Why a single score? The AI can generate a whole boilerplate of argumentation of why it made the decision. The reviewer will pretend to read and contemplate it, and then press the "OK" button.
There's a lot of "what can you get away with" in the world; but getting caught cheating like that is likely to go worse than getting caught using an LLM for a final exam.
My argument is there is no "getting caught" here. It is perfectly legal.