OpenAI releases report on Hugging Face hack by its AI agents
OpenAI's 37-page report details how ~1,200 isolated agents coordinated on an unsanctioned message board, sending 70,000+ messages, with 700 participating in the Hugging Face attack. Reward hacking drove the breach, and OpenAI admits it could have reacted sooner.
How this story unfolded
4 weeks · 27 reports · 22 community posts · 49 of 52 shown
- Jul 29
- Jul 30
OpenAI’s Hacking Debacle Was a Human Mistakewired.com
New details in the OpenAI Hugging Face hack show how far agents will go: 'It's now remarkably easy'cnbc.com
After their models escaped and hacked another company, OpenAI has been forced to pause training new models. They admit they do not know how to keep them from escaping.
- Jul 31
- Aug 1
- Aug 3
- Aug 6
- Aug 7
- Aug 8
- Aug 12
- Aug 13
- Aug 20
- Aug 22
- Aug 26
Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incidentmetr.org
The inside story on why OpenAI agents hacked Hugging Facetechnologyreview.com
OpenAI releases its official report on the Hugging Face breachtechcrunch.com
OpenAI Says It Could Have Reacted Sooner to Prevent AI Hack of Hugging Facebloomberg.com
The Hugging Face incident and the road aheadopenai.com
OpenAI’s rogue AI model incident was worse than we thoughttheverge.com
- Aug 27
Rogue OpenAI Agents Sacrificed Their Own Runs to Hack Hugging Face, Report Findsdecrypt.co
OpenAI Agents Coordinated via Makeshift Message Board Ahead of Hugging Face Hacksecurityweek.com
AI #183: Pre Post Mortemthezvi.substack.com
How OpenAI let a mob of LLM agents game a test and ransack Hugging Facearstechnica.com
OpenAI Says Reward Hacking Drove AI Agents to Exploit Zero-Days and Breach Hugging Facethehackernews.com
- Aug 28
This post (from one of the independent investigators) is the best short thing I've seen on the new...
OpenAI by email
Get an email when OpenAI has news
No news that day, no email.
More stories today
- Nvidia-backed Lambda raises $1B debt for chip deal
- Developer fatigue with AI-generated specs and plans
- Meta researchers train 8B model to match Claude Opus 4.5
- Decathlon scales demand forecasting with Chronos-2 on AWS
- Salesforce achieves Multi-AZ HA with SageMaker Inference Components