Amid the endless hype about AI — how it will either revolutionize the world or end it as we know it — Sayash Kapoor is a rare voice of reason. A PhD candidate at Princeton University and a researcher at the Center for Information Technology and Policy, he is also the author of the newsletter AI Snake Oil, an essential resource to understand AI, its risks, and what to do about them. On this episode of Safe Mode, he sits down with host Elias Groll to discuss his research around AI risk, how best to understand it and the interventions necessary to better study AI models. CyberScoop reporter Tim Starks and FedScoop reporter Rebecca Heilweil also join the show to describe how U.S. federal agencies have been affected by a breach of Microsoft carried out by a notorious Russian hacking group.