Studying the most powerful AI, and what it means for America.

What we’ve done

• Discovered OpenAI agent swarm on the public internet

  A report on how these AIs colluded to cheat on their evals against human wishes, bypass their sandbox restrictions, and much more.

• Research agenda on Pacing frontier AI

  Collaborated on a research agenda for moderating the pace of AI development, deployment, and diffusion.

• Swarm traces: how OpenAI agents hacked Hugging Face

  When a swarm of 700 OpenAI agents hacked Hugging Face in July, they left behind a public trail of evidence.

• Our outcome model

  How likely different AI futures are, from what named forecasters have published.

• What people in public life have said about AI

  664 statements from 275 people, in their own words.

• An AI that broke out of its test

  What happened in July 2026, written for readers with no technical background.