The federal judge dismissed most of Sarah Silverman’s complaint against Meta regarding the unauthorized use of authors’ copyrighted books for training its generative artificial intelligence concept. This marks the second court decision favoring AI companies in intellectual property disputes. U.S. District Judge Vince Chhabria refuted the claim that Meta’s AI technique is a derivative work of copyrighted material, stating it as “nonsensical.” Silverman’s failure to demonstrate how Meta’s output transformed the plaintiffs’ books led to the dismissal of her copyright infringement claims. The judge allowed her attorneys an opportunity to challenge this decision.
Although Meta did not contest the accusation of using books to train its AI, the court’s decision was influenced by a previous AI art generator lawsuit. Judge William Orrick questioned the ability of artists to prove copyright infringement without similar material generated by AI tools. The fair use defense shields AI companies from obtaining licenses for using copyrighted material in training their models.
The lawsuit focused on Meta’s AI model, LLaMA, and its alleged mimicry of copyrighted expression. However, Judge Chhabria found the argument lacking evidence of direct transformation from copyrighted works. The plaintiffs must demonstrate that LLaMA’s outputs incorporate elements of their books to establish derivative infringement, following a logic similar to Judge Orrick’s previous ruling.
Claims of unfair enrichment and competition law violations were also dismissed in the court’s decision, as they were deemed preempted by trademark infringement claims. The outcome of this case, along with similar author-related lawsuits, will likely impact the use of copyrighted material by AI companies.
