Welcome to the 89th edition of our weekly newsletter! I’m Chenghao Sun, a fellow with the Center for International Security and Strategy (CISS) at Tsinghua University, Council Member of The Chinese Association of American Studies and a visiting scholar at the Paul Tsai China Center of Yale Law School (fall 2024).
ChinAffairsplus is a newsletter that shares articles by Chinese academics on topics such as China’s foreign policy, China-U.S. relations, China-Europe relations, and more. This newsletter was co-founded by my research assistant, Xueyu Zhang, and me.
Through carefully selected Chinese academic articles, we aim to provide you with key insights into the issues that China’s academic and strategic communities are focused on. We will highlight why each article matters and the most important takeaways. Questions and feedback can be addressed to sch0625@gmail.com
Today, we have selected an article written by Xueyu and me, which focuses on the AI-Nuclear Nexus.
Summary
At present, strategic anxieties in the “Third Nuclear Age” and the logic of pursuing deterrence through technological integration are driving some nuclear-armed states to explore incorporating AI into nuclear weapons systems. AI contributes to nuclear deterrence through three major pathways: strengthening nuclear strike capabilities, improving nuclear situational awareness, and empowering nuclear command, control, and communications systems (NC3).
The AI-nuclear nexus may alter the strategic balance, undermine strategic stability, and weaken psychological and normative constraints such as the nuclear taboo. It may also challenge the nuclear non-proliferation regime through “intelligentized” arms expansion. Looking ahead, stronger bilateral and multilateral dialogue and improved crisis-management mechanisms are needed to help cool down an emerging intelligentized arms race.
Why It Matters
The integration of AI and nuclear weapons has drawn significant attention from policymakers, military practitioners, and academics worldwide, as it may directly undermine strategic stability, reshape crisis decision-making processes, fuel nuclear arms races, and increase the risk of nuclear war.
Although existing scholarship has revealed the far-reaching strategic implications of the militarization of AI and the intelligentization of nuclear deterrence systems, further exploration of the underlying motivations is required, specific integration pathways, and their mechanisms of impact. This paper examines the process and implications of integrating AI into nuclear deterrence systems, explores the motivations behind this integration and its strategic consequences, and proposes preliminary approaches to addressing these challenges through global governance and bilateral and multilateral cooperation. In doing so, it seeks to fill gaps in the existing literature and provide meaningful implications for policymaking.
Key Points
1. Motivations for the Integration of Artificial Intelligence and Nuclear Weapons
(1) Securing First-Mover Advantages and Alleviating Strategic Anxiety
As geopolitical and technological competition among major powers intensifies, the international community increasingly recognizes AI’s military application as a key factor shaping future warfare. The pursuit of first-mover advantages, intertwined with strategic anxiety, has prompted some states to explore AI to strengthen nuclear strike capabilities, maintain strategic advantages, and safeguard national security.
On the one hand, nuclear powers such as the United States and Russia are exploring AI applications to enhance nuclear deterrence and consolidate strategic advantages. On the other hand, states such as the United States remain highly vigilant about other countries’ technological development, viewing their growing military capabilities as a potential threat and increasing defense investment accordingly.
(2) Reshaping the Strategic Balance through AI as a Technological Lever
For states with smaller nuclear arsenals, AI may function as a ‘technological lever,’ shifting deterrence from numerical superiority toward precision and helping offset disadvantages in arsenal size. The integration of AI and nuclear weapons may compensate for the limited size of some states’ nuclear arsenals and enhance nuclear deterrence without expanding their stockpiles. Meanwhile, advances in conventional weapons, including unmanned combat platforms and precision-strike capabilities, may enable non-nuclear states to counter nuclear-armed states, potentially prompting the latter to strengthen their nuclear forces. For major nuclear powers such as the United States and Russia, AI-nuclear integration may also enable an overall enhancement of nuclear strike capabilities. AI-enhanced ISR may also create ‘one-way transparency,’ allowing technologically superior states to better track adversary nuclear deployments and missile movements, potentially weakening their second-strike capability.
(3) Integration as Deterrence
Even without translating into actual destructive capabilities, the trend toward AI-nuclear integration itself can have a deterrent effect. First, even if such capabilities are not fully realized, the belief that a state possesses them may be sufficient to generate deterrence. Second, a key feature of AI-enabled military development is its potential to produce “discontinuous improvements” in capabilities, with weapon performance undergoing qualitative rather than gradual changes at certain stages. Thus, even limited experimental integration may cause states on the “opposite side” of technological leaders to worry that the survivability of their nuclear weapons is under threat.
2. Pathways to AI-Nuclear Integration
(1) AI Empowers Nuclear Weapons Deployment and Enhances Strike Effectiveness
First is “effective deployment,” which involves applying AI to optimize the performance of nuclear weapons delivery platforms, enhance deployment flexibility, and balance nuclear strike capabilities with nuclear survivability. Second is “precision guidance,” which combines reinforcement learning with guidance technologies to enhance missiles’ adaptability in complex electromagnetic and weather environments and improve strike effectiveness. Finally, “coordinated penetration,” which uses AI to enhance missiles’ ability to penetrate adversary air defense and missile defense systems, reach target areas, and complete strike missions.
(2) AI-Enabled Nuclear Threat Warning and Situational Awareness Systems
AI-enabled unmanned autonomous reconnaissance systems are becoming an important complement to existing nuclear situational awareness capabilities. As technology advances, unmanned surface vehicles (USVs), unmanned underwater vehicles (UUVs), autonomous underwater vehicles (AUVs), and unmanned aerial vehicles (UAVs) can operate for extended periods in challenging environments, including polar regions, the open ocean, and the deep sea, improving the efficiency of military operations. At the same time, AI can process massive amounts of multi-source data to assist in identifying, classifying, and continuously tracking potential nuclear force targets, thereby enhancing nuclear situational awareness.
(3) AI-Enabled Nuclear Command, Control, and Communications Systems (NC3)
AI can provide decision support. In high-pressure environments where there is little room for maneuver, the introduction of AI can serve two purposes: expanding the range of available decision options and improving the quality of decision-making. Meanwhile, AI can help enhance the security and resilience of nuclear command, control, and communications (NC3). In infrastructure maintenance, AI can draw on machine learning, data mining, and predictive analytics to continuously monitor relevant weapons systems, identify equipment degradation, abnormal conditions, and potential vulnerabilities, and support early warning of failures, maintenance scheduling, and vulnerability remediation.
3. The Impact of AI-Nuclear Integration on Strategic Stability
(1) The Multifaceted Impact of AI-Nuclear Integration on Strategic Stability
Firstly, the integration of AI and nuclear weapons will alter the static nuclear balance of power among states. Static capabilities primarily refer to the number of nuclear weapons a state possesses and its credible delivery capabilities, while “static force stability” refers to a state of nuclear equilibrium in which both sides possess reliable nuclear retaliatory capabilities. The introduction of AI is changing the traditional paradigm in which the number of nuclear weapons determines whether a state reaches the threshold of nuclear deterrence.
Secondly, the integration of AI and nuclear weapons will undermine crisis stability among nuclear-armed states, compress decision-making time, and increase the risk of war arising from misunderstanding and miscalculation. First, AI-assisted decision-making may increase the likelihood of states taking risks in nuclear crisis scenarios. Second, AI may amplify the risk of miscalculation in crisis decision-making environments. Third, AI-assisted decision-making systems may accelerate the pace of nuclear war, causing crises to escalate more rapidly into nuclear conflict.
Thirdly, the integration of AI and nuclear weapons will undermine arms race stability among nuclear powers. Major powers will face greater uncertainty in assessing each other’s technological capabilities, military capabilities, and strategic intentions, increasing the risk of miscalculation and intensifying the security dilemma. States may also adopt a “worst-case assumption,” assuming that an adversary’s technological capabilities are superior or rapidly catching up, thereby increasing investment and driving a spiral of arms race escalation.
Finally, the integration of AI and nuclear weapons may weaken the ethical constraints in traditional nuclear strategic competition. On the one hand, AI systems may struggle to grasp the non-quantifiable political and psychological elements of brinkmanship and resolve signaling, making it difficult to perceive the fear, hesitation, and willingness to compromise of opposing decision-makers under high-pressure conditions or to interpret diplomatic signals in a timely manner. On the other hand, AI’s quantitative assessment of the risks and benefits of nuclear strikes, combined with the ambiguity of responsibility arising from AI-assisted decision-making, may weaken the constraining force of the “nuclear taboo.”
(2) Emerging Risks of AI-Nuclear Nexus: Technical Deficiencies, Loss of Control, and Ethical Challenges
First, unlike traditional nuclear forces built around missile numbers and delivery capabilities, AI-enabled nuclear deterrence systems rely heavily on algorithms, data, and communication links. If critical nodes are attacked or poisoned, the system may generate misidentification, false warnings, or anomalous decision recommendations, escalating risks during crises. Second, “Human-in-the-Loop” (HITL), which retains human intervention, may exist only formally under extreme high-pressure conditions, making effective human intervention difficult in practice. Finally, AI systems’ self-learning and adaptive capabilities may give rise to “value drift” and risks of goal misalignment.
4. Prospects for International Governance amid the Intelligentization of Nuclear Weapons
(1) Deficiencies and Challenges in the Existing Governance Framework
First, some technologically advanced states continue to deepen the integration of AI and nuclear deterrence systems to maintain security advantages and deter “adversaries,” intensifying the security dilemma among states. At the multilateral and bilateral levels, the international community has proposed several initiatives and minimum consensuses on AI and nuclear risks, but these remain limited in scope, weak in binding force, and supported by inadequate verification mechanisms, making it difficult to effectively constrain the process of “intelligentized arms expansion”. Moreover, AI and other intelligent technologies are dual-use in nature. Their potential to generate discontinuous improvements in nuclear strike capabilities also challenges traditional nuclear disarmament verification frameworks centered on warhead numbers and delivery-system scales.
(2) Rebuilding Nuclear Strategic Stability in the Era of “Intelligentized” Arms Competition
Rebuilding strategic stability in the “intelligentized arms era” should begin with aligning risk perceptions, focus on crisis stability and arms race stability, and aim to prevent AI military applications from undermining nuclear deterrence stability and intensifying the security dilemma. This requires a gradual governance pathway from risk research to policy communication, from expert consensus to behavioral norms, and from bilateral exploration to multilateral adoption.
First, the international community should further support the United Nations and relevant disarmament bodies in conducting systematic research on AI-nuclear integration, while promoting cross-national and interdisciplinary arms-control communities to deepen shared understanding of related risks. Second, the risks of AI-nuclear integration should be incorporated into bilateral dialogues among nuclear powers, using official, semi-official, and Track II mechanisms to strengthen communication, build mutual trust, and reduce miscalculation. Third, building on bilateral dialogue, relevant consensus should gradually be extended to the multilateral level. For example, the China-US principle of “ensuring that nuclear weapons remain under effective human control at all times” could gradually become a basic principle accepted by the P5 nuclear-weapon states.
Conclusion
As strategic competition among major powers intensifies and the military potential of emerging disruptive technologies becomes increasingly apparent, the integration of artificial intelligence (AI) with nuclear weapons has become a significant trend in the development of nuclear forces that can no longer be overlooked. In today’s complex and volatile international security environment, persistent strategic mistrust and acute security dilemmas among some major powers, compounded by accelerated efforts by the United States and other countries to advance military applications of AI, make the systemic risks arising from AI–nuclear integration impossible to ignore. If decisions on the use of nuclear weapons were ultimately left to algorithms, humanity would face grave risks.
In the age of AI, China should further engage in and help lead efforts to strengthen global strategic stability, contributing Chinese wisdom and proposals to address the risks of AI–nuclear Nexus. On the one hand, China should actively participate in the governance of military AI safety and security, promote the development of broadly representative and universally applicable international norms, and advocate fundamental principles. On the other hand, it should strengthen strategic communication and dialogue on governance with major nuclear powers and leading AI powers, exploring practical cooperation on crisis management, the verification of early-warning information, and the boundaries between human and machine decision-making. Only by encouraging countries to move from mutual deterrence toward joint risk governance can we prevent AI from becoming a catalyst for nuclear war and lay a firmer foundation for lasting peace in the age of AI.
About the Author
Sun Chenghao (孙成昊): Professor Sun is a Senior Fellow and the head of the U.S.-Europe program at the Center for International Security and Strategy (CISS), Tsinghua University. He is a member of Munich Young Leaders with the Munich Security Conference, a visiting scholar at Yale Law School (2024), a visiting fellow at the Centre for Humanitarian Dialogue (2025), and a fellow at the Arms Control Negotiation Academy (2025-2026). His research interests include U.S. domestic and foreign policy, China-U.S. relations, transatlantic relations, AI, and International security, etc.
Zhang Xueyu (张学玉): Zhang Xueyu is a Research Assistant at the Institute for Global Cooperation and Understanding, Peking University.
About the publication
The Chinese version of this article was published in the Journal of Strategy and Decision-Making (《战略决策研究》). This journal is a CSSCI Extended Edition source journal published by Guangdong University of Foreign Studies. Focusing on international affairs, the journal emphasizes international strategic studies and addresses major contemporary issues at the global, regional, and country levels. It values the academic rigor, contemporary relevance, and innovativeness of research and is committed to advancing theoretical and policy research on international strategy.
