Meta took a break from talking to book publishers in regard to licensing deals to get training data for some of its generative AI models. This has been revealed by the new court documents connected to the case Kadrey v. Meta Platforms. New filings sent to the court on Friday include parts of depositions of Meta employees that were taken by lawyers representing the plaintiffs in the case. These documents show that some Meta employees have concluded that negotiating AI training data licenses for books might not be scalable. Sy Choudhury, who is in charge of Meta’s AI partnership projects, said that Meta’s attempts to get writers interested and involved were met with very slow uptake in engagement and interest. According to the transcript, Sy Choudhury said, “I don’t recall the entire list, but I remember we had made a long list from initially scouring the Internet of top publishers, et cetera and we didn’t get contact and feedback from — from a lot of our cold call outreaches to try to establish contact.” This is one of many cases in the U.S. courts that pitted AI companies against writers, news outlets, visual artists, musicians and other copyright owners. OpenAI, Anthropic, and other technology companies have been accused of using their work to train chatbots without permission or payment. Meta attempts to get copyright permissions Authors like Sarah Silverman and Ta-Nehisi Coates are among the plaintiffs. Since the case was first filed in 2023, the lawyers for the plaintiffs have changed their lawsuit several times. The most recent amended complaint from the plaintiffs’ lawyers says that Meta, among other things, compared pirated books to copyrighted books that were offered for license. This was to see if it made sense to try to get a licensing deal with a publisher. Meta is also accused of training some of its AI models, including the famous Llama series of “open” models, with “shadow libraries” that contain stolen e-books. The lawsuit says that Meta may have used torrenting to protect some of the libraries. People who use torrenting to share files across the internet have to “seed,” or upload, the files they want to get at the same time. The plaintiffs said this was a form of copyright infringement. The court files show that Meta stopped trying to license some AI-related books in early April 2023. They explained that the problem was with timing and other logistics. According to a transcript, Choudhury said that some publishers, especially fiction book publishers, did not actually own the rights to the material that Meta was thinking about licensing. Choudhury said, “I’d like to point out that the — 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 […] And so it would take a long time to engage with all their authors.” In the same light, a transcript shows that Choudhury said that Meta has stopped licensing efforts linked to AI development at least once before. He said, “I am aware of licensing efforts such, for example, we tried to license 3D worlds from different game engines and game manufacturers for our AI research team […] we got very little engagement to even have a conversation.” The US AI industry, which has begun to face competition from China, is facing many lawsuits. In fact, OpenAI and Meta say that being forced to pay copyright holders for their material could kill the US AI industry. Fair use or copyright? – US AI lawsuits in 2025 Both plaintiffs and defendants have something to say. Tech companies have argued that their AI systems make fair use of copyrighted material by studying it to learn to create new, transformative content. Copyright owners argue that the companies unlawfully copy their works to generate rival content that threatens their livelihoods. Some people who own copyright, like big record labels, the New York Times, and a number of best-selling authors have kept pressing their claims. On the other hand, some content producers, like Reddit, News Corp, and the Financial Times, started licensing their work to tech companies on their own. Thomson Reuters and Ross Intelligence dispute shows how judges may handle fair use arguments. Thomson Reuters said Ross stole intellectual property from Westlaw and used it to make an AI-powered legal search engine. Invoking fair use, Ross denied doing anything wrong. 🚨🇺🇸COURT RULES COPYRIGHTED MATERIAL CAN’T BE USED FOR AI TRAINING A U.S District Judge just crushed AI firms, ruling that training on copyrighted content is not fair use. The case, Thomson Reuters vs. Ross Intelligence, could set a devastating precedent for AI companies… pic.twitter.com/w685u7zjKS — Mario Nawfal (@MarioNawfal) February 12, 2025 A federal judge in Delaware said that Ross Intelligence could not copy the technology, information, and company’s material in order to make a competing legal platform based on AI. The decision by U.S. Circuit Judge Stephanos Bibas Ross Intelligence is the first decision made by a U.S. court on the closely watched issue of fair use in AI-related copyright litigation. Clearly, AI companies could have it rough this year with copyright accusations. Cryptopolitan Academy: How to Write a Web3 Resume That Lands Interviews - FREE Cheat Sheet
