japanese AI stack underscore the technical feasibility of their vision. However, the path from technical prototype to widespread adoption and sustainable infrastructure is fraught with structural challenges that the original World Wide Web, with its comparatively simpler technical stack and less fragmented digital landscape, never quite faced in the same way.

The early web flourished due to a set of universally accepted, open protocols and a relatively low barrier to entry for content creation and infrastructure deployment. Today, the generative AI space is dominated by colossal, proprietary machine learning models and massive compute resources, making true decentralization and interoperability exceedingly complex. Who defines the standards for this new “public AI”? How are conflicts resolved, and how is long-term maintenance funded without either centralizing power or fragmenting into incompatible silos? These aren’t just technical questions; they are fundamental issues of data governance and algorithmic ethics.

Big tech companies like DeepMind and Salesforce are contributing to Current AI not purely out of altruism; rather, they are securing a strategic foothold. Their involvement allows them a seat at the table in defining future public AI standards and helps mitigate potential regulatory pressure by demonstrating a commitment to “responsible” AI. This allows them to shape the narrative around “public AI” without fundamentally altering the core business model of their own proprietary models, ensuring they remain key players in any emerging ecosystem.

Redefining Scale Beyond Silicon Valley Metrics

Bdeir’s rejection of “scale” as the ultimate measure—calling it “the Big Tech paradigm”—is perhaps Current AI’s most radical proposition. She envisions success not in user numbers or market share, but in an Indigenous elder in the Brazilian Amazon passing down ecological knowledge in their own language via a locally developed tool. This human-centric metric reorients the entire discussion around the utility and cultural preservation of technology, rather than its monetization potential.

Yet, the practicality of sustaining such a distributed, context-specific network without the economies of scale that attract capital and talent remains a formidable hurdle. The existing digital world is a deeply embedded ecosystem of powerful, proprietary platforms, each vying for user attention and data. Can a public alternative, however noble in its aims, compete effectively for mindshare and resources without some form of strategic growth? The goal is not just to build alternatives, but to make them viable, attractive, and resistant to absorption by the very forces they seek to counterbalance.

Current AI’s commitment to community control and bespoke solutions for specific cultural contexts offers a compelling counter-narrative to the prevailing AI orthodoxy. However, the real work lies in transforming this noble intent into a durable, self-sustaining global network that actively resists the centralization and commercialization inherent to so much of today’s technology. The aspiration is clear: a truly open AI. The path to building it, while avoiding the pitfalls of existing power structures, is anything but.