Zuckerberg approved Meta’s use of ‘pirated’ books to train AI models

[Drfun]

A group of authors is accusing Mark Zuckerberg of approving Meta's use of pirated books to help train its AI models. According to their court filing, Meta used a large database called LibGen, which contains books and articles that were shared without permission.



Internal Meta messages show that some of the company's AI team warned that LibGen's content was pirated. Despite this, Zuckerberg reportedly gave the green light to use it. The authors claim that using this material could harm Meta's ability to negotiate with regulators.

The authors behind the lawsuit include Ta-Nehisi Coates and comedian Sarah Silverman. They're suing Meta for using their books to train the Llama language model, which powers the company's chatbots.

LibGen, also known as Library Genesis, is a "shadow library" that started in Russia and contains millions of books. Last year, a New York court ordered LibGen's operators to pay $30 million in damages for copyright violations.

The use of copyrighted material in training AI has become a big issue. Creators are upset that their work is being used without permission, affecting their income and business.

In the court filing, the authors pointed to a memo where Zuckerberg's initials, "MZ," were used to show that he approved the use of LibGen. The filing also mentioned that Meta engineers talked about accessing LibGen data but were hesitant to do so because it didn't feel right to use a corporate laptop for file sharing.

In a previous court decision, a judge dismissed claims that Meta's AI outputs violated the authors' copyrights. However, the judge allowed the authors to revise their case, which they've now done. They plan to add new claims based on the recent evidence.

Meta has not yet commented on the latest developments.