I think the interesting question is- who is responsible when copyrighted material is provided as an output and is inadvertently used, openai, or the user?