Some scammers who appear in Musk's replies impersonating him use similar or identical avatars and screen names in order to perpetrate their fraud. Removing and preventing the creation of this type of account would not even require "AI". It hasn't happened yet because Twitter lacks the will to do it, for whatever reason.
Calculating the levenshtein distance from users, the average difference between the avatars, and looking at basic interactions in reply-to-reply-to-users isn't AI. That's heuristic. And it should have been implemented ages ago.
