Recent legal documents in an AI copyright lawsuit against Meta reinforce previously reported claims that the organization has “paused” negotiations with book publishers regarding licensing agreements for training data used in several of its generative AI models.
These documents pertain to the case Kadrey v. Meta Platforms — one of numerous lawsuits making their way through the U.S. court system, involving AI firms and authors, along with other holders of intellectual property. Generally, the defendants in these lawsuits—primarily AI companies—have contended that training on copyrighted material constitutes “fair use.” In contrast, the plaintiffs—those holding copyright—strongly disagree.
The latest submissions made to the court include excerpts from depositions of Meta employees, taken by the plaintiffs’ attorneys, and suggest that some employees believed negotiating licenses for AI training data from books might not be feasible in the long run.
One of the depositions revealed that Sy Choudhury, who spearheads Meta’s AI partnership initiatives, indicated that the company had experienced “very slow uptake in engagement and interest” from publishers during outreach efforts.
“I can’t recall the complete list, but I do remember we compiled a lengthy list by initially researching top publishers online,” Choudhury stated according to the transcript. “Unfortunately, we didn’t receive much response or feedback from many of our outreach attempts to try and make contact.”
Choudhury remarked, “A few did engage with us, but the number was quite limited.”
Court records reveal that Meta halted certain book licensing endeavors related to its AI initiatives in early April 2023 due to “timing” and various logistical challenges. Choudhury noted that some publishers, particularly those specializing in fiction, actually did not possess the rights to the material under consideration for licensing.
“In the fiction category, we quickly learned from our business development team that many of the publishers we were in talks with claimed they did not have the rights to license the data to us,” Choudhury elaborated. “This meant it would take a considerable amount of time to engage with all their authors.”
During his testimony, Choudhury acknowledged that Meta had previously paused efforts related to AI licensing, according to the transcript.
“For instance, we attempted to license 3D worlds from various game engine and manufacturer companies for our AI research team,” Choudhury noted. “Similar to what I’ve described for fiction and textbook data, we received minimal interest to even have a dialogue […] Consequently, we opted to build our own solution.”
The plaintiffs’ legal team, representing bestselling authors like Sarah Silverman and Ta-Nehisi Coates, has submitted several amendments to their complaint 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 amended complaint alleges that Meta, among other charges, cross-referenced certain pirated works with copyrighted titles available for licensing to identify whether pursuing agreements with publishers was justifiable.
Moreover, the complaint accuses Meta of utilizing “shadow libraries” of pirated e-books to train various AI models, including the widely used Llama series labeled as “open.” The complaint suggests that Meta may have accessed some of these libraries through torrenting. Torrenting, which enables the distribution of files online, requires users to “seed,” or upload, the files they wish to obtain—an act that the plaintiffs argue constitutes copyright infringement.
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
