Meta can train its train its Llama large language models with the copyrighted material of 13 bestselling authors — including Sarah Silverman, Ta-Nehisi Coates and Junot Diaz — without their permission, a judge found Wednesday. (Photo by Jaque Silva/NurPhoto via AP)
Like Judge Alsup, Judge Chhabria concluded that the AI technology at issue in the cases is transformative, but did not make that the basis for his ruling in favor of Meta on the question of fair use. Rather, Judge Chhabria excoriated the plaintiffs — who include Sarah Silverman, Ta-Nehisi Coates and Junot Diaz — as failing to present meaningful evidence that Meta's use of their works to create Llama impacted the market for their books.
"Meta has defeated the plaintiffs' half-hearted argument that its copying causes or threatens significant market harm," Wednesday's decision states. "That conclusion may be in significant tension with reality, but it's dictated by the choice the plaintiffs made to put forward two flawed theories of market harm while failing to present meaningful evidence on the effect of training LLMs like Llama with their books on the market for those books."
The judge summarized the theories as contending, first, "that Llama is capable of reproducing small snippets'" from their books and, second, "that Meta, by using their works for training without permission, has diminished the authors' ability to license their works for the purpose of training large language models."
The judge added: "Both of these arguments are clear losers."
Still pending is the authors' claim that Meta infringed their copyrights by downloading millions of pirated books through peer-to-peer networks. Neither party moved for summary judgment on that matter. Judge Chhabria also said he would issue a separate order granting Meta's motion for summary judgment on the plaintiffs' claim that it violated the Digital Millennium Copyright Act by removing authorship information from their works.
Judge Chhabria indicated during summary judgment arguments last month that he was inclined to find that Meta's LLM is transformative, a finding that typically supports fair use. But he said his ruling would hinge on the effect that Meta's action would have on the market for the authors' books.
Judge Chhabria emphasized that the consequences of his ruling are limited, noting that the case is not a class action and that his decision impacts only the 13 plaintiffs, "not the countless others whose works Meta used to train its models."
"And, as should now be clear," Wednesday's decision states, "this ruling does not stand for the proposition that Meta's use of copyrighted materials to train its language models is lawful. It stands only for the proposition that these plaintiffs made the wrong arguments and failed to develop a record in support of the right one."
Boies Schiller Flexner LLP, one of the firms representing the plaintiffs, said in a statement to Law360 Wednesday that it disagreed with the court's opinion.
"The court ruled that AI companies that 'feed copyright-protected works into their models without getting permission from the copyright holders or paying for them' are generally violating the law. Yet, despite the undisputed record of Meta's historically unprecedented pirating of copyrighted works, the court ruled in Meta's favor. We respectfully disagree with that conclusion."
Meta said it appreciated the court's decision.
"Open-source AI models are powering transformative innovations, productivity and creativity for individuals and companies, and fair use of copyright material is a vital legal framework for building this transformative technology," a Meta spokesperson said in a statement to Law360 Wednesday.
The lawsuits against Meta and Anthropic are among dozens nationwide that accuse tech companies of ripping off copyrighted material to develop generative AI models. The cases against Meta and Anthropic include allegations that LLM training materials included millions of copyrighted works taken from pirating sites, and the presiding judges differed in their takes on the issues.
Judge Alsup said Anthropic must face a trial regarding the pirated books it downloaded, even if the AI technology is transformative. Judge Chhabria said that "unmitigated piracy" did not tip the scales against Meta on fair use, finding that content downloads were ultimately used for a transformative purpose and that the plaintiffs failed to provide evidence that Meta's use of so-called "shadow libraries" caused them significant market harm.
Judge Chhabria expressed more concern than U.S. District Judge William Alsup about the potential for AI-generated works to flood the market with books, songs, images and other material that could "dramatically undermine the market for those works" and the incentive for humans to create them.
In his opinion, Judge Chhabria drew a contrast between his view and Judge Alsup's perspective on the potential market harm that AI-created works could have on the original materials the technology is trained on, saying he disagreed with one of Judge Alsup's analogies.
"Such harm would be no different, [Judge Alsup] reasoned, than the harm caused by using the works for 'training schoolchildren to write well,' which could 'result in an explosion of competing works,'" Judge Chhabria said. "But when it comes to market effects, using books to teach children to write is not remotely like using books to create a product that a single individual could employ to generate countless competing works with a miniscule fraction of the time and creativity it would otherwise take."
Judge Chhabria called Judge Alsup's analogy "inapt" and said it was "not a basis for blowing off the most important factor in the fair use analysis," referring to market harm.
Judge Chhabria also called it "ridiculous" to say adverse rulings against tech companies facing copyright lawsuits would stop the development of generative AI.
"These products are expected to generate billions, even trillions, of dollars for the companies that are developing them," Judge Chhabria said. "If using copyrighted works to train the models is as necessary as the companies say, they will figure out a way to compensate copyright holders for it."
The major rulings in the dozens of pending AI copyright cases are expected to go through years of appeals until the U.S. Supreme Court weighs in.
The authors are represented by Joseph R. Saveri, Cadio Zirpoli, Christopher K.L. Young, Holden Benon and Aaron Cera of Joseph Saveri Law Firm LLP, David Boies, Joshua Irwin Schiller, Maxwell V. Pritt, Jay Schuffenhauer, Joshua M. Stein, Margaux Poueymirou, David L. Simons and Jesse Panuccio of Boies Schiller Flexner LLP, Elizabeth J. Cabraser, Daniel M. Hutchinson, Reilly T. Stoler, Rachel Geman, Kenneth S. Byrd and Betsy A. Sugar of Lieff Cabraser Heimann & Bernstein LLP, Bryan L. Clobes, Mohammed A. Rathur and Alexander J. Sweatman of Cafferty Clobes Meriwether & Sprengel LLP, Nada Djordjevic, Amy E. Keller, David A. Straite and James A. Ulwick of DiCello Levitt, Scott J. Sholder and CeCe M. Cole of Cowan DeBaets Abrahams & Sheppard LLP, and Matthew Butterick.
Meta is represented by Bobby Ghajar, Mark Weinstein, Kathleen Hartnett, Judd Lauter, Elizabeth L. Stameshkin and Colette Ghazarian of Cooley LLP, Angela L. Dunning of Cleary Gottlieb Steen & Hamilton LLP and William Thomas Marks, Anna Manchester Stapleton and Kannon K. Shanmugam of Paul Weiss Rifkind Wharton & Garrison LLP.
The case is Richard Kadrey et al. v. Meta Platforms Inc. et al., case number 3:23-cv-03417, in the U.S. District Court for the Northern District of California.
--Editing by Emily Kokoll and Amy French.
For a reprint of this article, please contact reprints@law360.com.
