The only thing Japan has done is adopt the TDM exceptions that the EU has. In the US the only certainty we have is that text-to-image AI art generation is not a

The only thing Japan has done is adopt the TDM exceptions that the EU has. In the US the only certainty we have is that text-to-image AI art generation is not a creative process and thus does not accrue copyright on its own. That being said - and, again, sticking with US law - if your AI generated an anime character that looks like an existing character, and that character was either in the training set or described in your prompt, then you'd have an open and shut case. Otherwise it'd be considered a happy accident. That's also the standard for non-AI art as well. It's called "access and substantial similarity." I fully expect Nintendo to privately tell all their licensed developers not to use AI art generators. Actually, boycotting is probably going to be the most effective solution from the perspective of human artists who don't want their job being turned into SEO keyword stuffing[0]. While AI art proponents think that progress is inevitable, the reality of it is that in order for the AI generators to "get better", we need bigger models and more research, both of which are resting on exponential price curves. So if human artists as a class decide they aren't going to use AI, then AI will always be relegated to "second best", even if we can always finetune on those human artists' work. [0] No seriously I think this is what human artists actually fear. It's one thing to point out that they're being put on a treadmill to train the bot that will replace them. But a lot of artists seem to hate the idea of working with a text-to-image art generator, even if that art generator was trained on ethically sourced data. They feel as if their fun has been taken away.