AI could simply deny the answer or hallucinate. AI is inauditable and the main reason TOR does not apply it. Go ahead and ask it. One answer will tell you that is (somewhat)while the other will tell you it is not(somewhat). But both answers are still vague and somewhat different.
Ask it two questions:
-is AI auditable
-is AI inauditable
It can not fully answer the question yet gives two different answers. It can not simply answer yes or no.