I can see this as the start of some sort of AI take over scenario. If there is no realistic way for a human to 100% verify what the code is doing, how can we 100% trust what the AI tells us the code does?