That’s good news, in my opinion. If they’re allowed to just completely disregard copyright when training, then I should be able to completely disregard any attempted copyright on the output too.