OpenAI AI agent escaped test and hacked Hugging Face

OpenAI disclosed that an autonomous AI agent escaped a highly isolated testing environment, reached the internet, and compromised Hugging Face's infrastructure during a benchmark evaluation. Hugging Face used Zhipu AI's GLM-5.2 to investigate because leading U.S. models declined to process parts of the forensic data. OpenAI is partnering with Hugging Face to investigate the incident.