In this scenario OpenAI would have to have Huggingface's full cooperation in the deception, right? That seems challenging to achieve.