It's fascinating watching AI companies defend scraping the world's copyrighted content to train their models, while simultaneously accusing competing AI companies of scraping their AI's output to train competing models.