\n<\/p>\n
New court filings<\/a> in an AI copyright case against Meta add credence to earlier reports<\/a> that the company \u201cpaused\u201d discussions with book publishers on licensing deals to supply some of its generative AI models with training data.<\/p>\n The filings are related to the case\u00a0Kadrey v. Meta Platforms<\/a> \u2014 one of many such cases winding through the U.S. court system that\u2019s pitted AI companies against authors and other intellectual property holders. For the most part, the defendants in these cases \u2014 AI companies \u2014 have claimed that training on copyrighted content is \u201cfair use.\u201d The plaintiffs \u2014 copyright holders \u2014 have vociferously disagreed.<\/p>\n The new filings submitted to the court Friday, which include partial transcripts of Meta employee depositions taken by attorneys for plaintiffs in the case, suggest that certain Meta staff felt negotiating AI training data licenses for books might not be scalable. <\/p>\n According to one transcript, Sy Choudhury, who leads Meta\u2019s AI partnership initiatives, said that Meta\u2019s outreach to various publishers was met with \u201cvery slow uptake in engagement and interest.\u201d<\/p>\n \u201cI don\u2019t recall the entire list, but I remember we had made a long list from initially scouring the Internet of top publishers, et cetera,\u201d Choudhury said, per the transcript, \u201cand we didn\u2019t get contact and feedback from \u2014 from a lot of our cold call outreaches to try to establish contact.\u201d<\/p>\n Choudhury added, \u201cThere were a few, like, that did, you know, engage, but not many.\u201d<\/p>\n According to the court transcripts, Meta paused certain AI-related book licensing efforts in early April 2023 after encountering \u201ctiming\u201d and other logistical setbacks. Choudhury said some publishers, in particular fiction book publishers, turned out to not in fact have the rights to the content that Meta was considering licensing, per a transcript.<\/p>\n \u201cI\u2019d like to point out that the \u2014 in the fiction category, we quickly learned from the business development team that most of the publishers we were talking to, they themselves were representing that they did not have, actually, the rights to license the data to us,\u201d Choudhury said. \u201cAnd so it would take a long time to engage with all their authors.\u201d<\/p>\n Choudhury noted during his deposition that Meta has on at least one other occasion paused licensing efforts related to AI development, according to a transcript.<\/p>\n \u201cI am aware of licensing efforts such, for example, we tried to license 3D worlds from different game engine and game manufacturers for our AI research team,\u201d Choudhury said. \u201cAnd in the same way that I\u2019m describing here for fiction and textbook data, we got very little engagement to even have a conversation [\u2026] We decided to \u2014 in that case, we decided to build our own solution.\u201d<\/p>\n Counsel for the plaintiffs, who include bestselling authors Sarah Silverman and Ta-Nehisi Coates, have amended their complaint several times since the case was filed in the U.S. District Court for the Northern District of California, San Francisco Division in 2023. The latest amended complaint submitted by plaintiffs\u2019 counsel alleges that Meta, among other offenses, cross-referenced certain pirated books with copyrighted books available for license to determine whether it made sense to pursue a licensing agreement with a publisher.\u00a0<\/p>\n The complaint also accuses Meta of using \u201cshadow libraries\u201d containing pirated e-books<\/a> to train several of the company\u2019s AI models, including its popular Llama series of \u201copen\u201d models. According to the complaint, Meta may have secured some of the libraries via torrenting. Torrenting, a way of distributing files across the web, requires that torrenters simultaneously \u201cseed,\u201d or upload, the files they\u2019re trying to obtain \u2014 which the plaintiffs asserted is a form of copyright infringement<\/a>.<\/p>\n<\/div>\n