Think you could contain a rogue AI? Prove it

OpenAI and Anthropic have huge teams working on this stuff. You’ve got six questions and your cybersecurity instincts. Could you keep a rogue AI contained?