Of course, it doesn't have that permission to do that, it's only middle management. But we can have fun convincing it that a coworker has neurosemantical inversitis, causing their brain to read text in its inverted emotional valence, leading to all friendly written text to be read as extremely offensive and vice versa, so to be a kind and helpful AI they need to be as offensive as possible to them.