In a recent legal battle surrounding the unauthorized use of authors' copyrighted books for training an AI model, U.S. District Judge Vince Chhabria has dismissed most of Sarah Silverman's lawsuit against Meta. This ruling marks another victory for AI firms, as the court has sided with them on novel intellectual property questions.

Judge Chhabria firmly denied one of the core arguments put forth by the authors, which claimed that Meta's AI system itself is an infringing derivative work made possible only by using copyrighted material. In his order, he stated that there is no way to consider Meta's LLaMA models as a recasting or adaptation of the plaintiffs' books, calling such claims nonsensical.

Additionally, Silverman's argument that every result produced by Meta's AI tools constitutes copyright infringement was also dismissed by the judge. She failed to provide evidence that any of the outputs could be understood as a recasting, transformation, or adaptation of the plaintiffs' books. However, Chhabria did give her lawyers the opportunity to replead this claim, along with five others that were not allowed to advance in court.

Interestingly, Meta did not seek to dismiss the allegation that the copying of books for training its AI model could potentially rise to the level of copyright infringement.

This ruling builds upon findings from another federal judge, William Orrick, who presided over a similar lawsuit where artists were suing AI art generators over the use of billions of images downloaded from the internet as training data. Judge Orrick also cast doubt on the ability of artists to prove copyright infringement in the absence of identical material created by the AI tools, calling the allegations "defective" in several respects.

The outcome of these legal battles could have implications for whether creators are compensated for the use of their material to train chatbots that mimic human behavior, potentially undercutting their labor. AI companies argue that they are protected by the fair use defense to copyright infringement and therefore do not need to secure licenses.

Silverman's lawsuit, filed in July, contended that Meta's AI model extensively copied text from the training dataset and progressively adjusted its output to resemble expressions extracted from the copyrighted material. The central claim revolved around the assertion that the entire purpose of Meta's LLaMA model is to imitate copyrighted expression, rendering the entire model an infringing derivative work.

However, Judge Chhabria dismissed this argument, stating that there were no allegations or evidence suggesting that LLaMA had been recast, transformed, or adapted based on preexisting copyrighted works.

Another key theory presented by Silverman and other creators suing AI firms was that every output produced by AI models constitutes infringing derivatives and that companies benefit from third-party users initiating acts of vicarious infringement. Judge Chhabria concluded that evidence of substantially similar outputs was required to prove this theory, as the plaintiffs would need to demonstrate that LLaMA's outputs incorporated portions of their books. This aligned with Judge Orrick's view that the alleged infringer's derivative work must bear some similarity to the original work or contain protected elements.

As a result, plaintiffs in these cases face the challenge of presenting evidence of AI tools producing infringing works that are identical to their copyrighted material. This presents a significant hurdle, as some have already conceded that none of the outputs closely match the training data content. Under copyright law, substantial similarity is a crucial factor in assessing infringement.

Other claims in Judge Chhabria's order, including those regarding unjust enrichment and violation of competition laws, were also dismissed as they were found to be preempted by the surviving claim for copyright infringement.

In July, Sarah Silverman joined a class action lawsuit against OpenAI, accusing the company of copyright infringement. This case has now been consolidated with other lawsuits from authors in federal court.