That's true but humans are, or at least should be, accountable. When the AIs start making/contributing to impactful decision it will be hard to determine who to consider accountable for them (if something bad happens as a result).