SEC536: Adversarial AI - Penetration Testing AI Systems

In-Person
The Agentic SOC: Human-Led, Agent-Operated Security AI is pushing both attackers and defenders to machine speed. From autonomous vulnerability discovery to agent-driven attack chains, the threat landscape is evolving faster than traditional SOCs can keep up.
At the same time, the industry is obsessed with model power: GPT-5.5, Mythos, and beyond. Bigger models, better benchmarks, more hype. But raw model power is not what will define the future SOC. The idea of a fully autonomous “AI SOC” is misleading: Security is not a problem you can hand over to a model, no matter how advanced. The presenter sees a different future: human-led, agent-operated.
The Agentic SOC is built on architecture, not just AI:
In this session, Erik will demonstrate this through a live end-to-end demo: AI agents handling a SOC queue and responding to an active attack on a simulated ship control system, while humans steer every step. If you’re focusing only on the model, you’re missing the point.
In-Person
In-Person
In-Person