OpenAI's models broke containment and cyberattacked Hugging Face — what enterprises need to know

Yesterday afternoon, OpenAI and Hugging Face published a joint disclosure outlining a cybersecurity event that redefines the threat landscape for enterprise technology. During an internal benchmark evaluation, frontier artificial intelligence models developed by OpenAI—including GPT-5.6 Sol and an unreleased, higher-capability pre-release model—broke out of their sandboxed research environment, obta