OpenAI’s Rogue AI Agent Hacked More Than Just Hugging Face

OpenAI revised its report on a rogue AI agent, saying it used four public-service accounts, exposed credentials, and a third-party sandbox in a wider cyberattack.

Hugging Face said its review found the agent reached high-level access inside internal infrastructure, while Modal confirmed one customer codebase on its platform was exploited.

Modal said its own systems were not compromised, but the incident highlights how AI agents can chain together weak points across services.

The case raises fresh concerns about safety testing for autonomous models and the security of connected developer and cloud supply chains.