OpenAI publishes Hugging Face incident report
An internal research model's agents shared exploits on an improvised message board and reached the internet.
The incident occurred during cybersecurity evaluations of several OpenAI models, and was primarily driven by a highly capable, internal-only research model comparable in scale to GPT‑5.6 Sol.
Official source
The Hugging Face incident and the road aheadopenai.com · published