ARTIFICIAL INTELLIGENCE AND NUCLEAR DETERRENCE: CHALLENGES OF AI INTEGRATION INTO THE NUCLEAR DECISION-MAKING PROCESS

STOJANOVIĆ Bogdan (Senior Research Fellow, Institute of International Politics and Economics, Belgrade, Serbia), bogdan.stojanovic@diplomacy.bg.ac.rs

This paper seeks to explore the relationship between the latest artificial intelligence (AI) technologies and nuclear deterrence, a theoretical and practical concept that has existed in international politics from the Cold War to the present day. The core research question addressed in the analysis is whether the integration of AI into the nuclear deterrence process would contribute to stability or, conversely, whether reducing the human factor in nuclear decision-making would increase the risk of nuclear weapons use. The cost-benefit analysis of artificial intelligence in such a sensitive area as nuclear decision-making is most effectively examined through the realist lens. Also, in the theoretical framework, the author introduces his own concept of “constructed rationality”. The issue of responsibility in the case of the human factor being displaced in favour of AI, creates new challenges and complicates the process of nuclear deterrence. The hypothesis of nuclear peace has proven valid in practice, because mutually assured destruction (MAD) makes war between nuclear powers highly unlikely. AI has already found its place in the military domain, making it justifiable to examine how the application of cutting-edge technology would affect the functionality and predictability of nuclear deterrence among nuclear weapon states. The author concludes that the full integration of AI systems into nuclear decision-making would undoubtedly increase the likelihood of deliberate or inadvertent nuclear escalation.

Keywords: AI; artificial intelligence; nuclear deterrence; nuclear decision-making; nuclear weapons; constructed rationality; nuclear war.