The Territory
What kind of resource is AI, and what governance principles follow from its nature? The answer requires understanding where AI capabilities actually come from and recognizing what that origin means for how these systems should be governed.
The sufficient condition is access to the accumulated knowledge of humanity.
Large-scale AI systems are trained on the recorded output of human civilization. Scientific literature, legal reasoning, historical narrative, creative works, technical documentation, and the accumulated exchange of ideas across cultures and centuries constitute the corpus from which these systems derive their capabilities.
The conceptual relationships they learn, the reasoning structures they internalize, and the patterns of understanding they approximate were not invented by any engineering team. They were produced through generations of human intellectual effort.
Engineering effort and compute investment are necessary conditions for building these systems. They are not sufficient conditions. The sufficient condition is access to the accumulated knowledge of humanity. Remove the human knowledge commons from the equation and the engineering investment produces nothing. Reduce its diversity and richness and the capabilities degrade.
AI systems are neither purely private inventions nor purely public resources.
Recognizing the commons origin of AI capability does not diminish the genuine achievement that large-scale AI development represents. Advances in algorithm design, distributed computing, hardware optimization, and large-scale system architecture require substantial expertise and sustained financial investment. These contributions are real, they generate value, and they legitimately support commercial return.
What they do not do is transform the underlying knowledge from which AI systems learn into private property. AI systems are hybrid entities. Their capabilities arise from the interaction between private technical investment and collectively produced knowledge.
The representational structure of human knowledge was not created by the training process. It was made computationally legible by it.
Intellectual property law has long distinguished between discovery and invention. Laws of nature, mathematical relationships, and abstract ideas are not patentable precisely because they are discovered rather than invented. What already exists in the world cannot be owned by whoever first identifies it.
The representational structure of human knowledge, approximated by AI systems through training, falls on the discovery side of this line. Neural networks trained on human knowledge converge toward shared internal representational structures that reflect the relational organization of human understanding. These structures were not created by the training process. They were made computationally legible by it. The distinction matters for how IP claims over AI systems should be understood and bounded.
Understanding what AI is clarifies what the governance question actually is.
Understanding AI as a hybrid system built from both collective knowledge and private investment does not automatically resolve the governance question. It does, however, clarify what the governance question actually is.
It is not primarily a question about the regulation of technology. It is a question about the governance of a commons a shared resource, created collectively over centuries, that is now being processed at scale for private commercial benefit without corresponding obligation to the communities that created it.
The standard commercial and legal frameworks applied to AI development were not designed for resources with these characteristics. Applying them without modification produces enclosure as a structural outcome not an exception, but the predictable result of applying standard logic to an unusual resource.