A federal judge has dismissed most of Sarah Silverman‘s lawsuit against Meta over the unauthorized use of authors’ copyrighted books to train its generative artificial intelligence model, marking the second ruling from a court siding with AI firms on novel intellectual property questions presented in the legal battle.
U.S. District Judge Vince Chhabria on Monday offered a full-throated denial of one of the authors’ core theories that Meta’s AI system is itself an infringing derivative work made possible only by information extracted from copyrighted material. “This is nonsensical,” he wrote in the order. “There is no way to understand the LLaMA models themselves as a recasting or adaptation of any of the plaintiffs’ books.”
Another of Silverman’s arguments that every result produced by Meta’s AI tools constitutes copyright infringement was dismissed because she didn’t offer evidence that any of the outputs “could be understood as recasting, transforming, or adapting the plaintiffs’ books.” Chhabria gave her lawyers a chance to replead the claim, along with five others that weren’t allowed to advance.
Notably, Meta didn’t move to dismiss the allegation that the copying of books for purposes of training its AI model rises to the level of copyright infringement.
The ruling builds upon findings from another federal judge overseeing a lawsuit from artists suing AI art generators over the use of billions of images downloaded from the Internet as training data. In that case, U.S. District Judge William Orrick similarly delivered a blow to fundamental contentions in the lawsuit by questioning whether artists can substantiate copyright infringement in the absence of identical material created by the AI tools. He called the allegations “defective in numerous respects.”
Some of the issues presented in the litigation could decide whether creators are compensated for the use of their material to train human-mimicking chatbots that have the potential to undercut their labor. AI companies maintain that they don’t have to secure licenses because they’re protected by the fair use defense to copyright infringement.
According to the complaint filed in July, Meta’s AI model “copies each piece of text in the training dataset” and then “progressively adjusts its output to more closely resemble” expression extracted from the training dataset. 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 “not viable” in the absence of allegations or evidence suggesting that LLaMA, short for Large Language Model Meta AI, has been “recast, transformed, or adapted” based on a preexisting, copyrighted work.
Another of Silverman’s main theories — along with other creators suing AI firms – was that every output produced by AI models are infringing derivatives, with the companies benefiting from every answer initiated by third-party users allegedly constituting an act of vicarious infringement. The judge concluded that her lawyers, who also represent the artists suing StabilityAI, DeviantArt and Midjourney, are “wrong to say that” — because their books were duplicated in full as part of the LLaMA training process — evidence of substantially similar outputs isn’t necessary.
“To prevail on a theory that LLaMA’s outputs constitute derivative infringement, the plaintiffs would indeed need to allege and ultimately prove that the outputs ‘incorporate in some form a portion of’ the plaintiffs’ books,” Chhabria wrote. His reasoning mirrored that of Orrick, who found in the suit 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 plaintiffs across most cases will have to present evidence of infringing works produced by AI tools that are identical to their copyrighted material. This potentially presents a major issue because they have conceded in some instances that none of the outputs are likely to be a close match to material used in the training data. Under copyright law, a test of substantial similarity is used to assess the degree of similarity to determine whether infringement has occurred.
Other dismissed claims in Chhabria’s order include those over unjust enrichment and violation of competition laws. To the extent they’re based on the surviving claim for copyright infringement, he found that they’re preempted.
Meta didn’t immediately respond to a request for comment.
In July, Silverman also joined a class action against OpenAI accusing the company of copyright infringement. The case has been consolidated with other suits from authors in federal court.