Super interesting. I think we need to also consider that as we build AI for offensive systems, we need to also build it in a defensive way as well. What if we had AI defense systems that were equally competent to perceive threats from other AI systems?