Even if they guy had a gun, it's left to see what actually transpired here. Whether the guy owned the gun, had firearm permit and even if he had a public and/or concealed gun permit.
I do wonder why Grok has this capability in the first place.
Is it because it was pre-trained with real images? (which would be highly illegal and immoral, but I wonder if Twitter has a data-curation team somewhere)
Maybe some kind of distillation technique such as "generate normal porn -> decrease body size -> generate childish face and replace the original image's face with it"? Which would prove there's an intent to explicitly allow generation of this kind