I’d think, in this case, you’d still have to legally acquire the content. So meta pirating a ton of books would still be piracy, the act of piracy would be illegal. Scraping the entire internet for publicly posted data isn’t illegal, however, so that’s still “fair game”. Of course, the internet is full of illegally posted content, so I’m not sure how you account for all that when training AI (most model makers probably don’t bother to try). “Sure, we trained our model on Disney movies, but user FartFace6969 posted them to youtube mirrored with the audio pitched up, your honor!”
Comment on The U.S. Copyright Office’s Draft Report on AI Training Errs on Fair Use
Steve@communick.news 1 week ago
This would be interesting.
That would mean if you are training an AI model at home, you can copy any movie show or music you want, legally.
greybeard@lemmy.one 1 week ago
Butterbee@beehaw.org 1 week ago
I’m certain that’s not what that means. It only means that if Anthropic or Google wants to train an AI model THEY can steal any content they want.
NotSteve_@lemmy.ca 1 week ago
Step 1: be a mega rich corporation