• Meta lawsuit update: Meta has won an AI training copyright lawsuit by 13 book authors. The only remaining issue is whether Meta violated copyright by torrenting books for Llama model training.
• Judge's order: Judge Vince Chhabria granted Meta's motion for summary judgment. Authors and Meta will meet on July 11 to discuss the plaintiffs' claim. Chhabria suggested authors may struggle due to lack of evidence.
• Torrenting relevance: Torrenting from one shadow library (LibGen, over 80.6 terabytes) is "at least potentially relevant" in a few ways. It's relevant to bad faith (first factor of fair use), and could show Meta's use benefited the library and supported unauthorized copying.
• Counting against Meta: Most peer-to-peer file-sharing cases are found to be copyright infringement. Some of the libraries Meta used have been found liable. But authors haven't shown how Meta's downloading benefits pirate libraries.
• Book use connection: There's no separation between Meta downloading books for Llama training and the transformative use. Meta's ultimate use was transformative, so was the downloading.
• Possible outcomes: Authors may show Meta contributed to BitTorrent network. More will be revealed as the case advances. One potential outcome is publishers may make licensing easier. If other authors win, it could set a precedent. Chhabria suggested Meta may renew licensing talks if a stronger fight is raised. If AI companies use only public domain works, there's less excuse for torrenting.