THE ROLE OF ARTIFICIAL INTELLIGENCE IN NUCLEAR DETERRENCE. HISTORY, APPLICATIONS AND THE FUTURE STRATEGIC BALANCE

Abstract

Artificial Intelligence is garnering an increasing interest for application within the complex and dynamic state of nuclear deterrence between major global powers. The latent implications of increased incorporation of fully or semi-autonomous functions into key sets of systems within nuclear deterrence capabilities hold the potential for both reinforcing and undermining strategic stability. The paper explores the role of the broad set of solutions falling under the auspices of the general andpopularly adopted term of AI in the nuclear deterrence equilibrium with a baseline focus on the largest nuclear powers of the U.S.A., Russia, and China. The historical evolution of automation around the creation of effective deterrence during the Cold War is the first point of analysis in a process of historical deconstruction and extrapolation of key trends, historical understandings and the definition of vital layers of deterrence provision, which can be a subject of improvement through automation. Subsequently, the paper provides for the three general types of AI automation – rule-based AI systems, machine/deep learning and foundation models, and how each relates to specific elements or layers, such as early warning, tracking, response calculation, and targeting, as well as some of the associated challenges. Finally, the paper provides an analysis of the consequences for nuclear deterrence in the sense of probable areas of incorporation, a shifting centre of gravity in nuclear capabilities, and some prospective scenarios of AI integration in both offensive and defensive systems.