AI risk discourse shifts towards centering people; but can we trust it?