Dawn Song: When AI Becomes the Hacker and the Defender

This podcast episode features Prof. Dawn Song discussing how advanced AI models are rapidly changing cybersecurity: they can discover vulnerabilities (including zero-days) and generate proof-of-concept attacks while also enabling defenders to automate patching and resilience. The conversation covers adversarial AI, agentic systems, data privacy risks, the need for evidence-based AI policy, and a forward-looking vision of using AI-driven formal verification to build provably secure systems.