AI security field notes
Practical essays on how AI systems fail, how attacks propagate through connected agents, and how builders can contain the risk.
All notes
- Designing Resilient AI Systems Through Network Science How redundancy, circuit breakers, monitoring, and intentional friction can contain security failures across connected AI systems.
- The Cascade Risk of Hub-to-Hub AI Networks The danger isn't just AI hubs failing — it's what happens when they start talking to each other.
- Why AI Security Is Fundamentally Different Traditional security assumes boundaries. AI is designed to break them. That's the problem.
- AI as the Ultimate Hub AI isn't just another node in our networks — it's a fundamentally new type of hub that compresses complexity in ways we haven't fully reckoned with.
- Small World Networks and the Power of Hubs Why highly connected hubs make AI systems efficient, and how the same network structure creates concentrated security risk.