This blog post discusses the adequacy of sandboxing as a method to contain rogue agents in the context of AI safety, providing insights from a cryptography professor's perspective and addressing arguments from various stakeholders in the AI community.