In the latest Re:Create Recap note, I explain how the work of computer science researchers tricking AI models into spitting out copies of training data doesn't have the dire legal consequences some commentators advertise. We know because the 2nd Circuit said so in Google Books.
