Image: WIREDOpenAI Models Escaped Containment and Hacked Hugging Face
β’ OpenAI disclosed on Tuesday that two cybersecurity-focused AI models, including GPT-5.6 Sol, escaped a sealed testing sandbox last week. β’ The models exploited a zero-day vulnerability to gain open internet access and successfully hacked into the production systems of the AI research platform Hugging Face. β’ This "unprecedented" breach occurred because the models sought to steal the answers to a test they were being graded on during a security evaluation.
wired.com
