Can AI keep from being so easily offended?