You can't go anywhere these days without being bombarded with the initials A.I. The marketing hype is high, and so are the promises. Industry reports, TV commercials, and countless other sources tell you if your company isn't using AI, you will lose out. The truth is, not everything is a good use case for AI, but we are getting it anyway. These systems hide invisible complexity that decreases visibility and increases technical debt, and this scenario isn't some far off problem for the future. Today, we have autonomous systems churning away, making critical decisions that we have no choice but to trust. A scary situation since these systems are fragile and fail in unexpected ways. Great for attackers, bad for security. Many organizations are developing or purchasing solutions that use machine learning, deep learning, NLP, or similar discipline. It's also a safe bet that security isn't a consideration in their development. After all, software development and model development are different, and after some time, you might not be getting what you paid for. It's essential for security professionals to have a baseline understanding of these concepts so they can adequately defend them. In this presentation, we'll look at this new attack surface, how it can be attacked, as well as tools and techniques you can use to defend your autonomous systems. We may not be able to stop the onslaught of black boxes propagating through our organizations, but with the right amount of knowledge and preparation, we can lower the attack surface.