Work on Security Instead of Friendliness? — AI Alignment Forum