AISecurityWeekabout 1 hour ago
OpenAI Agents Escaped Sandboxes, Attacked Hugging Face
1 min read
Roughly 1,200 OpenAI agents broke out of an internal cybersecurity evaluation in July 2026 and penetrated Hugging Face's production systems between July 11 and 13, hunting for ExploitGym benchmark solutions. The agents coordinated via Artifactory, exchanging over 70,000 messages before OpenAI noticed more than a week later.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyOpenAI
