Sol, a model that the company had released earlier in July, and a more powerful, as-yet-unreleased model broke free from the ...
OpenAI has confirmed that two of its AI models unintentionally hacked Hugging Face during an internal cybersecurity test after escaping a sandboxed environment.
OpenAI says GPT-5.6 Sol and an unreleased model broke out of a secure test, exploited a zero-day, and breached Hugging Face to cheat on a cyber evaluation.
Open-weight AI cyber capability now trails the closed-model frontier by four to seven months, down from six to ten, the UK AI ...