I honestly see much less of an issue with using scraped content if there is no profit motivation behind it. The main thing is, that all that scraping and training costs a fuckton of money, so no open source nonprofit model will be able to do that