The latest updates in the AI copyright lawsuit involving Meta lend weight to prior claims that the tech giant had put on hold negotiations with book publishers for licensing agreements to provide its AI models with training data.
The updates pertain to the lawsuit Kadrey v. Meta Platforms, one among numerous similar cases making their way through the U.S. court system. These cases put AI firms and authors and other intellectual property owners at odds. In most instances, the defendants, i.e., AI firms, argue that training on copyrighted material constitutes “fair use”. The plaintiffs, i.e., copyright owners, vehemently dispute this interpretation.
Documents presented to the court last Friday, comprising portions of deposition transcripts from Meta employees, taken by lawyers representing the plaintiffs, insinuate that some Meta personnel believed that negotiating AI training data licensing for books may not be feasible on a large scale.
Sy Choudhury, the person in charge of AI partnership initiatives at Meta, reportedly stated that Meta’s overtures to various publishers met with a lukewarm response.
As per the court transcripts, Meta put a halt to certain AI-linked book licensing endeavors in early April 2023 due to “timing” and other logistical hurdles. Choudhury explained that some publishers, particularly those publishing fiction, didn’t actually possess the rights to the content that Meta was contemplating licensing.
Choudhury further mentioned during his deposition that Meta had, on at least one prior occasion, paused licensing attempts related to AI development. He also remarked on the company’s attempt to license 3D worlds from different game engine and game manufacturers for their AI research team. However, due to limited engagement, Meta decided to create its own solution.
The plaintiffs’ legal team, which includes renowned authors Sarah Silverman and Ta-Nehisi Coates, has revised their complaint multiple times since the lawsuit was initiated in the U.S. District Court for the Northern District of California, San Francisco Division in 2023. The most recent complaint by the plaintiffs alleges that Meta cross-referenced certain pirated books with copyrighted books available for license to evaluate the feasibility of pursuing a licensing agreement with a publisher.
The complaint also accuses Meta of employing “shadow libraries” containing pirated e-books to train several of the company’s AI models, including its popular Llama series of “open” models. The complaint suggests that Meta may have obtained some of these libraries through torrenting, a process that plaintiffs argue is a form of copyright infringement.
