Lawyers are doing their jobs–there is just no legal basis, given how a LLM is built, that the mere act of training a model or generating a latent space infringes copyright. That said, Anthropic literally just got shit on for pirating a shit ton of training data but not the training itself–because actually downloading copies of a work without license is and always has been infringement.
No? Anthropic’s judge literally continued the case over the basis they admitted to torrenting (i.e., pirating) a huge number of books for their training repo. Meta did not do that afaik
Lawyers are doing their jobs–there is just no legal basis, given how a LLM is built, that the mere act of training a model or generating a latent space infringes copyright. That said, Anthropic literally just got shit on for pirating a shit ton of training data but not the training itself–because actually downloading copies of a work without license is and always has been infringement.
Meta got shit on for pirating books.
Anthropic got shit on for not pirating books.
No? Anthropic’s judge literally continued the case over the basis they admitted to torrenting (i.e., pirating) a huge number of books for their training repo. Meta did not do that afaik