A federal judge has allowed AI-related copyright cases to advance the case against Meta, but he has dismissed some of the cases.
In Kadrey vs. Meta, authors such as Richard Kadrey, Sarah Silverman, and Ta-Nehisi Coates alleges that Meta violates intellectual property rights by using the book to train the Lama AI model, and that the company removes copyright information from the book, hiding the alleged infringement charges.
Meanwhile, Mehta argued that the training qualifies as fair use, and that the case should be dismissed because the author lacks a litigation position. In court last month, US District Judge Vince Chhabria appeared to show he was opposed to firing, but he also criticised what the author’s legal team deemed as “excessive” rhetoric.
In Friday’s ruling, Chhabria wrote that the allegations of copyright infringement “a clear, concrete injury sufficient for its position,” and the author also appropriately claims that META intentionally deleted CMI (copyright management information) to conceal copyright infringement.”
“To sum up, these allegations raise “rational, if not particularly strong reasoning,” in which Meta removes the CMI to prevent Llama from outputting the CMI and reveals that it is trained with copyrighted materials,” writes Chhabria.
However, the judge dismissed the author’s claims relating to California’s Comprehensive Computer Data Access and Fraud Act (CDAFA). Because they “doesn’t claim that Meta has accessed a computer or a server.”
The lawsuit has already given us some glimpses into how Meta approaches copyright, and the court application from the plaintiffs alleges that Mark Zuckerberg granted the llama team permission to train the model using copyrighted works, and that other meta team members discussed the use of legally questionable content for AI training.
The court is currently considering many AI copyright cases, including the lawsuit against the New York Times Open.