OpenAI states that the Hugging Face security incident involves multiple OpenAI models, including GPT-5.6 Sol and a more powerful pre release model.