No Market Harm, No Case
The authors had alleged that Meta trained LLaMA on their copyrighted works without permission, in some instances allegedly acquiring the material via pirated websites. But the court was unmoved, stating that the plaintiffs didn’t prove their book sales or licensing opportunities were negatively affected by LLaMA’s development.
The judge had previously hinted that Meta’s use might be deemed “transformative,” another major pillar supporting a fair use defense. However, his final ruling rested more squarely on the lack of demonstrable economic impact.
Legal Firepower on Both Sides
The case featured a who’s who of legal heavyweights, with the authors represented by powerhouse firms such as Boies Schiller Flexner LLP, Joseph Saveri Law Firm LLP, and Lieff Cabraser Heimann & Bernstein LLP. Meta, meanwhile, was backed by Cooley LLP, Cleary Gottlieb, and Paul Weiss, among others.
None of the parties involved issued an immediate comment following the ruling.
A Precedent-Setting Moment for Generative AI
This decision may embolden tech companies developing AI to continue mining copyrighted material for training purposes, as long as their use remains transformative and doesn’t directly undercut the source material’s market. The implications stretch far beyond LLaMA, affecting every developer training generative AI on vast corpora of internet-based text.
The ruling also adds to the growing body of case law in favor of fair use in the context of machine learning, a battleground that is rapidly becoming one of the most legally consequential arenas in tech.
