Thinking about abuse potential, is there such a thing as irreversible finger-printing of media generated like this? So that even bad actors couldn't hide the fact that it was generated by AI.