The Legal Argument
Moral arguments establish what is right. Legal frameworks establish what is enforceable. Governing the AI commons requires both and the legal theories that best describe what is actually happening are not the ones currently being applied.
Copyright was built for one-to-one copying.
Copyright infringement the theory most commonly applied to AI training disputes requires demonstrating substantial similarity between a protected work and an allegedly infringing output. This test was designed for one-to-one copying and is poorly suited to the large-scale distillation of patterns from millions of works.
The harm at issue in AI training is not the reproduction of specific expressions. It is the systematic extraction of value from contributions to the knowledge commons without compensation or governance participation for the communities that made those contributions. Copyright, as currently constructed, cannot adequately address this harm.
Value extracted from the commons and retained without corresponding obligation.
Unjust enrichment requires demonstrating that a defendant was enriched, that the enrichment came at the plaintiff's expense, and that retaining the benefit without compensation would be unjust.
Applied to AI training: organizations that train on the commons are enriched by the resulting capabilities; that enrichment derives from the aggregated contributions of knowledge producers who received nothing for it; and retaining the full commercial benefit without contribution to the commons or its governance is unjust given the absence of meaningful consent or participation.
Some resources cannot be privately enclosed.
American law recognizes that certain resources are held in trust for the public and cannot be alienated by any government or private actor. The public trust doctrine holds that navigable waters, the atmosphere, and certain other natural resources are common to all, incapable of private appropriation.
The accumulated representational structure of human knowledge, made computationally explicit through AI training, qualifies for analogous treatment. It was not created by any single generation and cannot be alienated by any single generation without violating the intergenerational obligations that commons governance imposes. The public trust framework provides a constitutional anchor for commons claims that does not depend on finding an infringed copyright or an enforceable contract.
Empirically derived conditions for commons sustainability.
Elinor Ostrom's empirical research on long-enduring common-pool resource institutions identified the governance conditions under which commons resources are sustainably managed. Her findings drawn from fisheries, irrigation systems, and forests translate directly to the AI knowledge commons and provide an evidence-based framework for institutional design.
The relevant conditions: clearly defined boundaries between commons resources and proprietary ones; proportional equivalence between benefits derived from the commons and contributions made to its maintenance; collective governance arrangements that give affected communities real decision-making power; monitoring and graduated enforcement mechanisms; and nested governance structures that operate at multiple scales.
These are not abstract ideals. They are empirically derived conditions for commons sustainability and they provide a blueprint for AI governance institutions that is grounded in evidence rather than ideology.