I think openai needs to publish an ai Levenshtein distance api. Where you feed in a block of text and it tells you how close it is to something it’s ever generated. It can keep track of everything it’s ever generated and tell us if it seems familiar or not.