Report from Reuters
In Brief – In an order granting Meta summary judgement in an AI-training copyright infringement case brought by 13 authors, Federal Judge Vince Chhabria delivered an impassioned dissent to fellow judge William Alsup’s order released only a few days before that offered a strident defense of AI training as often being protected by the fair use copyright defense in a case involving other authors suing AI developer Anthropic. Judge Chhabria made clear that while the facts of this case and plaintiffs’ arguments, which he called “clear losers”, forced him to side with Meta, he did not want to. And he presented his thoughts on a “potentially winning argument—that Meta has copied their works to create a product that will likely flood the market with similar works, causing market dilution.”
Context – A fair use shootout between dueling federal judges. Wow. And neither order exceeds 40 pages. They are worth reading. Alsup focuses on the clearly transformative nature of Generative AI (GAI), a revolutionary tool to create digital content, and considers that to be the key fair use factor defending training. Chhabria’s rebuttal argues that the most important factor is that GAI will “harm the market” for basically any kind of digital content it can create, saying “the potentially winning argument” is not that GAI will regurgitate the training material, but instead that it will generate massive volumes of cheap content that is “similar enough to compete with the originals and thereby indirectly substitute for them.” He calls this “indirect market substitution”. On that argument, he said, “the plaintiffs’ presentation is so weak that it does not move the needle,” and concludes that his ruling does not mean he thinks Meta’s use of copyrighted materials to train its LLMs is lawful, but “stands only for the proposition that these plaintiffs made the wrong arguments and failed to develop a record in support of the right one.” And we conclude that despite both judges siding with AI developers their decisions don’t remotely resolve the fair use issue and instead point to the fact that judge-created rules are going to take more years, many more judges, and likely the US Supreme Court.