As DOD increasingly relies on artificial intelligence, a question has arisen: How can one be sure that the AI models are working the way they should?