That's one thing I've been asking myself a lot recently: Is it realistic to expect that AI can properly attribute the source material in its output anytime soon

That's one thing I've been asking myself a lot recently: Is it realistic to expect that AI can properly attribute the source material in its output anytime soon? IMHO this is a non-negotiable requirement if the field wants to be taken serious. How can I trust the output if I cannot verify the correctness of the input? And then there's the legal aspect: in music, sampling is legal nowadays, but you have to ask the original author for permission and then pay royalties based on how much your own creation is based on the original sample. AI isn't much more than automated sampling, how can AI generated content ever hope to avoid a legal quagmire if it cannot properly attribute its source material?