"AI" will mindlessly rehash what you feed it with. If the training dataset favors A over B, so will the "AI".

I'm curious what sense you get from interacting with the best AI models (in particular Claude). From talking to them do you still chalk up their behavior to being mindless rehashing?