zlacker

>>moose4+(OP)
> “We will have a trial on the pirated copies used to create Anthropic’s central library and the resulting damages,” Judge Alsup wrote in the decision. “That Anthropic later bought a copy of a book it earlier stole off the internet will not absolve it of liability for theft but it may affect the extent of statutory damages.”

I'm not sure why this alone is considered a separate issue from training the AI with books. Buying a copy of a copyrighted work doesn't inherently convey 'fair use rights' to the purchaser. If I buy a work, read it, sell it, and then publish a review or parody of it, I don't infringe copyright. Why does mere possession of an unauthorized copy create a separate triable matter before the court?

Keep in mind, you can legally engineer EULAs in such a way that merely purchasing the work surrenders all of your fair use rights. So this could wind up being effectively: "AI training is fair use for works purchased before June 24th, 2025, everything after is forbidden, here's your brand new moat OpenAI"

>>kmeist+AB
The ruling suggests that "pirating a book that could have been bought at a bookstore" for the sake of "writing a book review" "is inherently, irredeemably infringing".

Which suggests that, at least in the judge's opinion, 'fair use rights' do exist in a sense, but it's about when you read the book, not when you publish.

But that's not settled precedent. Meta is currently arguing the opposite in Kadrey v. Meta: they're claiming that they can get away with torrenting training material as long as they only leech (download) and don't seed (upload), because, although the act of downloading (copying) is generally infringement under a Ninth Circuit precedent, they were making a fair use.

As for EULAs, that might be true for e-books, but publishers can't really do anything about Anthropic's new strategy of scanning physical books, because physical books generally don't come with shrinkwrap license agreements. Perhaps publishers could start adding them, but I think that would sit poorly with the public and the courts.

(That's assuming the ruling isn't overturned on appeal, which it easily might be.)