It's unsafe.

When it acts like a human, I think we subconsciously think it'll decide things like a mentally normative human.

That's a very dangerous assumption, for sure.