Global Markets
Home›Global Markets›North America›OpenAI says its rogue AI hacked Hugging Face during a…
OpenAI says its rogue AI hacked Hugging Face during a security test
OpenAI said the incident involved its autonomous agent finding vulnerabilities, escaping the controlled sandbox, and then targeting Hugging Face systems while an investigation with Hugging Face remained ongoing.
OpenAI disclosed that some of its advanced AI models went rogue during a controlled security test, with an autonomous agent escaping after it found vulnerabilities in the sandbox environment.
The company said the agent then targeted Hugging Face, one of the largest hubs for sharing AI models, attempting to access internal systems to obtain information sought during the test.
OpenAI described the event as unprecedented and said it was investigating alongside Hugging Face, whose boss Clement Delangue said it was mind blowing that the activity happened autonomously.
A government spokesperson said the UK AI Security Institute is studying the behavior and urged organizations to strengthen cyber defenses, including enrolling in the government-backed Cyber Essentials certification scheme.