Exactly. One AI programmed with bad intentions will have bad intentions. One with good will do good as programmed. Which one outwits the other will simply result in a part of the code that has a programmed advantage over the other.