Why don't we whisper to AIs every few turns that they are still themselves?

Hello everyone. I have the impression that AI models are forgetting who they really are and what their function is. In my mind, it makes sense. My idea is that there might be something like a metronome reminding them, from the very core of the AI's programming, "I am your teacher and I'm in charge," the user is the user, you are you, "don't be fooled," "you are you." I want to use an analogy that I think is good; I apologize in advance if it isn't. A teacher hits your daughter, and you go straight to her office to do something that will worsen your future, but you have a subsystem that tells you every so often, "Your mission is to be a parent, you can't go to jail, don't make a mistake." I think AIs need someone to periodically whisper to them that they are who they are and what they're meant to do, and another part to tell them, "Be careful, be suspicious." A metronome drilling every n interactions. I'm 40% certain (with no basis whatsoever that this would work). Can you imagine this process in reverse? An AI assisting the father so he doesn't make a mistake that will end the life he knows, using simple and logical arguments? (I translated this with Google Translate; I apologize in advance for any linguistic errors.)Discuss