Out of the Loop Again: How Dangerous is Weaponizing Automated Nuclear Systems?

2025-05-01

AI Summary

This study investigates whether and how AI-enabled automated nuclear command systems enhance the credibility and coercive efficacy of nuclear threats. Method: Employing a preregistered game-theoretic model, political-psychological experiments (with UK parliamentarians and public samples), and analysis of AI militarization, the research examines perceptions of threat credibility under varying decision-making architectures. Contribution/Results: Findings reveal that dehumanized, “tying-hands” AI launch mechanisms significantly increase threat credibility in specific contexts—outperforming both ambiguous threats and human-mediated ones. This challenges foundational assumptions in classical nuclear deterrence theory regarding controllability and rational communication. The study provides the first empirical evidence that AI-driven automation can reconfigure the psychological foundations of deterrence. Critically, it demonstrates that heightened technical credibility does not necessarily improve strategic stability; instead, it may exacerbate misperception and crisis escalation, revealing structural risks to international security.

Technology Category

• Humans and AI: Voting
• Game Theory and Economic Paradigms: Adversarial Learning
• Philosophy and Ethics of AI: Safety, Robustness & Trustworthiness

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Are nuclear weapons useful for coercion, and, if so, what factors increase the credibility and effectiveness of nuclear threats? While prominent scholars like Thomas Schelling argue that nuclear brinkmanship, or the manipulation of nuclear risk, can effectively coerce adversaries, others contend nuclear weapons are not effective tools of coercion, especially coercion designed to achieve offensive and revisionist objectives. Simultaneously, there is broad debate about the incorporation of artificial intelligence (AI) into military systems, especially nuclear command and control. We develop a theoretical argument that explicit nuclear threats implemented with automated nuclear launch systems are potentially more credible compared to ambiguous nuclear threats or explicit nuclear threats implemented via non-automated means. By reducing human control over nuclear use, leaders can more effectively tie their hands and thus signal resolve. While automated nuclear weapons launch systems may seem like something out of science fiction, the Soviet Union deployed such a system during the Cold War and the technology necessary to automate the use of force has developed considerably in recent years due to advances in AI. Preregistered survey experiments on an elite sample of United Kingdom Members of Parliament and two public samples of UK citizens provide support for these expectations, showing that, in a limited set of circumstances, nuclear threats backed by AI integration have credibility advantages, no matter how dangerous they may be. Our findings contribute to the literatures on coercive bargaining, weapons of mass destruction, and emerging technology.

Problem

• Assessing credibility of automated nuclear threats
• Exploring AI integration in nuclear coercion
• Evaluating risks of weaponizing autonomous systems

Innovation

• Automated nuclear launch systems enhance threat credibility
• AI integration improves nuclear coercion effectiveness
• Reduced human control signals stronger resolve