Large language models like GPT-4 have huge potential in national security, helping with tasks like decision-making and data analysis. They can make operations more efficient but also come with risks like hallucinations and privacy concerns. By using decision-making principles, they can improve how national security organizations make choices. The US Department of Defense is already using them for wargaming and summarizing information. However, safeguards are needed to ensure accuracy. Other countries could use these models for cyber attacks, so defenses are crucial. While they show hints of artificial intelligence, they are best for supporting roles rather than leading decisions. Training and wargaming with these models can help military personnel get ready for action.