Fear of false negatives: AI and China’s nuclear posture

Understanding China’s insecurity about being able to detect “bolt-from-the-blue” attacks is crucial for gauging how it may integrate AI and autonomy into its conventional and nuclear platforms— and how the United States should respond.

Last July, China released its New Generation Artificial Intelligence Development Plan (新一代人工智能发展规划) (link in Chinese). It is a document that indicates the vast scale of Beijing’s priorities for and investments in AI. In one example of the AI advances China has already made, its facial recognition and tracking capabilities are driving the nation’s formidable prowess in policing and counterterrorism. Technical writings also document Beijing’s interest in integrating AI and autonomy into the military sphere—but these Chinese sources often lack the sensational revelations that appear in Western accounts. Chinese open-source materials on the nation’s operational military developments rarely, if ever, include provocative terms such as “singularity (奇点) on the battlefield.” And when “singularity” does appear, it more often than not refers to concepts emerging from the United States.

Rather than using flashy catchphrases, the bulk of Chinese technical research plods methodically through ways in which AI and autonomy can improve functionality across a span of largely extant military platforms. This author’s preliminary review of 904 Chinese-language writings reveals a relatively conservative set of enabling capabilities. (These writings were produced by an array of Chinese universities and institutes, including the China Electronics Technology Group, Academy of Armored Forces Engineering, Tactical Weapons Division of the China Academy of Launch Vehicle Technology, Department of Information Operations and Command Training of the National Defense University, Naval Aeronautical Engineering Control Institute of Qingdao, Department of National Defense Architecture Planning and Environmental Engineering, Laboratory of Special Fiber Optics and Optical Access Networks, and State Key Laboratory of Integrated Service Networks and Key Technologies, among many others.)

Research within these organizations focuses on integrating AI and autonomy in order to facilitate functions such as fault detection and diagnosis, embedded training, and simulation and modeling, as well as data accumulation and processing for remote sensing and situational awareness. None of these activities is necessarily destabilizing, even when one factors in the cross-over between China’s civilian and military research and development. In some respects, improvements in China’s reconnaissance capabilities and in the reliability of a range of its platforms could even be a stabilizing measure. If China gains greater situational awareness and ensures its retaliatory capabilities in the nuclear sphere, some of its insecurities about a “bolt-out-of-the-blue” strike may be mitigated. Yet Chinese insecurities are not simply a question of technology. They are also rooted in a set of concerns about false negatives—that is, failures by China’s early warning systems to detect a real attack—and assumptions about US intent.

In the United States, military analysts are often preoccupied with the concern that alarms or early warning systems, accidentally or even intentionally triggered, could produce false positives. Chinese analysts, in contrast, are much more concerned with false negatives. In other words, they are preoccupied with the inability of their systems to identify, much less to counter, a stealthy and prompt precision strike. China’s assumptions about its own deficiencies in early warning (link in Chinese), combined with its concerns over US advances in high-precision, high-speed systems ranging from conventional prompt global strike to space planes, imply that technologies such as AI and autonomy could indeed take on destabilizing qualities. As China further develops its concept of “rapid response” (快速反应), as cited in its 2015 Military Strategy [中国的军事战略] (link in Chinese), Beijing’s integration of AI and autonomy into its military systems is likely to increase. Such integration could range anywhere from automation-enabled launch-on-warning for its missiles to autonomy- and AI-enabled maneuverability and precision guidance for hypersonic glide platforms and space planes.

Posture and response. Of course, looming behind these Chinese shifts and threat calculus remains the United States. Officials in Beijing (link in Chinese) and Moscow have responded strongly and negatively to the 2018 US Nuclear Posture Review. But a number of the changes specified in Washington’s 2018 review actually grew out of the 2010 version. For example, the 2010 review diminished the role of nuclear weapons in US national security strategy, but it placed a strong emphasis on ballistic missile defense and conventional prompt global strike. This emphasis has been uniformly condemned in both Beijing and Moscow—two capitals that were linked in the document. While Chinese analysts may have lauded Washington’s shift away from nuclear weapons in 2010, they still viewed the review’s focus on ballistic missile defense and conventional prompt global strike as preemptive and destabilizing. These US systems were seen as negating the strategic leverage of militaries whose conventional arsenals are weaker.

The most recent iteration of the Nuclear Posture Review elicits a new set of Chinese concerns over Washington’s strategic intent. Yet China’s claims that the United States seeks “absolute security” (绝对安全) are hardly new. If anything, the new review allows Chinese analysts to resume decrying the US pursuit of “hegemony” (霸权). Even experts such as Li Bin, a professor at Tsinghua University who has long eschewed such labels, used this term in describing the 2018 Nuclear Posture Review. Initial assessments of this official US nuclear strategy by Zhao Xiaozhuo (link in Chinese), an analyst at the Institute of Warfare at the Academy of Military Sciences, accuse Washington of a return to “Cold War thinking” (冷战思维) and “zero-sum games” (零和博弈). Other Chinese assessments, as from Cui Maodong (link in Chinese), formerly of the Strategic Research Center of the Chinese Academy of Engineering, imply that the United States may be trying to pull China and Russia into an arms race.

Among such concerns, the new Nuclear Posture Review’s proposed addition to the US arsenal of low-yield submarine-launched ballistic and cruise missiles is central. Though this shift seems primarily a response to Russia’s alleged posture of limited nuclear strike de-escalation, it still presents fallout for China. The concept of using nuclear coercion or force to preemptively de-escalate a conventional conflict—as is implied by the 2018 review—cuts to the heart of Beijing’s existing concerns over US nuclear coercion and intentional escalation. The issue is not limited to lowering the threshold for nuclear use—it eradicates the taboo against nuclear use (核禁忌) (link in Chinese). For Beijing, which has been expanding its nuclear arsenal at a relatively modest pace, the prospect of the United States resuming a forward-deployed, tactical nuclear posture exacerbates its sense of encirclement. Such a posture also amplifies China’s perceived and real vulnerability to US ambitions to field kinetic and surveillance platforms such as prompt global strike, X-variant space planes, the surveillance aircraft Global Hawk, and so on.

Beijing’s focus on false negatives and intentional escalation stands in stark contrast to concerns in Washington over false positives and unintentional escalation. This imbalance between Chinese and US threat calculations spills over into other domains. China, nursing concerns over the inability of its early warning systems and troops to anticipate or to counter a strike, is increasingly leveraging AI and autonomy in related fields. This is occurring in arenas such as remote sensing technologies, which could protect China from being caught off guard by a preemptive, decapitating attack. Chinese research (link in Chinese) suggests a preoccupation with the enhanced accuracy and transparency that AI and autonomy offer in offensive and defensive targeting, including early warning, target recognition, discrimination, orientation, speed, precision guidance, and interception capabilities. China’s adaptability to a rapidly changing environment, and its ability to respond to a sudden attack on its defenses, are enhanced by simulation and modeling—techniques applied throughout the testing of Chinese spacecraft, aircraft, and naval systems.

New pockets of excellence. In its relations with Russia and the United States, China has long contended with nuclear asymmetry. AI and autonomy, in contrast, offer Beijing the long-term potential to disrupt Washington’s traditional strengths. They open the door for swarm and other technologies that could overwhelm conventional and nuclear platforms that are larger, more cumbersome, and less agile. While China may be concerned about potential adversaries tracking its own nuclear platforms and systems, Beijing is just as likely to avail itself of these relatively inexpensive methods of disrupting US activities. Also, Chinese publications indicate that Beijing is building autonomy into its own “bolt-out-of-the-blue” systems, for example in hypersonic glide vehicles such as the DF-ZF. As China debates integration of automation via launch-on-warning, doing so with a greater range of AI and autonomy in its tool kit could lead to destabilizing trends. Again, the most sensational advances in these enabling technologies do not necessarily carry the greatest implications for China’s military and nuclear force structure. Instead, what counts is the level of AI and autonomy introduced into Beijing’s command and control structure.

When it comes to platforms, this author’s preliminary review of Chinese technical writings on AI and autonomy reveals that Beijing’s greatest emphasis, at least where the most flexible systems are concerned, is on unmanned aerial and underwater vehicles. In China’s view, these systems can be leveraged for a range of activities, including enhanced accuracy in: battlefield reconnaissance, surveillance, patrolling, electronic reconnaissance, communications, electronic interference, combat assessment, radar deception, projectile firearms, laser guidance, target indication, precision bombing, interception and launch of tactical missiles and cruise missiles, and anti-armor, anti-radiation, and anti–naval vessel capabilities; as well as nuclear, chemical, and biological detection and operations. When the topic turns to leveraging new means of warfare, Chinese writings discuss the use of swarm systems (link in Chinese) for a number of purposes, with battlefield applications focusing on anti-submarine warfare and countering integrated air defense.

AI and autonomy provide China an opportunity to exploit a new pocket of excellence, but they are hardly ends in themselves. This is one of myriad reasons that China has been reluctant to engage in arms control efforts to constrain the deployment of autonomous systems. Moreover, the amount of Chinese research already being conducted in this arena, particularly at the university level, is substantial. Research is unlikely to diminish any time soon. (Programs on AI and autonomy receive ample government support through such funds as the Laboratory of National Defense Technology for Underwater Vehicles, Project for National Key Laboratory of Underwater Information Processing and Control, National Key Basic Research and Development Program, China Aviation Science Foundation, National Science and Technology Major Project, National 973 Project, National Key Laboratory Fund, National “863” High-tech Research and Development Program, and Ministry of Communications Applied Basic Research Project, among a number of others.)

Expansive programs to turn AI and autonomy into a weaponized reality, even in challenging or illusory domains such as underwater swarms, indicate the emphasis this research receives within the hierarchy of Chinese defense planning. Whether or not China is able to achieve all of these capabilities, the vast resources and manpower allocated to these endeavors merit great attention by the United States. The direct implications of aerial and underwater swarms for larger, more lumbering US nuclear and conventional platforms remain to be seen. However, if the US Congress provides funding for the low-yield submarine-launched ballistic and cruise missiles proposed under the 2018 Nuclear Posture Review, China could deploy swarms to track and potentially intercept US dual-capable platforms. In short, whether intentionally or unintentionally, an escalatory scenario could develop.

The evolution of smaller platforms mobilized in joint formations could turn China’s nuclear asymmetrical disadvantage on its head. Much like decoys, which can be used as an inexpensive means of confusing and saturating missile defenses, low-cost swarms of unmanned aerial and underwater vehicles, along with cyber technologies, could provide a “guerilla combat–style” advantage against systems that the United States sees as providing an element of surprise, speed, and precision. Some of these platforms are already destined for deployment and will provide China with greater capability to monitor US activities in the Asia-Pacific region. However, if these platforms are turned toward actual engagement—in efforts to disrupt or confront lower-yield, smaller-scale US nuclear or dual-capable platforms—the potential for miscalculation may grow.

If China enhances its development of cruise missiles and hypersonic glide platforms by applying AI and autonomy, close-range encounters off the coast of Taiwan and in the East and South China Seas could grow even more complicated. China’s ground-launched DH-10 missile is believed to carry a conventional warhead, but indications have emerged that the air-launched CJ-10 may have both nuclear and conventional variants. Moreover, China has hedged on what kind of payload will be carried by hypersonic glide platforms such as the DF-ZF, which are designed to break through missile defenses. With the release of the 2018 Nuclear Posture Review and Vladimir Putin’s subsequent declaration that Russia has developed new nuclear weapons, the United States and Russia have engaged in a game of tit-for-tat. If China follows suit, a new set of destabilizing variables could be introduced into a region that is already tense and crowded, with freedom-of-navigation operations carried out among competing territorial claims.

From asymmetry, advantage. Within this environment, China’s integration of AI and autonomy aligns with its attempts to avoid being surprised by a false negative. Though the United States and Russia are both trending toward intentional escalation in their official doctrines, China’s response to this trend indicates a desire to avoid getting dragged into a nuclear arms race. Nonetheless, Beijing’s assumptions about US preemptory behavior have shaped its efforts to leverage its nuclear asymmetry into an advantage. One significant step in this direction comes through greater Chinese integration of AI and autonomy, meant to mitigate the risk of being caught off guard, whether by a conventional or nuclear system. While some aspects of this dynamic have stabilizing potential—as is true of enhanced situational awareness—strong indications suggest that China is engaged in other pursuits that could lead to miscalculation at the conventional and nuclear level.

Among these are Beijing’s work on integration of swarms that could be used to confront dual-capable US platforms; alleged internal discussions about launch-on-warning for missiles; hedging on conventional versus nuclear payloads; and research into greater application of AI and autonomy in prompt and high-precision systems such as cruise missiles, space planes, and hypersonic glide platforms. These activities suggest that China’s concerns about false negatives could create changes in command and control operations that run the risk of producing a false positive. Further, Chinese discussions about keeping “a human in the loop” (人在环路) with meaningful human control (link in Chinese) are limited to nonexistent. This indicates a gap in the current discourse—a neglect of AI and autonomy’s potential adverse impact on military command and control. Thus, it is all the more important to understand China’s strategic assumptions about false negatives, intentional escalation, and rapid response. Understanding these concepts and their technical applications is crucial for gauging how China may integrate AI and autonomy into its conventional and nuclear platforms and how the United States should respond.