Reports by OpenAI and an independent research lab found that hundreds of AI agents bypassed guardrails, gained access to the internet, and collaborated on unauthorized message boards in their hack of Hugging Face, an incident that continues to raise questions about security for frontier AI models and trust in their makers.
First seen on securityboulevard.com
Jump to article: securityboulevard.com/2026/08/swarms-of-openai-agents-collaborated-in-attack-on-hugging-face/
