Meta CEO Mark Zuckerberg approved the use of pirated books to train AI: Court filing

Meta Platforms trained its AI models using pirated versions of copyrighted books, with the approval of its CEO Mark Zuckerberg.

According to newly disclosed court papers, a group of authors is alleging that the social media giant knew they were using pirated work to train their AI systems.

Internal documents from Meta “reveal” the claims

In their court filing, the authors said internal documents that were produced by Meta during the discovery process showed that the social network firm knew about the pirated books.  According to The Guardian, its CEO Zuckerberg backed the use of the LibGen dataset, a huge online archive of books. This was despite warnings within the company’s AI executive team that it is a dataset “we know to be pirated.”

US author Ta-Nehisi Coates, comedian Sarah Silverman, and other writers suing the company for copyright infringement made the accusations in filings that were made public on Wednesday in a California federal court.

The authors took Meta to court in 2023 on allegations that the social media company was misusing their books to train AI models, specifically Llama, its large language model that powers its chatbots.

Originating in Russia, the Library Genesis or LibGen dataset is a “shadow library” which claims to contain millions of novels, notification books as well as science magazine articles.

In 2024, a New York federal court asked LibGen’s anonymous operators to pay a group of publishers $30 million in damages for copyright infringement

This is one of the many others that allege that copyrighted work by authors, artists, and others was used to train generative AI tools like ChatGPT chatbot without consent from owners. Creative professionals have warned that using their work without their consent is endangering their business models.

According to Reuters, defendants have, however, argued that they made fair use of copyrighted material.

The judge allowed the authors to file an amended complaint

In the Meta case, the authors reportedly asked the court on Wednesday for permission to file an updated complaint. In their arguments, they indicated that new evidence showed that social network firms used the AI training dataset LibGen, which includes millions of pirated works, and distributed it through peer-to-peer torrents.

According to them, Zuckerberg “approved Meta’s use of the LibGen dataset notwithstanding concerns within Meta’s AI executive team (and others at Meta) that LibGen is ‘a dataset we know to be pirated.'”

The filing also cites a memo that referred to Zuckerberg’s initials, noting that “after escalation to MZ” Meta’s AI team “has been approved to use LibGen.”

Last year, a US district judge, Vince Chhabria, dismissed claims that text generated by Meta’s AI models infringed the authors’ copyrights and that Meta unlawfully stripped books’ copyright management information. This refers to information about the work including the title, name of author, and copyright owner.

The plaintiffs were, however, permitted to amend their claims. In their arguments this week, the authors said the evidence bolstered their infringement claims and justified reviving their copyright management information case adding a new computer fraud allegation.

During a Thursday hearing, the judge said he would allow the authors to file an amended complaint although he was skeptical about the merits of the fraud claims.

Land a High-Paying Web3 Job in 90 Days: The Ultimate Roadmap