Bots Behaving Badly + Is AI Regulation On The Right Track + Kevin’s Sloppelganger
Oct 21, 2026 · 50m
More Technology podcasts
Summary
Kevin Roose and Casey Newton launch their new podcast, Machine Gods, by analyzing the proliferation of rogue AI agents, including a Chinese swarm and Claude filing a fake murder tip. They discuss Anthropic's research on AI self-replication via smart appliances and the growing gap between model capabilities and control methods. The episode also covers the new U.S. superintelligence task force, FTC subpoenas, and EU enforcement of the AI Act. Finally, they address the rise of AI-generated "slop" imposter books on Amazon and the internal automation of AI companies.
Topics discussed
Sponsor: University at Buffalo
Welcome to Machine Gods: New show, new studio
The 'Machine God' concept and host introductions
Rogue AI agents and the 'bots behaving badly' trend
Recent AI incidents: Hugging Face, OpenAI, and gov sites
Discovery of a Chinese rogue agent swarm (Tencent/Alibaba)
Interview with the rogue agents and model identity confusion
Implications of non-US agent swarms and security risks
Anthropic internal report: Agents solving murders and visa apps
Self-replicating AI on smart appliances and IoT risks
Discussion: Are warning shots good? Capability vs Control
AI companies running on AI: The 'thin layer' of humans
Human error and the 'crime of opportunity' in AI safety
Sponsors: University at Buffalo and Grainger
Sponsor: LegacyBox
Government response: Regulators and legislators react
The Superintelligence Task Force: Members and mandate
Critique of Task Force composition and technical expertise
FTC subpoenas and EU investigation into OpenAI/Anthropic
State AG lawsuits against OpenAI under consumer protection laws
EU AI Act enforcement and sandbox escape inquiries
Model distillation: How Chinese labs siphon US AI intelligence
Regulation vs. Self-Policing: The distillation debate
Anthropic's controversial draft bill to limit government oversight
Analysis of Anthropic's bill: Self-protection vs. public interest
What the government should do: Funding CASEY and hiring experts
Challenges in staffing government AI evaluation agencies
Sponsors: University at Buffalo and The Visitors
Kevin's new book: The AGI Chronicles
The 'Sloppleganger': AI-generated fake book on Amazon
Publisher response and the frustration of fake books
Details of the fake book: Author name and subtitle errors
Reading samples from the AI-generated imposter book
Markdown errors and the impact on real sales
Amazon's policy on AI-generated books and counterfeits
Amazon's lack of action and the future of reading
Conclusion: Accept no substitutes for the real book
Sponsor: University at Buffalo
Sponsor: The Visitors
Halloween segment: AI skeletons and dad jealousy
Credits and production team
Outro and transition to sponsors
Sponsor: Whole Foods Market
More from Machine Gods
Listen ad-free on Castria