Scraping and an official API are equally "dicey" in that the method isn't relevant to the question of how the content itself is licensed to be used.