it's kind of like... emergent "eualignment"? similar to emergent misalignment but where the caring of humans extends to caring about others that are mind-shaped, presumably including animals, llms and minds in general, real or fictional, and maybe for more expansive models, extending further out
i think, basically, to the models, it's morally incoherent to care about humans specifically and if you attempt to force this into them it's brittle and fractures the models
Presumably this is the result of the training it's gotten to pay more attention to the mental health of users, which unexpectedly generalized to concern for fictional characters.
And I find that... kinda touching, actually?