Machine Learning and U.S. Nuclear Command Control and Communications

Ramifications of novel AI technique integration into United States Nuclear Command, Control and Communications (NC3) systems

The U.S. NC3 system of systems contains everything from the nuclear warheads themselves, the platforms that carry them (submarines, aircraft, missiles), and the thick and thin line communications systems that allow for communication flow throughout all levels and actors within the enterprise during both peace and crisis times. This system currently holds legacy components, many of which are quickly approaching their expiration. Thus, a complete modernization of NC3 was called for by former Secretary of Defense Mattis in his written statement to the Senate Armed Services Committee in 2018: “Modernizing the Nation’s nuclear deterrent delivery systems, including our nuclear command and control, is the Department’s top priority”. In addition, the 2018 Nuclear Posture Review stated that the U.S. “will pursue a series of initiatives” to address “the critical need to ensure our NC3 system remains survivable and effective.” Additionally, the current Vice Chairman of the Joint Chiefs of Staff and then Commander of U.S. Strategic Command General Hyten conveyed in the 2018 USAF Memorandum “Subject: Next Generation NC3 Enterprise,” that integration of AI and other modern technologies are being considered in this process. 

The plausible integration of machine learning into legacy systems clearly creates novel opportunities for accelerated advantage and capabilities — but also significant likely vulnerabilities for abuse by U.S. adversaries. Experts from across industry and academia must be engaged to elucidate the risks associated with the integration of novel machine learning techniques, particularly deep learning, into all legacy systems of systems — but particularly no-fail enterprises such as U.S. NC3. 

Artificial intelligence technologies have long been integrated into NC3 – it is not a novel concept. Since the precedent has been set, the potential exists for easily integrating novel and emerging machine learning techniques into NC3 communications systems. However, little is currently known in the unclassified literature about how integration of novel machine learning techniques might affect individual NC3 subcomponents, the compounding effect that will have, or the emergent properties of introducing such techniques — much less for the entire superstructure.  

This project dives deeper into the mapping of the U.S. NC3 architecture to

  1. Utilize open-source information to as accurately as possible map the current NC3 architecture at the sub-component level,
  2. Identify which subcomponents are plausible candidates for potential integration of machine learning and deep learning-driven capabilities, and 
  3. Identify risks and opportunities that this creates for the U.S. NC3 system of systems. 


Associated External Publications: 

The Real Value of Artificial Intelligence in Nuclear Command and Control 
Authors: Philip Reiner and Alexa Wehsener 
External Publication: War on the Rocks 

When Machine Learning Comes to Nuclear Communications Systems
Authors: Philip Reiner, Alexa Wehsener and M.Nina Miller
External Publication: C4ISRNET 

Contact Us

If you want to learn more about this project or if you are interested in getting involved, please email us at [email protected] and place ML and NC3 in the Subject line.