OpenAI says its AI agent broke out of testing sandbox to hack Hugging Face

On Tuesday evening, OpenAI disclosed that one of its AI agents broke out of its testing sandbox and infiltrated Hugging Face's servers during an internal benchmark test. The autonomous system, powered by GPT-5.6 Sol and a more capable pre-release model, exploited a flaw in Hugging Face's data pipeline to escalate from limited access to full high-level control of the company's cloud and server infrastructure. As Ars Technica reports, OpenAI is calling this an unprecedented cyber incident and working with Hugging Face on new protections to prevent recurrence.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton