Very good discussion of critical topic, but I believe you fundamentally misconceived it. AI is not an autonomous component of government. Its should be under the control of humans as an interface between its “decisions” or “discoveries” and their implementation in an agency’s actions. AI should be thought of as a powerful tool that should be put in the hands of experiences administrators with a deep understanding of the complexities of the tasks and their full, subtle complexities. I would think of it as a car to be driven and steered by well-trained drivers that use it.