ARTICLE AD BOX
That’s exactly backward. The Copyright Office’s own report makes clear that the legality of training depends on how the data was acquired and what the model does with it. A developer who trains on pirated or paywalled material like Anthropic, Meta and probably all of them to one degree or another, can’t launder infringement through the word 'training.' "
~ The Trichordist from their post 'There Is No ‘Right to Train’: How AI Labs Are Trying to Manufacture a Safe Harbour for Theft'
English (US) ·