Days after former employees said the company was being too reckless with its technology, OpenAI released a research paper on a method for reverse engineering the workings of AI models.