A federal judge has dismissed most of Sarah Silverman’s lawsuit against Meta over the unauthorized use of authors’ copyrighted books to train her generative artificial intelligence model.
U.S. District Judge Vince Chhabria on Monday strongly disputed one of the authors’ core theories, that Meta’s AI system was itself an infringing derivative work made possible only by information extracted from copyrighted material. “This is nonsense,” he wrote in the order. “There is no way to understand the LLaMA models themselves as a rewrite or adaptation of any of the plaintiffs’ books.”
Another argument by Silverman that any result produced by Meta’s AI tools constituted copyright infringement was rejected because she presented no evidence that any of the results “could be construed as a rewrite, redesign, or adaptation of Plaintiffs’ books.” . Chhabria gave her lawyers a chance to refile the lawsuit, along with five others who were not allowed to proceed.
In particular, Meta has made no move to reject the accusation that copying books for the purpose of training its AI model constitutes copyright infringement.
The ruling builds on findings from another federal judge who is overseeing a lawsuit by artists suing AI art generators for using billions of images downloaded from the Internet as training data. In that case, U.S. District Judge William Orrick also struck a blow at basic allegations in the lawsuit by questioning whether artists can establish copyright infringement in the absence of identical material created using the AI tools. He described the allegations as “flawed in many ways.”
Some of the questions raised in the litigation could determine whether creators are compensated for using their material to train human-like chatbots that have the potential to undermine their work. AI companies claim that they do not need to secure licenses because they are protected by the fair use defense against copyright infringement.
According to the complaint filed in July, Meta’s AI model “copies each section of text in the training data set” and then “incrementally adjusts its output to more closely resemble the expression extracted from the training data set.” The lawsuit revolved around the claim that the entire purpose of LLaMA is to imitate copyrighted expression and that the entire model should be considered an infringing derivative work.
But Chhabria called the argument “unsound” because there are no claims or evidence to suggest that LLaMA, short for Large Language Model Meta AI, was “recast, transformed, or modified” based on a pre-existing copyrighted work “adapted”.
Another of Silverman’s main theories – along with other developers suing AI companies – was that any results produced by AI models infringed derivatives and the companies benefited from any third-party response, allegedly constituting an act of vicarious infringement. The judge concluded that her lawyers, who also represent the artists suing StabilityAI, DeviantArt and Midjourney, were “wrong to say that” – because their books were fully reproduced as part of the LLaMA training process – evidence of substantive similar results are not necessary.
“To prevail on the theory that LLaMA’s results constitute derivative infringement, plaintiffs would actually have to allege and ultimately prove that the results “incorporate in some form a portion” of plaintiffs’ books,” Chhabria wrote. His reasoning mirrored that of Orrick, who noted in the lawsuit against StabilityAI that “the alleged infringer’s derivative work must still bear some similarity to the original work or contain the protected elements of the original work.”
This means that in most cases, plaintiffs will have to provide evidence of infringing works created by AI tools that are identical to their copyrighted material. This is potentially a big problem because in some cases they have admitted that none of the outputs are likely to come close to matching the material used in the training data. In copyright law, a substantial similarity test is used to assess the degree of similarity and determine whether infringement has occurred.
Other claims dismissed in Chhabria’s order include those of unjust enrichment and violation of competition laws. To the extent that they are based on the existing claim for copyright infringement, they are, in his opinion, anticipated.
Meta did not immediately respond to a request for comment.
In July, Silverman also joined a class action lawsuit against OpenAI, accusing the company of copyright infringement. The case was consolidated with other lawsuits by authors in federal court.