In March, a largely overlooked, 90-page Government Accountability Office study revealed something interesting: This summer, the Pentagon is getting a new AI Strategy.
Between shaping ethical norms for AI and establishing a new Chief Data and AI Officer, it’s clear top brass have big plans for the technology, though the report is light on the details. Released in 2018, the last AI Strategy laid the scaffolding for the U.S. military’s high-tech competition with China. But over the past four years one thing has become apparent: The United States needs a balanced approach to AI investment — one that doesn’t simply guard against threats, but also imposes costs on a Chinese force that sees AI as the key to victory.
Undoubtedly, a military conflict between the United States and China would be catastrophic, and every effort must be taken to avoid such an outcome through diplomatic means. Still, “thinking about the unthinkable” is necessary for U.S. defense planners to prioritize investments and identify potential deficiencies in U.S. force posture and readiness. It’s worthwhile, then, to consider a worst-case situation where, within the decade, the U.S. and Chinese militaries become embroiled in large-scale combat in or around the Indo-Pacific.
While AI is probably not going to determine the outcome of a U.S.-Chinese war, the bottom line is that the technology would augment U.S. and Chinese military capabilities in important ways. Chinese investments in AI prioritize near-term offensive capabilities. In particular, AI could play a significant role in the People’s Liberation Army’s efforts to disrupt and degrade the U.S. battle network and compensate for its own deficiencies in the undersea and electromagnetic domains. Meanwhile, U.S. investments could improve operational readiness and “jointness” — the ability of service branches to work together — for a global military that is spread increasingly thin. At the same time, AI will introduce new vulnerabilities for both China and the United States — particularly concerning data security.
How AI Could Enhance Chinese Capability
The most likely sources of a potential U.S.-Chinese conflict, such as a Chinese invasion of Taiwan or a contest over some South China Sea feature, would likely feature the full spectrum of civil and military information operations aimed at deterring U.S. intervention and degrading U.S. allies’ will to fight. AI could play a dominant role in each of these missions. The Network Systems Department of the People’s Liberation Army, for example, may try using generative language models to synthesize and amplify content on Facebook and Instagram, as it has done using botnets and other non-AI tools around Taiwanese elections. The Chinese military is also likely to wage a similar campaign to discredit U.S. military activities or sow division with partners, including Australia and Japan.
Soon after the start of a conflict, the People’s Liberation Army would likely attack U.S. sensor and communication networks, and several different kinds of machine-learning applications could aid this task. A cadre of scientists at the People’s Liberation Army National University of Defense Technology, for example, specializes in “fuzzing,” using machine learning to identify vulnerabilities in an adversary’s computer networks. Experts also point to AI’s role in attacking or defending critical infrastructure in Taiwan, Japan, Australia, or the United States.
Chinese planners also aim to use AI for electronic countermeasures and operations across the electromagnetic spectrum. For example, analysts from anquan neican (a Chinese journal for cybersecurity research) are optimistic about cognitive electronic warfare — using AI to analyze incoming radar signals, and then automatically adapting one’s own output to optimize jamming. But several other applications of AI also play a role in electronic spectrum operations. In 2020, for example, the People’s Liberation Army awarded equipment contracts for swarms of drones equipped with modular radar-jamming systems, which could be flown near U.S. carrier strike groups, military installations in Japan and South Korea, or shared facilities in the Philippines. Many systems under development by Chinese universities and military research institutions are explicitly designed to counter U.S. drone systems and swarm concepts. Chinese companies have already exported drones to Nigeria, the United Arab Emirates, and Egypt, among others. However, while some People’s Liberation Army experts contend that these drones have been “battle tested,” others are less sanguine about their capabilities in a real conflict.
Moreover, the People’s Liberation Army may attempt to use AI to enhance the lethality and reach of its surface ships and anti-access and area denial systems, which could hold U.S. forces at risk during a crisis. China’s current approach to territorial defense relies on hundreds of short- to long-range ballistic missiles that would target U.S. aircraft carriers and strike aircraft based in mainland Japan, Okinawa, South Korea, and as far away as Guam. As early as 2016, Wang Changqing, director of the General Design Department of the China Aerospace Science and Industry Corporation, claimed that the company’s next generation of cruise missiles would use AI to adapt to specific combat conditions, being capable of adjusting flight profiles and even warhead yield. Chinese defense industry engineers appear inspired by the U.S. Long-Range Anti-Ship Missile, which uses AI to improve accuracy and achieve more flexible targeting.
Finally, the People’s Liberation Army is building a wide array of autonomous vehicles and extensive undersea sensor networks that make use of AI and big-data analytics. These systems may be useful in recording and transmitting the locations of U.S. undersea vehicles, and would be crucial to overcoming the Chinese military’s disadvantages in undersea warfare. Large unmanned submarines, such as the HSU-001 and Haishen-6000, could be equipped with sea mines to deny the U.S. Navy access to undersea space between the first and second island chains, or to restrict access to the Taiwan or Luzon Straits.
Of course, AI has the potential to revolutionize Chinese operations in countless other ways, including through predictive maintenance, logistics, and back-office tasks not discussed in depth in this article. In any case, it is clear that the People’s Liberation Army is banking on the technology to create asymmetric advantages vis-a-vis the United States.
On the U.S. Side: Jointness, Maintenance, and Targeting
Succeeding in combat with the Chinese military would require the highest level of coordination between U.S. military services. The U.S. Department of Defense has long recognized the importance of joint operations, but struggles to implement this vision in practice. Since 2022, the department has been working to develop a Joint All-Domain Command and Control concept, which rests on using AI to analyze sensor data combined in a single network. While sharing data across services would enable commanders to make more informed decisions, the architecture could also use machine-learning systems to identify targets and make weapon recommendations, thereby speeding up engagement times. U.S. forces are also leveraging AI in an effort to fuse disparate sensor inputs and create a common operational picture to more effectively identify and disrupt Chinese area denial systems, providing a safer environment for U.S. forces to operate. At present, however, Joint All-Domain Command and Control is still just a concept. There is a huge gulf between the seamless integration senior leaders envision in concept, and the reality that many operators struggle to log into their isolated computer networks each day.
The U.S. military is also trying to use AI to improve year-round force readiness, especially through predictive maintenance and logistics. Readiness will be especially important given that a potential U.S.-Chinese clash could happen with little warning. When the Department of Defense launched the Joint Artificial Intelligence Center in 2018, its first goal was to deliver AI-enabled predictive maintenance systems to forecast equipment breakdowns before they occur. AI-based maintenance could offer a significant improvement on the military’s current preventive maintenance system, which directs servicemembers to schedule and conduct maintenance at routine intervals. As of 2022, predictive maintenance software has already been deployed in support of the F-35 Lighting II. By using AI to monitor subsystem health and predict component failures on tighter intervals, U.S. defense planners aim to ensure that the greatest number of fighters are operational in the event of a U.S.-Chinese conflict. As this technology matures, preventive maintenance systems could be deployed across the joint force, increasing readiness for surface vessels and other platforms.
If a U.S.-Chinese crisis unfolds in the Indo-Pacific (most likely in or around the East or South China Seas), U.S. logistics forces would be asked to undertake a herculean feat: transporting supplies en masse halfway across the world. The challenge would become even more daunting in a protracted conflict. This is precisely why U.S. logistics forces have endeavored to combine cloud services with AI, to make more informed decisions about when to move supplies. The U.S. Army has already contracted IBM to provide cloud services and access to Watson — a question-answering AI computer — to store and process logistics data. By expanding intelligent logistics beyond the Army, U.S. forces hope to more efficiently coordinate plans to deliver the personnel, equipment, and supplies they need for successful operations. However, cloud storage also has the potential to introduce new vulnerabilities. In an operation known as “Cloud Hopper,” for example, Chinese hackers from the Ministry of State Security penetrated the cloud services of IBM, among other U.S. defense contractors. Data security will be the biggest impediment to the U.S. military’s intelligent logistics efforts.
Intelligent aerial vehicles could also benefit U.S. forces. Specifically, U.S. defense planners hope to use smart drones to support manned fighters and supplement nearby-based U.S. air forces that would likely be outnumbered by Chinese counterparts. Companies such as Boeing and Kratos have already developed autonomous drone prototypes that act as escorts or “loyal wingmen” for platforms like the F-35 and F/A-18. In addition to autonomous navigation, these drones may use AI for independent combat missions, surveillance, and cognitive electronic warfare. Within the next five years, it is possible that autonomous platforms will be able to analyze surveillance data, identify and respond to enemy aircraft’s electronic-warfare threats, and conduct battle damage assessments. These abilities could be a boon to joint forces that could access and further process collected data as part of Joint All-Domain Command and Control.
Beyond logistics, AI has a role to play in long-range fires. In a fight with the Chinese military, U.S. forces would be severely constrained by a demanding “shot doctrine” — the need for a high volume of missile launches to guarantee destruction — when targeting Chinese surface vessels and land-based defense systems. In principle, AI could help U.S. forces conserve munitions by providing adaptive targeting in response to battle-damage assessment, so the U.S. military doesn’t waste munitions on targets that have already been taken out. Specifically, the United States could use intelligent standoff weapons to increase the effectiveness of initial strike operations. Long range air-to-ground missiles could also allow U.S. forces to engage targets from a distance far enough to evade defensive fire. Recently, the U.S. military has attempted to incorporate AI with standoff missiles and other munitions. For example, the U.S. Air Force’s “Golden Horde” project seeks to use intelligent battle-damage assessment and munition communication to eliminate wasted missiles. If a target is assigned a multi-missile salvo but the first missile destroys the target, other missiles from the salvo will autonomously identify the damage, consider alternate targets, and navigate towards any within reach. U.S. defense planners hope that this would prevent expensive waste and dramatically cut down the time needed to destroy critical targets. Such capabilities could benefit initial U.S. strikes against People’s Liberation Army infrastructure and area denial weapons — however, the technology required for intelligent standoff missiles is likely still a few years away at best.
Takeaways and Recommendations
In a potential conflict, AI would offer distinct benefits for both Chinese and American forces. While People’s Liberation Army capabilities remain inferior in many respects, Chinese military leaders are investing in AI to offset U.S. military advantages. In his Center for a New American Security report on China’s AI strategy, Gregory Allen points out that China’s government sees AI as a promising military “leapfrog” development, meaning that it offers advantages to the People’s Liberation Army despite its lagging behind on development of other technologies. Rather than psychological or cognitive operations, as several scholars have suggested, our analysis of Chinese military investments suggests that AI’s most immediate and profound effects will come from intelligent munitions, as well as from maintenance and logistics systems that are already under development. Accordingly, we recommend two lines of effort to increase the likelihood of U.S. victory in a near-term U.S.-Chinese war: expanding investment in counter-AI research and adopting zero-trust architectures for the development of U.S. AI systems.
First, the United States must be prepared to degrade and counter the People’s Liberation Army’s evolving suite of AI systems. Department of Defense planners should continue to regulate Chinese access to advanced equipment, data, and capital to hinder the availability and utility of AI systems. As one of us found last year, very few of the People’s Liberation Army’s AI equipment suppliers are named in U.S. end-user export control lists. More of them should be. Gaps in the U.S. export control framework, combined with a lack of situational awareness, continue to allow the Chinese military to access U.S. technology and capital in pursuit of AI.
Simultaneously, the United States can impose costs on AI-reliant Chinese forces by embracing advances in the fields of counter-AI and counter-autonomy. In responding to China’s growing AI power, it is important for U.S. leaders to avoid a strategy that is reactive, defensively oriented, and which might become yet another area where the United States is on the wrong side of the cost curve. Counter-autonomy could help to avoid that outcome. Specifically, the Department of Defense should invest more in adversarial machine learning techniques — finding and exploiting weaknesses in Chinese AI models by feeding them specific data inputs. In a 2020 white paper, the Defense Science Board recommended that the department use counter-autonomy “to defend against increasingly autonomous systems deployed by adversaries, and to ensure that U.S. autonomous systems are not vulnerable to adversary countermeasures.” Despite the recommendation, the Department of Defense has not publicly created a senior counter-autonomy leadership position, nor has it invested in related research projects or sought to acquire systems designed specifically to understand and defeat Chinese AI platforms. But doing so would be crucial to ensure U.S. success in a near-term crisis.
Second, the United States should continue to invest in its own AI capabilities to remain competitive with the People’s Liberation Army. Chinese military leaders have long recognized their forces’ deficiencies in conducting joint operations — but U.S. forces, too, have a long way to go before they can effectively and reliably operate joint command and control. By using AI to intelligently incorporate sensors from across different services, the United States is more likely to keep its edge when conducting joint operations. The Department of Defense should also invest further in intelligent munitions. It has long relied on large salvos, which quickly deplete the U.S. missile inventory in a variety of combat scenarios. Successfully fielding intelligent standoff weapons to curtail this deficiency would be invaluable during a potential battle with the Chinese military.
Finally, the U.S. military must limit its own network vulnerabilities as it develops AI systems and updates its command-and-control system. This should include adopting zero-trust architectures for cybersecurity, which continuously validate every step of the network development process. Doing so will help thwart China’s plans to use machine learning to identify and attack vulnerabilities to critical U.S. networks — for example, through fuzzing. Furthermore, the Department of Defense should continue prioritizing rigorous test and evaluation, verification, and validation of AI systems department-wide. There is no room to cut corners in the name of speedy deployment. As the Institute for Defense Analyses concluded in 2018, “rare but catastrophic failures are harder to avoid in this context than they are in commercial settings.”
It is a tall order to predict if or how the United States and China may find themselves embroiled in a conventional war. However, the United States needs to be prepared to leverage AI for such a contingency. Failing to do so risks ceding the advantage to Chinese military planners. While the contest for “AI dominance” may only marginally affect the outcome of a near-term U.S.-Chinese conflict, that battle is America’s to lose.
Alex Stephenson is a China military technology research assistant at Georgetown University’s Center for Security and Emerging Technology and former surface warfare officer in the U.S. Navy.
Ryan Fedasiuk is a research analyst at Georgetown University’s Center for Security and Emerging Technology and an adjunct fellow at the Center for a New American Security.
The views expressed here are those of the authors and not those of the U.S. Navy, the U.S. Department of Defense, or any part of the U.S. government.
Image: China Military Online