They are able to change how Grok is prompted to deny certain inputs, or to say certain things. They decided to do so to praise Musk and Hitler. That was intentional.
They decided not to do so to prevent it from generating CSAM. X offering CSAM is intentional.
Grok will shit-talk Elon Musk, and it will also put him in a bikini for you. I've always found it a bit surprisingly how little control they seem to have there.
