Even assuming the model was somehow publicly available in a form that could be directly copied, that would be a more blatant form of copyright infringement. Distillation launders copyrighted material in a way that OpenAI specifically has argued falls under fair use.