Can AI respond when your friend says something offensive?