In its latest public threat report, OpenAI said some individuals had asked its chatbot to outline social media "listening" tools and other monitoring concepts, violating the startup's national security policy.