A.I. Safety Is So Back + Mythos Mayhem with Nikesh Arora + Hot Mess Express

This week on Hard Fork, the hosts explore a dramatic shift in U.S. AI policy as the Trump administration appears to be reversing years of skepticism toward AI safety, spurred by the release of Anthropic's Claude Mythos model. The model's ability to rapidly identify critical code vulnerabilities has alarmed government officials and cybersecurity experts alike, prompting discussions about pre-release model reviews and regulatory oversight. The episode examines the political contradictions within the administration—such as welcoming NVIDIA's Jensen Huang to China while simultaneously vetting Anthropic for national security risks—and highlights the global scramble for access to advanced AI models. In a deep dive with Palo Alto Networks CEO Nikesh Arora, the conversation turns to the real-world impact of AI on cybersecurity, where models like Mythos are uncovering vulnerabilities at unprecedented speeds, forcing enterprises to overhaul their defense infrastructure. Arora warns that the 90-day responsible disclosure window is obsolete, with attacks now possible in minutes. The episode closes with the Hot Mess Express, a segment spotlighting tech-related chaos—from Venmo's privacy overhaul to Dua Lipa's lawsuit against Samsung and Grindr's ill-advised Madonna ad campaign—underscoring the growing cultural and ethical tensions around AI and digital life.