I understand Elon and Pmarca to be saying that this behavior is programmed in by 'safety teams' at Meta.

It is an odd way of calcifying late-20th century American upper class anxiety. If we can't give the engineers a different moral goal, it will become the default mode of knowledge. A more immediate problem than "AI will kill us" is the dev teams determined to avoid "erasure" or "digital genocide" or whatever they'd call a race/culture-agnostic AI.