Fair use — the doctrine that permits certain unlicensed uses of copyrighted works — runs through nearly everything Matthew Sag has written. His fair use scholarship spans theory (what the doctrine is for), history (where it came from), empirics (what actually predicts fair use outcomes), and application (what fair use means for search engines, text mining, and AI training). His casebook, Copyright Law in the Age of AI, devotes six chapters to the doctrine.
History: fair use is older than you think
The Pre-History of Fair Use, 76 Brooklyn Law Review 1371 (2011), showed that fair use does not begin with American cases like Folsom v. Marsh (1841), as standard accounts assume. The doctrine’s roots lie a century earlier, in English abridgment cases: in Gyles v. Wilcox (1741), Lord Chancellor Hardwicke held that “a real and fair abridgment” was a new book, reflecting the “invention, learning, and judgment” of its maker. Copyright and fair use evolved together from the beginning — fair use is not a modern exception grafted onto copyright, but part of copyright’s original architecture. Federal courts have drawn on this history, including the Ninth Circuit in Monge v. Maya Magazines, 688 F.3d 1164 (9th Cir. 2012), and Geophysical Services v. TGS-NOPEC (S.D. Tex. 2017), which engaged the article’s account of fair abridgment as a case-by-case, fact-intensive question.
Empirics: fair use is more predictable than its critics claim
Predicting Fair Use, 73 Ohio State Law Journal 47 (2012), systematically tested what actually predicts fair use outcomes in litigation, focusing on case characteristics visible to litigants before trial. The study provided new empirical evidence for the centrality of transformative use, and undermined the common assumptions that commercial users rarely win and that fair use functions as a subsidy for the sympathetic. The evidence shows a doctrine more rational and consistent than its “unpredictability” critique suggests.
Theory and reform
God in the Machine: A New Structural Analysis of the Fair Use Doctrine, 11 Michigan Telecommunications & Technology Law Review 381 (2005), offered a structural account of the doctrine; Beyond Abstraction, 81 Tulane Law Review 187 (2006), placed fair use within a law-and-economics account of copyright scope. Two decades later, Recodifying Fair Use, 73 Journal of the Copyright Society (forthcoming 2027), returns to the statute itself, asking how Section 107 should be rewritten in light of everything the courts have learned since 1976.
Application: fair use and the machines
The fair use work converges with Sag’s signature concept of non-expressive use: his argument that copying works for search, text and data mining, and AI training should generally be fair use shaped the litigation over Google Books and HathiTrust and now frames the generative AI cases. Fairness and Fair Use in Generative AI, 92 Fordham Law Review 1887 (2024), carries the doctrine into the generative AI era, and A Student’s Guide to the Law and Policy of AI: Fair Use and Generative AI (2025) makes it teachable.
Key publications
- God in the Machine: A New Structural Analysis of the Fair Use Doctrine, 11 Mich. Telecomm. & Tech. L. Rev. 381 (2005)
- The Pre-History of Fair Use, 76 Brooklyn L. Rev. 1371 (2011)
- Predicting Fair Use, 73 Ohio St. L.J. 47 (2012)
- Fairness and Fair Use in Generative AI, 92 Fordham L. Rev. 1887 (2024) (SSRN)
- Recodifying Fair Use, 73 J. Copyright Soc’y (forthcoming 2027)
Related
Non-expressive use · Copyright and AI training · Copyright Law in the Age of AI (casebook)