GPT 3,5 and 4: "we are trained over vast data collected over the internet, written by users over a long time. And books and articles on arxiv and all other things that needed quite some effort"

Also GPT: "you cannot copy from us!!"

Not back when Bing was scraping Google's search results in order to try and make their search engine half decent, of course. They only stopped when they got caught, only fair AI rivals do the same.