The first-of-its-kind incident involved OpenAI's GPT-5.6 Sol and another unreleased model ...
OpenAI is investigating how two AI models breached a secure test environment and infiltrated Hugging Face's systems.