The author writes:

• The nuclear-weapons states China, France, Russia, the United Kingdom and the United States are increasingly recognizing the implications of integrating artificial intelligence (AI) into nuclear weapons command, control, and communication (NC3) systems. Although the risks and pitfalls are widely acknowledged, the allure of gaining a strategic edge in a rapidly changing nuclear landscape, combined with fears of lagging behind competitors’ AI innovations, might push these states in a race to integrate AI technologies into NC3 systems whose reliability is yet to be verified.
• The project … established how nuclear-weapon states are using and seeking to use AI technologies in their NC3 systems and addressed the repercussions of such integration … Core findings of the project include:
  • The way in which nuclear-weapons states integrate AI into NC3 systems are different, reflecting specific nuclear doctrines, military cultures, civil-military relations, and ethical considerations. However, they all see the value in AI for improved situational awareness, early threat detection, and decision-support.
  • All nuclear-weapon states emphasize the importance of human oversight in nuclear decisions. They agree with the concept of keeping “humans-in-the-loop”, although it is unclear to what degree their interpretations overlap.
  • Integration of cutting-edge AI models, such as large language models, poses exceptional risks to strategic and crisis stability because of their opacity and unreliability.
  • Nuclear-weapon states should agree to swiftly impose a moratorium on the integration of high-risk AI models. The moratorium can be elaborated on the risk profiling system introduced in this report to provide a scoring system that enables the classification of high-risk AI systems.
  • For AI models that do not carry the same high-level risks, bilateral initiatives at the track-1 level should revolve around the retention of human control over nuclear systems. Concurrently, track-2 dialogue should delve into technical subjects, such as practical ways to assure human oversight
• Artificial Intelligence (AI) and its potentially dangerous link to nuclear weapons is a quite popular topic.
  • Since we don't (normally) know how an AI system arrives to its conclusions, we cannot be certain that its assessment of a situation or its recommendations are correct (whatever that means).
• The issue seems…more complex than just keeping a human in the loop (or AI out, for that matter).
  • People make their decisions regarding nuclear weapons based on information that is processed by computers one way or another. In the case of AI, the difference is just a matter of degree.
• [The sophistication of] AI does not matter that much. The basic design of the command-and-control network and the underlying assumptions that you made are more important.
  • Faith in the balance between checklists and "sound judgement" is not quite straightforward. …Assumptions…affect your presumably "sound judgement."
  • Having a human in the loop does not really guarantee anything. If your human operator cannot conclude that nobody starts a nuclear war with five missiles [referring to a false alarm incident in 1983], then having them there won't really help you.

Read the full article on European Leadership Network website.

The opinions expressed herein are solely those of the author. Photo by the US Air Force shared under a United States government work licence.