Parv Mahajan comments on Comparing risk from internally-deployed AI to insider and outsider threats from humans