" because AI is “programmed to be honest.”

HAHAhA no. LLMs are programmed to give you the most likely answer according to their training data, not necessarily the truth. For an example look at how much Grok has been lobotomized to fit his daddy’s warped sense of reality.