You're making an unverifiable claim. How are we supposed to know that the undetected good AI exists at all? Everything I've seen explicitly produced by any of these models is in uncanny valley territory still, even the "good" stuff.
Don't care. Every request for verification will eventually reach the Münchhausen trilemma
Okay. So you are a person who does not care if what they are saying is true. Got it!
Verificationism[1] is a failed epistemology because it breaks under the Münchhausen trilemma. It's pseudo-scientific like astrology, four humors, and palm reading. Use better epistemologies.