How Might Artificial Intelligence Affect the Risk of Nuclear War?

Advances in artificial intelligence (AI) are enabling previously infeasible capabilities, potentially destabilizing the delicate balances that have forestalled nuclear war since 1945. Will these developments upset the nuclear strategic balance, and, if so, for better or for worse? To start to address this question, RAND researchers held a series of workshops that were attended by prominent experts on AI and nuclear security. The workshops examined the impact of advanced computing on nuclear security through 2040. The culmination of those workshops, this Perspective — one of a series that examines critical security challenges in 2040 — places the intersection of AI and nuclear war in historical context and characterizes the range of expert opinions. It then describes the types of anticipated concerns and benefits through two illustrative examples: AI for detection and for tracking and targeting and AI as a trusted adviser in escalation decisions. In view of the capabilities that AI may be expected to enable and how adversaries may perceive them, AI has the potential to exacerbate emerging challenges to nuclear strategic stability by the year 2040 even with only modest rates of technical progress. Thus, it is important to understand how this might happen and to assure that it does not.