This one makes me wonder: AI can find and generate attacks on an IT system faster that a hacker can - so inevitably our IT security systems are going to be turned over AI defense bots. So that sounds like Even Steven right. But then some malicious young mind that happens to be *creative* finds a hack that no-one ever thought of before, and exactly as you note, since no-one ever thought of it before AI cannot think of it at all, so in effect, these system will be exactly vulnerable to *human* attack. Thus the current big push for Agentic AI, where at least one of the idea/hope/aspirations is actually to reach credible synthetic 'thinking' ability. It would be a hoot if they achieve without ever even understanding what it is a human brain.