Can AI politely tell someone that something they said offended you?