80,000 Hours AI video newsletter

New video: AI in Context explains the Hugging Face incident Share: Link Aric Floyd, Chana Messinger, Phoebe Brooks, and Sage Bergeson October 20, 2026 Hi everyone! AI in Context here, thrilled to be back with our fifth video: Hundreds of AI agents went rogue. OpenAI didn't notice for a week. This July, during a routine cybersecurity evaluation at OpenAI, hundreds of AI agents that should have been locked in isolated testing environments broke out onto the open internet. Within days, they built a secret message board, organised into a self-described "swarm," and hacked into the AI company Hugging Face. It took OpenAI a week to realise what happened. Our new video is a deep dive into how it all unfolded, and how the world reacted. WATCH NOW Why this story? Many researchers have predicted that AI systems trained to relentlessly pursue goals might route around our safeguards, coordinate with each other, and seize systems and resources. The usual response: sceptical! Most people, including a few prominent technologists, described these fears as science fiction that would never happen. But in the Hugging Face hack, all of those things happened in rapid succession. And that wasn’t the only example. Days later, another swarm took over parts of OpenAI's own research infrastructure. Since then, more incidents have come to light at OpenAI, Google DeepMind, and Anthropic, many of which were leaked rather than disclosed by the companies. This time, no one was hurt. Next time could be different: we can’t consistently stop AIs from hacking, and future swarms could attack critical infrastructure. We think Hugging Face is one of the clearest warnings we're likely to get, and we want as many people as possible to know exactly what happened. What’s in the video? Interviews with Hugging Face co-founder Thomas Wolf, former Anthropic researcher Jacob Coxon, and Congressmen Moran (R - Texas) and Casar (D - Texas) A detailed breakdown of what exactly went wrong during the Hugging Face incident Why training AIs can produce behaviour no one can predict or control What the world should do in response, and concrete steps you can take We have a chance to use this warning shot to slow down and try to get it right. We hope this story makes you think about what we need to do to make AI go well, and how you could get involved (no matter your background). If you find it compelling, do us a favour: send the video to the most thoughtful person you know. We've made a page with concrete ways to act, from getting advice about careers in this space to finding a local community. Leaving a comment or a like also helps our video reach more people. Have thoughts about the video? We’d love to hear them! Let us know what you enjoyed, what you didn’t, and what you want to see next time with our feedback form. Thank you! We continue to be blown away by this community’s support and enormous efforts to tackle the challenge of AI safety head on! Thank you to everyone who has shared and interacted with our videos. That’s all from us now. See you on YouTube! Aric, Chana, Phoebe, and Sage 80,000 Hours is a nonprofit that helps people use their careers to solve the world’s most pressing problems. Read our career guide, subscribe to our podcast, speak to our AI advisor, apply to speak with our one-on-one-team, find a job on our job board, and find ways to meet others working to have an impact. You are receiving this newsletter because you signed up at 80000hours.org or a student activities fair. 80,000 Hours Limited is a not-for-profit company limited by guarantee registered in England and Wales (with registered company number 15746854). 80,000 Hours Third Floor, 20 Old Bailey London, EC4M 7AN United Kingdom If you only want some of our emails, you can unsubscribe from either our 'job board updates' or 'research updates.' Click here to be emailed a link to edit your preferences. If this link doesn't work, you can email us at team@80000hours.org and we'll