Sorry, but that's entirely false. See my other comment. The US fair use test was created in 1841. The Google case only found that its book search product passed the test, including the publication of "snippets" not having a negative impact on the market for books. That doesn't mean every other arguably-similar project passes the test too. They would need to show that, for example, generated images do not impact the market for images made by the artists whose work was scanned - which is obviously not the case. The situation with generative neural networks is not at all "well settled" by the case about Google's book search.