They are antithetical really. AI is fundamentally about undefined behavior because we cannot define it better algorithmically (putting aside the AI algorithms t

They are antithetical really. AI is fundamentally about undefined behavior because we cannot define it better algorithmically (putting aside the AI algorithms themselves). Security is about avoiding such undefined behavior and only doing things that we expect. At best you have a very secured sandbox to keep the AI in, away from anything but user input and training data.