Executive Summary

Recent federal district court decisions in Kadrey v. Meta and Bartz v. Anthropic correctly recognized AI training as transformative fair use, finding it "quintessentially transformative." However, these rulings reveal concerning judicial attempts to fragment fair use analysis through novel requirements threatening innovation. This article argues AI training satisfies the four-factor fair use test by converting expressive text into mathematical representations serving entirely different purposes than original works. Unlike technologies facilitating direct copying, AI systems predominantly generate new content based on learned patterns, qualifying for protection under Sony Corp. v. Universal City Studios' substantial non-infringing use doctrine.

The article critiques Judge Chhabria's "market dilution" theory, which threatens to transform copyright from protecting expression into anti-competition regulation. Empirical evidence supports fair use: book sales increased 6.5% during 2024 despite widespread AI deployment, demonstrating complementary rather than substitutive functions. Restricting AI development would inflict economic damage while advantaging authoritarian competitors facing no such constraints. Courts must embrace transformative use analysis while rejecting requirements that handicap American innovation in this defining technological competition.