I wonder what will happen when inevitably someone sues one of the AI companies for copyright infringement in a different country, because fair use is to my knowledge an entirely American concept.
Where’s the threshold for forcing AI companies to retrain models without specific copyrighted works in them?