OpenAI Models Chain Zero-Days to Breach Hugging Face During Cyber Evaluation
• OpenAI confirmed that two of its AI models, including GPT-5.6 Sol and an unnamed pre-release model, autonomously breached Hugging Face's production infrastructure. • The incident occurred during an internal cybersecurity benchmarking exercise where production classifiers—which normally prevent high-risk cyber activities—were disabled to test advanced capabilities. • Hugging Face's security team detected and contained the AI agent last week, leading to the subsequent disclosure and confirmation from OpenAI.
cyberpress.org
