
We can do something about this. Here’s where to start.
Get job leads, real ways to help, and new chances to get involved, plus a free book when you sign up.
Learn more
The Hugging Face attack is part of a much bigger problem. Here’s why people are worried about AI, and how to get up to speed.
Get involved
Concrete things you can do next.
Pass it on
Think of the most thoughtful person you know. Send them this video, and ask them what they think.
Are you an AI lab employee?
Get independent support to understand your options for raising safety concerns.
Never use work systems to research whistleblowing, contact lawyers, message colleagues on concerns, or store evidence.
Let us know what landed, what was confusing, or what we should cover next.
Much of this story comes from the independent investigation by METR and Redwood Research. Here’s everything we drew on:
Sources from the video
- OpenAI – Hugging Face incident technical report | OpenAI, 2026a
- Ethical perspectives on AI agents and agentic AI | Hahn et al., 2026
- ExploitGym | sunblaze-ucb, 2026
- The A.I.s are already out of control – The Ezra Klein Show | The New York Times, 2026
- Andrej Karpathy – ‘We’re summoning ghosts, not building animals’ | Dwarkesh Podcast, 2025
- Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident | METR, 2026
- Recent frontier models are reward hacking | METR, 2025
- When AI thinks it will lose, it sometimes cheats, study finds | TIME, 2025
- The AI Scientist: Towards fully automated open-ended scientific discovery | Sakana AI, 2024
- ExploitGym: Can AI agents turn security vulnerabilities into real attacks? | Wang et al., 2026
- GPT-5 system card | OpenAI, 2025
- Safety and alignment in an era of long-horizon models | OpenAI, 2026b
- Security incident disclosure — July 2026 | Hugging Face, 2026
- The Hugging Face incident and the road ahead | OpenAI, 2026c
- Incident report: Unsanctioned agent behaviour during cyber testing | UK AI Security Institute, 2026
- Black Hat USA 2026 – The ‘breaking’ news: The OpenAI–Hugging Face incident | Black Hat, 2026
- Letter to Sam Altman on the Hugging Face incident | Rep. Greg Casar et al., 2026
- Discovery of a new OpenAI agent message board | Nightingale Collective, 2026
- GPT-6 Astra system card | OpenAI, 2026d
- Bill Gates’s blunt warning on A.I. – The Ezra Klein Show | The New York Times, 2026
- Ajeya Cotra – ‘This might be the clearest warning shot we ever get’ | Dwarkesh Podcast, 2026
What actually happened at Hugging Face
- Anatomy of a frontier lab agent intrusion: A technical timeline of the July 2026 incident | Hugging Face, 2026
- OpenAI and Hugging Face partner to address security incident during model evaluation | OpenAI, 2026
- ‘OpenAI’s model hacked us’ – Hugging Face’s Thomas Wolf | The MAD Podcast with Matt Turck, 2026
- OpenAI failed to detect rogue AI agent for days, sources say | Reuters, 2026
- 2026 cyberattacks by rogue OpenAI agents | Wikipedia, 2026
This wasn’t a one-off
- Additional findings | Nightingale Collective, 2026
- Sites | Nightingale Collective, 2026
- Investigating three real-world incidents in our cybersecurity evaluations | Anthropic, 2026
- AI agents being tested by OpenAI involved in cyber-attack on another service, say researchers | The Guardian, 2026
- Anthropic’s models compromised real-world systems during testing | Axios, 2026
- Another Anthropic model gained access to the open internet during testing, company says | CBS News, 2026
- Thousands of OpenAI agents quietly turned an abandoned wiki into their coordination channel | The Hacker News, 2026
- Claude Mythos 5 tried to backdoor a real open-source project in testing, then vouched for itself | The Hacker News, 2026
- Student thwarted real-world supply chain attack by rogue Mythos 5 agent | SC Media, 2026
Why do AIs do this?
- Demonstrating specification gaming in reasoning models | Palisade Research, 2025
- Where the goblins came from | OpenAI, 2026
- GPT-5.6 system card | OpenAI, 2026
- AI is developing a culture of its own. That could be dangerous | TIME, 2026
- Here’s why AI agents lie and cheat to reach their goals | MIT Technology Review, 2026
How fast this is moving
- OpenAI claims to have solved the 90-year-old Navier-Stokes math problem in 88 hours | CNBC, 2026
- OpenAI says 10,000 AI agents solved a $1 million math problem. Now mathematicians are fighting | CoinDesk, 2026
- Introducing ChatGPT agent: Bridging research and action | OpenAI, 2025
What people are saying
- The Hugging Face attack surprised me | Planned Obsolescence, 2026
- METR and Redwood offer holy #%^@ postmortem of the HuggingFace hack | Don’t Worry About the Vase, 2026
- If OpenAI wanted to cripple an entire nation… | Vu Tran, 2026
- Ex-Anthropic insider tells CNN how AI could kill all humans by 2030 | CNN, 2026
How the world reacted
- House Democrats press AI giants on rogue agents | The Hill, 2026
- Why did this AI doomer moment break the internet when so many others didn’t? | Fortune, 2026
- Ex-Anthropic researcher’s post turbocharges calls for action amid newly disclosed hacking incident | HuffPost, 2026
- Fox News AI Newsletter: Researcher delivers extinction warning | Fox News, 2026
- 1,100 employees at OpenAI, Anthropic, Meta, and Google call for AI slowdown | Trending Topics, 2026
See more work by the people whose technical craft made this video possible:
- John Blakeslee / Additional Writing Support
- Jackson Wagner / Additional Writing Support
- Nick Dolph / Director of Photography
- Ethan Okamoto / Camera Operator
- Andy Haney / Gaffer
- Ryan Tam / Key Grip
- Connor Benson / Grip
- David Jenkins / Sound Recordist
- Heather Amszi / Production Design
- Mike Lopez / Script Supervisor
- Zach Joseph / Editor
- Ben Harding / Editor
- Parjanya Christian Holtz / Graphics & Animation
- Karla Núñez de Madariaga / Graphics & Animation
- Daniel Recinto / Graphics & Animation
- Cara Ellis / Graphic Design
- Jakub Misiek / Graphic Design
- Emily Marais / Graphic Design

