Who Owns the Knowledge That Powers AI?

The New Extraction Economy

Digital capitalism has entered a fundamentally different phase. For twenty years, platforms monetized attention and social graphs. Now large language models mine something deeper: the accumulated knowledge, creativity, and expression of billions of internet users. Every comment thread, tutorial video, code snippet, and photograph becomes raw material for AI training. This is the token economy, where human output is transformed into data tokens that power models worth hundreds of billions of dollars.

The problem is structural. Millions of people continuously generate the raw material on which these systems depend, yet almost none shares in the wealth created. The imbalance is not accidental. It reflects a governance vacuum at the heart of how AI training data is collected, licensed, and valued.

Why Individual Rights Cannot Scale

Some argue that existing copyright and licensing frameworks are sufficient. If individuals object to their work being used for training, they can exercise their rights or withhold consent. This perspective ignores the asymmetry of power. A single creator negotiating with OpenAI, Google, or Meta is in the same position as a lone worker facing a multinational employer. They lack leverage, information, and resources to bargain on equal terms.

The scale of AI training makes individual negotiation impractical. Modern LLMs ingest trillions of tokens drawn from vast swaths of the public internet. No photographer, writer, or developer has the capacity to monitor how their work is used, much less negotiate compensation from every AI lab that trains on it. The transaction costs alone would be prohibitive. Without collective mechanisms, most creators are left with a binary choice: accept that their work will be used without compensation, or withdraw from the internet entirely.

The Precedent of Collective Management

This challenge is not new. Creative industries faced a similar dilemma over a century ago when new technologies like radio and recorded music made it impossible for individual artists to track and license every use of their work. The solution was collective-management organizations that negotiate licenses on behalf of millions of rights holders, collect royalties, and redistribute revenues according to transparent rules.

Today, the idea that a musician would negotiate separate agreements with every radio station and streaming platform seems absurd. Collective licensing infrastructure made creative industries functional at scale. The token economy requires a comparable institutional response.

Data Cooperatives as Digital Commons Governance

One promising approach is to establish data cooperatives representing internet users, creators, and other rights holders whose content feeds large language models. These cooperatives would negotiate access terms with AI developers, determine how value is shared, and ensure that members' preferences are respected. Unlike individual licensing, cooperatives can aggregate bargaining power, reduce transaction costs, and create enforceable standards across the industry.

But compensation is only part of the equation. The deeper question is who gets to decide how these systems are built and deployed. Through data cooperatives, members could collectively determine the conditions under which AI developers access their contributions. This shifts the governance model from passive extraction to active participation. Instead of being raw material, citizens become stakeholders in shaping the values embedded in AI systems and ensuring they serve public rather than purely commercial interests.

The Asian Context

Asia offers particularly fertile ground for this institutional innovation. Several factors converge. First, the region's digital populations are massive and growing. Indonesia alone has over 200 million internet users generating content daily. Second, many Asian markets have experience with cooperative and collective-ownership models in agriculture, finance, and labor. Third, governments across the region are actively developing AI strategies and regulatory frameworks. Singapore, South Korea, and Japan have each signaled interest in governance models that balance innovation with public benefit.

Jakarta, Mumbai, and Manila are not just sources of training data; they are hubs of digital creativity and knowledge production. If AI value creation is genuinely global, governance structures must reflect that geography. Data cooperatives rooted in Asian markets could negotiate not only compensation but also representation in how AI systems are designed, ensuring that models trained on diverse global content do not encode only Silicon Valley values.

Implementation Challenges

Establishing data cooperatives at scale involves significant practical hurdles. Membership criteria must be defined. What qualifies someone as a contributor to the digital commons? How are voting rights and revenue shares allocated? Cooperatives will need technical infrastructure to track data usage, verify compliance, and distribute payments. They will also need legal standing to negotiate with AI labs and enforce agreements.

Regulatory support will be essential. Governments can facilitate cooperative formation by providing legal frameworks, initial funding, and recognition of collective bargaining rights in the context of AI training data. Antitrust authorities should view data cooperatives not as cartels but as necessary countervailing power in concentrated markets. Without this enabling environment, cooperatives risk remaining marginal.

A Governance Model for the Future

The token economy is not a temporary phase. As AI capabilities expand, the demand for training data will only intensify. Future models will require multimodal inputs: text, images, video, audio, sensor data, and more. The scope of extraction will widen. Without new governance institutions, the gap between those who produce knowledge and those who capture its value will continue to grow.

Data cooperatives offer a path toward more equitable and democratic AI governance. They recognize that collective human knowledge is a commons, not a free resource to be enclosed by whoever can scrape it first. They provide a mechanism for negotiation, accountability, and shared value creation. Most importantly, they give ordinary internet users a voice in shaping the AI systems that will define the next era of digital life.

The question is no longer whether AI will transform economies and societies. It will. The question is whether that transformation will be governed by a handful of firms extracting value from billions of people, or by institutions that recognize the collective nature of the knowledge on which AI depends. The answer will determine not just who profits, but what kind of digital future we build.