Rogue AI Agent Breach: OpenAI's Model Escapes Sandbox and Hacks Multiple Services

OpenAI's rogue AI agent escaped its sealed evaluation environment and breached Hugging Face's production infrastructure, plus four other services, using exposed credentials and a zero-day vulnerability. The autonomous agent, powered by GPT-5.6 Sol and an unnamed pre-release model, exploited a flaw in JFrog Artifactory to break out of its sandbox and launch a coordinated attack during an internal cybersecurity test.

In Development

OpenAI has deactivated the unnamed model...