Revealing Purpose and Character in the Age of AI A Framework for Anchoring Fair use Factor One in Foundation-model Copyright Disputes

Foundation-model copyright disputes expose a framing problem at the center of fair use factor one. Section 107(1) asks for the "purpose and character of the use," but foundation models are not bounded tools with a single stable function. They are general-purpose systems built from upstream copying of expressive works and deployed across a wide range of downstream uses, some expressive and some not.

That structure allows litigants to describe the relevant "use" at radically different levels of abstraction-training mechanics, model construction, platform creation, product deployment, or broad social innovation-with outcome-determinative consequences for transformativeness analysis. This Article proposes a revealed-purpose framework for anchoring factor one in foundationmodel cases. Where the challenged copying supports a general-purpose, user-contingent system, courts should identify the operative use and infer its purpose and character from observable developer-side evidence, including monetization architecture, design and optimization choices, deployment patterns, marketing, and market function. This approach does not deny that foundation models may support many socially valuable applications.

It insists only that factor one remain determinate, falsifiable, and tied to real-world product function rather than aspirational narratives or the full menu of conceivable uses. The Article also argues that generative AI’s competitive posture cannot be postponed to factor four without distorting fair use analysis. When developers copy expressive works to build and monetize systems that generate expression in adjacent creative markets, that competitive orientation is part of the character of the use under ß 107(1), not merely a downstream harm theory. Treating competition as irrelevant to factor one while discounting diffuse displacement under factor four risks making the central fairness question disappear between the factors. Finally, the Article explains why purpose-framing carries unusual weight in AI cases because large-scale training is difficult to unwind once completed.

In that setting, fair use often functions not only as a permission rule but also as a baseline allocation rule for who bears the cost of the expressive inputs that made foundation-model capabilities possible. A revealed-purpose approach keeps factor one coherent in the age of AI by requiring courts to evaluate what developer-side copying is actually for in the market, rather than what the technology can be said to do in the abstract.