Yep, totally fair question, and one that’s being tested legally on many fronts. Rulings are generally siding with AI companies on the training side (using copyrighted works to train models is fair use) but there aren’t many decisions yet about output. The next few years will be interesting.