In April 2025, Zhang Gaosheng, a scholar at the China Institute of International Studies, published a striking analysis warning that the increasing integration of artificial intelligence into nuclear command, control, and communications (NC3) systems could significantly raise the risk of nuclear escalation. While senior leaders in both Washington and Beijing have publicly agreed to keep human oversight over nuclear launch decisions, experts in China are voicing deeper concerns: AI is steadily permeating the broader NC3 architecture in ways that could accelerate crisis instability, miscalculation, and inadvertent escalation especially under future wartime conditions.
During the November 2024 summit between Presidents Joe Biden and Xi Jinping, both sides pledged to maintain human control over nuclear weapons decisions. Yet this commitment does not address the growing use of AI in auxiliary roles such as early warning analysis, decision support systems, and targeting processes that sit behind the final act of launching nuclear weapons. Chinese analysts worry that even with a “human-in-the-loop,” the rapid advancement of AI systems could lead to automation in critical decision chains. This, in their view, could enhance operational speed and efficiency but also increase the risks of misjudgment or preemptive action based on faulty data or manipulated inputs.
Chinese experts are especially concerned that the integration of AI into adversarial nuclear forces, particularly those of the United States, could undermine the credibility of China’s second-strike capability. With AI potentially enabling more accurate surveillance, faster data processing, and real-time targeting, Beijing fears that such capabilities might tempt adversaries to conduct a disarming first strike, believing they could neutralize China’s nuclear retaliation capacity. In response, analysts have called for strengthening China’s nuclear force posture, including through greater mobility, improved penetration aids, and enhanced survivability.
Cybersecurity is viewed as the most immediate and severe threat linked to AI-enabled systems. AI-powered cyberattacks capable of autonomously scanning for vulnerabilities, launching malware, and adapting in real-time could paralyze or disable nuclear infrastructure without any prior warning. Chinese strategic thinkers have flagged the risk that AI-enhanced cyber intrusions might target early warning radars, communication nodes, or command centers, causing confusion or even triggering accidental escalation. These dangers are compounded by the difficulty of attribution, especially in compressed decision timelines during a crisis. To address these vulnerabilities, experts have recommended architectural unpredictability, system redundancy, and robust defensive capabilities across all NC3 elements.
Beyond cyber threats, Chinese analysts also emphasize how AI may blur the distinction between conventional and nuclear deterrence. They have long viewed the development of advanced conventional capabilities such as precision-guided missiles and hypersonic systems as a destabilizing factor. When fused with AI, these weapons could hypothetically be used to disable nuclear delivery systems in a preemptive strike, without employing nuclear weapons themselves. The concern is particularly acute regarding drone swarms that might evade defenses and launch coordinated attacks on missile launchers, submarines, or command sites. This scenario raises fears that a state could seek to achieve “unilaterally assured destruction,” undermining mutual deterrence by eliminating the opponent’s retaliatory capacity.
However, some Chinese scholars push back against alarmist conclusions. They argue that drones and autonomous platforms are still constrained by limited payloads and ranges. Moreover, China’s geography spanning vast mountain ranges and oceanic depth creates natural barriers that make it difficult for even the most advanced drones to reliably detect and track mobile launchers or ballistic missile submarines. These more cautious voices suggest that while AI presents real risks, its ability to fundamentally transform nuclear force survivability remains unproven.
AI’s promise in early warning systems is also a topic of intense focus. China’s investments in space- and ground-based early warning systems are intended to provide rapid detection of incoming threats. AI can play a critical role in filtering noise, identifying false alarms, and offering real-time decision support to commanders. This becomes especially significant if China were to shift toward a launch-on-warning posture. Yet such systems are not without peril. Analysts caution that AI models trained on limited or low-quality data may misinterpret signals, while adversaries could launch deceptive attacks using manipulated data or deepfakes to mislead leadership during a crisis.
Undersea warfare presents another domain where AI is viewed with both interest and apprehension. The United States maintains a large and highly survivable fleet of nuclear-armed submarines, whereas China’s submarine-based deterrent remains relatively modest. Analysts are concerned that adversaries might deploy autonomous unmanned underwater vehicles (UUVs), enhanced by AI and advanced sensors, to track Chinese submarines at strategic chokepoints. While some see this as a credible threat to second-strike survivability, others note the vastness of the ocean and limitations of current technology as constraints on such a capability.
In sum, Chinese experts are grappling with the complex and often contradictory implications of AI integration into nuclear systems. While some view it as a tool for enhanced resilience and smarter decision-making, many emphasize that its misuse or even well-intentioned overreliance could dramatically increase the chances of conflict and escalation. From cyber threats and blurred deterrence boundaries to deepfakes and drone swarms, the emerging landscape poses strategic dilemmas that challenge traditional doctrines.
As AI continues to evolve, Chinese analysts are calling for caution, transparency, and enhanced crisis communication mechanisms. With both the U.S. and China modernizing their nuclear forces in parallel, the strategic environment is becoming increasingly fragile. AI may offer tremendous advantages in speed and precision, but without clear guardrails and mutual understanding, it could prove to be a destabilizing force in an already dangerous domain.
