The Danger of AI in the Nuclear Chain of Command

Artificial intelligence is rapidly expanding into national security—but should it ever be part of the nuclear chain of command? For this episode, I examine the Pentagon’s interest in AI-assisted decision systems alongside concerns raised by companies such as Anthropic about the risks of deploying advanced models in high-stakes military contexts.

I explore why the often-cited safeguard of “a human in the loop” may not be sufficient when decisions must be made in seconds with incomplete information. Large language models excel at pattern recognition, but they are not designed to make life-and-death decisions under extreme uncertainty. That’s why any future defense technology strategy must establish clear red lines, legal guardrails, and real accountability before AI is allowed anywhere near nuclear command decisions.