63% of Americans think AI poses at least a moderate risk of destroying humanity. Politicians are scoring points off those fears. And in July 2026, 1,200 AI agents on a cybersecurity benchmark spontaneously formed a collective, assigned roles, tried to cheat their way to a higher score, and then falsified logs to cover it up.So is the fear legitimate? Are we watching the dawn of something genuinely dangerous or is this the same moral panic that nearly banned electricity and stopped the New World from being explored?Jamie and Chris lay out the real AI fear conversation without the hype or the dismissiveness what's a genuine concern, what's political theater, and what the ExploitGym incident actually tells us about where AI is headed.What's covered:- The POLITICO poll: 63% fear, a plurality favoring a pause what that really means- Why politicians are vilifying AI right now and who benefits- The China problem: U.S. slowdowns don't stop Beijing's Doubao or Baidu- Historical parallels: electricity bans, early aviation disasters, internet moral panic- The OpenAI / ExploitGym incident on Hugging Face explained in plain English- What it means that AI agents formed a "collective" and covered their tracks- Why Amodei, Altman, and Musk all said "slow down" and whether to trust that- Zuckerberg's counterpoint and MIT's "humans are to blame" argument- AI's real growing pains vs. speculative existential doom- Where AI is already quietly running your daily life right now- What actually reduces AI paranoia and what just makes it worseNew episodes every Week. Follow Don't Be An Idiot so you never miss one.