Averting AI Armageddon

As Russia and China increasingly aim to modernize their nuclear arsenals and integrate AI into their militaries, it is essential for policymakers to be aware of the risks posed by the AI-nuclear nexus. Dealing with China and Russia on issues at this nexus is likely to be difficult in the current diplomatic and military context, characterized by increasingly strained bilateral relationships between the United States and both China and Russia, along with an uptick in coordination between Beijing and Moscow. Nonetheless, there are still various steps that U.S. policymakers could take to bolster deterrence and stability with respect to these issues. These include:

• building knowledge and competency around issues at the AI-nuclear nexus;
• integrating AI into diplomatic initiatives related to nuclear and other strategic risks, and vice versa;
• establishing and promoting norms for the safe use of AI in relation to nuclear arsenals and other strategic capabilities;
• developing policy and technical criteria for assessing exactly how and when to keep humans in the loop on all nuclear-related processes;
• including AI technologies as a factor in oversight and reviews of the U.S. nuclear arsenal;
• investing in AI-enabled cyber and space capabilities to enhance defense and resilience, reduce incentives to attack those areas, and mitigate entanglement risks;
• consulting closely with U.S. allies about how AI will shape extended deterrence calculations related to both nuclear and conventional capabilities; and
• pursuing a comprehensive set of risk reduction and crisis management mechanisms with China and Russia while recognizing the obstacles to progress.

Failing to take these steps could leave the country and the world dangerously exposed to risks and ill-prepared to seize any opportunities arising from the increasingly salient AI-nuclear nexus.

Introduction

The nuclear order among major powers has fundamentally changed over the last few years. In particular, the People’s Republic of China (PRC) is building up its nuclear arsenal to make it numerically larger and technologically more sophisticated. As a result, the bipolar nuclear order—led by the United States and Russia—has given way to a more volatile tripolar one.¹

The same period has also seen the rapid advancement of artificial intelligence (AI) technologies, including for military applications.² AI breakthroughs have led many commentators to compare and contrast the technology’s current state with the advent of nuclear weapons and the field of nuclear arms control.³ Beyond those comparisons, though, the two areas overlap at what this report calls the AI-nuclear nexus. Understanding the nexus requires an assessment of each of the three countries and their militaries, their respective nuclear weapons and associated infrastructure, and potential military applications of artificial intelligence, as well as their effects on the global security environment.

This report provides an overview of this AI-nuclear nexus by synthesizing these disparate bodies of analysis and assessing the emerging dynamics in a way that is accessible to foreign policy and national security generalists who are not experts in any or all of the aforementioned topics. It proceeds in six parts. The paper starts by detailing trends in AI and potential military use cases. Then it examines the AI-nuclear nexus across three general categories: nuclear command, control, and communications; structural elements of the nuclear balance; and entanglement of AI-enabled conventional systems with nuclear risks. Third, the report assesses key factors affecting the level of risk, namely the role of humans, the degree to which AI systems become a single point of failure, and AI’s potential effects on the offense-defense balance. Fourth, it summarizes U.S., Chinese, and Russian efforts to modernize their nuclear arsenals and integrate AI into their militaries. Fifth, it lays out the diplomatic and military context and constraints for dealing with China and Russia on issues at the AI-nuclear nexus. Sixth and finally, the report concludes with recommendations for U.S. policymakers for how to approach these issues in ways that bolster both deterrence and stability.