Science Big AI firms have built their models by hoovering up copyrighted material from the internet as training data. They say this is legal, but copyright holders disagree – and this year they hit back in a major way
