A federal judge has dismissed most of Sarah Silverman’s lawsuit against Meta for unauthorized use of copyrighted books to train its generative artificial intelligence model, marking the second court ruling siding with AI companies on new intellectual property issues presented in the legal battle. .
U.S. District Judge Vince Chhabria on Monday denied 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. author. “This is absurd,” he wrote in the order. “There is no way to understand the LLaMA models themselves as a reworking or adaptation of any of the plaintiffs’ books.”
Another argument by Silverman that every result produced by Meta’s AI tools constitutes copyright infringement was rejected because she failed to provide evidence that any of the results “could be understood as a recasting, transformation or adaptation of the plaintiffs’ books.” Chhabria gave his lawyers the opportunity to reiterate their complaint, along with five other lawyers who were not allowed to move forward.
Notably, Meta has not rejected the claim that copying books for the purpose of training its AI model amounts to copyright infringement.
The ruling builds on the findings of another federal judge overseeing a lawsuit brought 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 dealt a blow to the lawsuit’s core arguments by questioning whether the artists could justify copyright infringement in the absence of identical material created by the tools of AI. He called the allegations “flawed in many respects.”
Some of the issues raised in the litigation could decide whether creators are paid for using their hardware to train human-mimicking chatbots that could reduce their labor. AI companies argue that they do not need to obtain 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 every piece of text from the training dataset” and then “gradually adjusts its output to more closely resemble” the expression extracted from the training dataset. training data. The lawsuit revolved around the claim that LLaMA’s sole purpose is to imitate copyrighted expression and that the entire design should be considered an infringing derivative work.
But Chhabria called the argument “unsustainable” in the absence of allegations or evidence suggesting that LLaMA, short for Large Language Model Meta AI, was “recast, transformed or adapted” based on a work pre-existing copyrighted material.
Another of Silverman’s main theories – as well as other creators suing AI companies – was that every result produced by AI models infringed derivatives, with the companies profiting from every response initiated by third-party users, which which allegedly constitutes an act of indirect infringement. The judge concluded that his lawyers, who also represent the artists suing StabilityAI, DeviantArt and Midjourney, are “wrong to say that” – because their books were entirely duplicated as part of the LLaMA training process – evidence of results substantially similar is not necessary.
“To prevail on the theory that LLaMA’s results constitute derivative infringement, plaintiffs would indeed 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 concluded in the lawsuit against StabilityAI that “the infringer’s alleged derivative work must always bear some similarity to the original work or contain the protected elements of the original work.”
This means that in most cases, plaintiffs will need to present evidence of infringing works produced by AI tools that are identical to their copyrighted material. This potentially presents a major problem as they have admitted in some cases that none of the results are likely to closely match the hardware used in the training data. Under copyright law, a substantial similarity test is used to assess the degree of similarity to determine whether infringement has occurred.
Other claims dismissed in Chhabria’s order include those related to unjust enrichment and violation of competition laws. To the extent that they are based on the surviving copyright infringement claim, he found that they were 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 filed by perpetrators in federal court.