I actually don’t think this is the case, since it’s just emulating actual behavior. In this case, real humans are talking like that, so if the AI adopts that in its training data, it’s not nonsensical.
It’s not really different from new slang getting passed in as training data and the AI using it.
I don’t agree with it, but the reason this religious argument (and most challenges of religion) falls flat, is because, to the true believer, their God is infallible, and so the idea of God making a mistake like that is on direct conflict with their core beliefs.