Meta Platforms (META.O) is facing serious allegations from a group of authors, including Ta-Nehisi Coates and comedian Sarah Silverman, who claim that the company used pirated versions of copyrighted books to train its artificial intelligence systems, including the Llama language model. The authors argue that this use was approved by Meta’s CEO, Mark Zuckerberg, according to newly disclosed court documents.
The authors, who filed a lawsuit against Meta in 2023 for copyright infringement, allege that internal Meta documents, produced during the discovery phase of the case, show the company was fully aware that the books it used were pirated. Meta has yet to comment on the allegations.
The lawsuit focuses on Meta’s use of the AI training dataset LibGen, a repository of pirated books that the authors claim was distributed through peer-to-peer torrents. The new evidence presented by the authors suggests that Meta executives, including Zuckerberg, were aware that LibGen’s contents were pirated but chose to proceed with using the dataset. Internal Meta communications reportedly confirm this.
The authors are seeking to update their complaint, asserting that the new evidence strengthens their case for copyright infringement. The suit also brings renewed attention to the ongoing legal battles over the use of copyrighted materials to train AI systems, with defendants arguing that such uses may fall under “fair use” doctrine.
In a previous ruling, U.S. District Judge Vince Chhabria dismissed claims related to copyright infringement and the alleged unlawful stripping of copyright management information (CMI) by Meta’s chatbots. However, during a hearing on Thursday, Chhabria indicated that he would permit the authors to file an amended complaint, despite his doubts about the validity of the fraud and CMI claims.