The assertion that Meta’s vocabulary model was inadequately trained on copyrighted material remained unchallenged by the compelling motion to dismiss.
In a legal battle over trademark infringement initiated by scholars, including Sarah Silverman, against Meta concerning its AI language training methods, a federal judge dismissed various claims.
The tech giant faced allegations from Silverman and other plaintiffs for allegedly misusing their copyrighted works to educate its LLaMA language models. Similar accusations were directed at OpenAI regarding the training of ChatGPT, which are still pending alongside the Meta lawsuit.
Judge Vince Chhabria’s ruling came as no surprise, as he sided with Meta in dismissing several claims during a recent hearing, as reported by Reuters. While Meta successfully had these claims dismissed, it notably contested the authors’ central argument that the utilization of their works to train LLaMA violated copyright laws.
Chhabria expressed his understanding of the plaintiffs’ main argument but found the remaining legal theories unclear and unsubstantiated.
The dismissed claims against Meta were broad and sought to establish the extent of Meta’s alleged infringement. Chhabria emphasized that the LLaMA concepts could not function without the data derived from the authors’ works, but he refuted the notion that the LLaMA versions themselves were replicas or adaptations of the plaintiffs’ books, deeming such claims illogical.
Furthermore, Chhabria rejected the plaintiffs’ assertions that every LLaMA output constituted an infringing derivative work, leading to vicarious copyright infringement. He also dismissed claims of Digital Millennium Copyright Act violations and unjust enrichment, stating that Meta did not breach any duty of care towards the authors by utilizing their works for training purposes.
Despite the setbacks, the authors’ case is not entirely closed, with the primary argument still under consideration. They have a 21-day window to submit a revised claim.
Requests for comments from representatives of the defendants and Meta by Rolling Stone remained unanswered at the time of reporting.
