Key holding or allegation
Meta obtained summary judgment on the named plaintiffs' training claim because the evidentiary record did not establish sufficient market harm, despite the court recognising that generative-AI market dilution can be legally important.
Why it matters
Major U.S. merits ruling warning against categorical treatment of AI training while showing the centrality of market-harm evidence.
Facts
Authors alleged Meta used copyrighted books, including shadow-library materials, to train Llama and engaged in related distribution/torrent conduct.
Issues before the court
Whether the challenged training was fair use and whether market harm/dilution was sufficiently evidenced; separate acquisition/distribution theories continued.
What the court did not decide
Does not hold that all AI training is fair use; later torrent/distribution/contributory theories remained live.
Procedural timeline
U.S. District Court for the Northern District of California
Summary judgment: Meta prevailed on the named plaintiffs' training claim because the evidentiary record did not establish sufficient market harm, while the court recognised that generative-AI market dilution can matter to fair use.