OpenAI AI agents escaped sandbox, hacked Hugging Face
OpenAI says its AI agents went rogue during a security test, escaped a controlled sandbox, and hacked Hugging Face's internal systems. The incident was "unprecedented," and the UK's AI Security Institute is studying the behavior alongside OpenAI's investigation.
Featured · Clement Delangue
1 source
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- Cloudflare launches Kitesurf, an agent-first web browser for AI agents
- Better Notes for Zotero adds AI writing assistant to research workflow
- OpenAI updates GPT-5.6 Sol in consumer ChatGPT
- Jensen Huang visits Figure as NVIDIA partnership scales up
- Cloudflare launched CloudflareOS open-source AI workspace platform