
Breakthrough Models and New Cyber Threats (with Evan Swarztrauber)
Until recently, concerns about AI agents conducting autonomous cyber operations were largely theoretical. The OpenAI–Hugging Face incident in July changed that, while Anthropic’s Mythos exposed software vulnerabilities buried for decades. Together, these incidents have forced policymakers to reconsider how governments and companies prepare for AI-enabled cyber threats. Can the federal government and private sector keep pace with increasingly capable models? Will open-source AI strengthen cyber defenses or make dangerous capabilities harder to contain? And can policymakers encourage companies…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.