Can AI react when being caught doing something wrong?