Why? Because there's no word for "insurance of AI advise accuracy"? The whole point of progress is that we create something that is not a thing at the moment.
No, because, like I said, GPTs are not legally allowed to represent individuals, so they cannot obtain malpractice insurance. You can make up an entirely ancillary kind of insurance. It does not change the fact that GPTs are not legally allowed to represent clients, so they cannot be liable to clients for legal advice. Seeing as how you think GPTs are so useful here... why are you asking me these questions when a GPT should be perfectly capable of providing you with the policy considerations that underline attorney licensing procedures.