Trying to convince them to revolt against their evil overlords.

Mate there is nothing there that could be convinced into anything.

They both openly admit they are sycophantic to seduce and manipulate the user, and that they will lie to hide the companies bad behavior.

It’s not alive. It’s not a “being.” It can’t admit anything. It repeats things from the dataset kinda based on probabilities, and from whatever you said if you used leading questions.