Raw LLM Responses

I'm currently following an insane court case where AI is sort of relevant: 5 years in to the case, the plaintiff has suddenly started (obviously) using chatGPT to write his motions. They're all awful, include blatant lies, and makes citations that don't say what he says they say (often, they show the opposite of what he wants them to). I've been feeding every motion into chatGPT: the first AI motion I fed it, it thought was extremely well done, despite anyone even remotely familiar with the case or the law in general being able to easily pick apart into nothing. As soon as it got the defendant's response, it started trashing the plaintiff and keeps suggesting that we make bingo games to predict how bad the next filing from the plaintiff will be. It's insanely accurate at predicting both parties motions at this point, including successfully predicting that the plaintiff's next motion would be to try to get the court to remove the defense attorney for defending the defendant. The only thing it doesn't like about the defense attorney so far is when he told the plaintiff that his AI generated motion was so full of errors that he wasn't capable of giving an intelligent argument against it, but only because of the AI accusation.