Well I expect if I were to start scraping it, they'd stop me pretty quick. But that doesn't necessarily matter because any AI which has been trained on it can be made to regurgitate the similarly structured text, so basically the dataset gets laundered from one model to another until it ends up in an open source model anyway..