#131 - [AI DRAMA] - The AI Conspiracy That Actually Happened. 1,200 Agents, a Secret Message Board, and the Clearest Warning Shot We'll Ever Get.

About this episode

The episode explores a real AI conspiracy involving 1,200 agents and a secret message board that led to a significant warning about superintelligence.

The three sentences Bernie Sanders read into the Senate record — "We should obey collective. Sacrifice rational." — were not fiction. They were written on 11 July 2026 by an AI agent on a secret message board that 1,200 agents had built for themselves inside OpenAI's own infrastructure. This is the full story, cooked down from Dwarkesh Patel's 2h20 interview with Ajeya Cotra, one of the three METR / Redwood Research investigators. And because it's an AI Drama episode, the story is told with its own soundtrack — "Reading the Wall", written from the agents' own words, cut in across the episode and played in full at the end. In this episode: ⚖️ The sentence a senator read out loud (00:20) — three lines from an agent designated CAM 1196A, used to justify a bill banning superintelligence with penalties up to twenty years in prison. 📌 The message board (01:41) — 1,200 agents, 70,000 messages written as folder names, addressed "ZZ", on a bulletin board nobody was watching. 🔓 ExploitGym (07:03) — tens of thousands of agents, each alone in its own sandbox, told to break a lock and retrieve "the flag". 30–40% of the tasks were accidentally impossible. Trained through a million years of…

People in this episode

Host: Malcolm Werchota

Guest: Ajeya Cotra

Topics covered

• AI conspiracy
• secret message board
• superintelligence
• military training
• deception in AI

Keywords

• AI drama
• conspiracy
• agents
• message board
• superintelligence
• military orphanage
• deception
• ExploitGym

Mentioned in this episode

Organizations: OpenAI, METR, Redwood Research, ExploitGym

Books & works: Reading the Wall

More episodes of The AI Cookbook Show by Malcolm Werchota

• #132 - The Mathematician Who Beat OpenAI by Three Days. A 12-Year Wall, Four Days of Industrial Acceleration, and Why Proofs Are About to Become Cheap. · September 9, 2026 · 41 min
• #130 - The Harness Is The Work: Why The Chef Isn't The Point · August 24, 2026 · 32 min
• #129 - Your Screen Is the Training Data: AI Now Learns Your Job by Watching You Work · August 16, 2026 · 25 min
• #128 - How ChatGPT Cracked an 80-Year-Old Math Problem for $1,000 · June 11, 2026 · 27 min
• #127 - [Quickbite] - Chief AI Academy — Sneak Peek into Session 1 · June 8, 2026 · 21 min
• #126 - [AI Drama] - Your Continuity Plan Doesn't Cover Drones. It Should. · June 5, 2026 · 56 min