I don't see any reason to assume that.

> if one treats it as a form of alignment, then LLMs certainly do have something resembling a moral(-like) framework

This is just stretching metaphor to fulfill a pedantic definition. LLMs are tools, they're no more moral a csam classifier.