New court documents in a copyright case involving AI against Meta reinforce previous reports that the company decided to temporarily halt discussions with book publishers for licensing agreements to provide training data for its AI models.

This case, known as Kadrey v. Meta Platforms, is one of many similar cases in the U.S. legal system where AI companies are in conflict with authors and intellectual property owners. The defendants, AI companies, argue that training on copyrighted material falls under “fair use,” while the plaintiffs, copyright holders, strongly disagree.

Recent court filings submitted on Friday included excerpts from depositions of Meta employees by attorneys representing the plaintiffs. These documents suggest that some Meta staff members believed negotiating AI training data licenses for books might not be scalable. According to the transcripts, Sy Choudhury, in charge of Meta’s AI partnership initiatives, mentioned that there was minimal engagement and interest from publishers when Meta reached out to them.

Choudhury explained that their attempts to establish contact with publishers did not yield positive results, leading Meta to pause certain efforts related to book licensing in early April 2023 due to timing and logistical challenges.

Choudhury revealed that many fiction book publishers did not have the rights to license content, making it challenging to engage with authors within a reasonable timeframe. He also mentioned another instance where Meta paused licensing efforts related to AI development for 3D worlds due to lack of engagement with game engine and game manufacturers.

The plaintiffs’ counsel, which includes well-known authors Sarah Silverman and Ta-Nehisi Coates, have made several amendments to their complaint since filing the case in 2023. The latest complaint alleges that Meta cross-referenced pirated books with copyrighted books to assess potential licensing agreements and used “shadow libraries” containing pirated e-books to train their AI models, including the Llama series. Torrenting was mentioned as a method to access these libraries, which the plaintiffs argue constitutes copyright infringement.