As China and the United States enter a new era, in which rivalry and confrontation are starting to extend across all dimensions of this bilateral relationship, artificial intelligence (AI) has emerged as a new frontier of strategic competition. Evidently, nations worldwide recognize the strategic significance of AI technologies to the future of economic development and military modernization. As such, it is hardly surprising that the United States and China have prioritized AI, though taking quite different approaches to policy. In the process, reactions to and perceptions of each other’s advances and ambitions have been a key influence upon these efforts. For instance, the Chinese government’s decision to devote strong state support to AI development was catalyzed to some extent by concerns over the perceived superiority of ‘Western’ AI, as demonstrated by AlphaGo’s mastery of the game of Go in the spring of 2016. The initial steps taken towards a U.S. national strategy on AI around that time also appear to have heightened Beijing’s concerns about the risks of falling behind in these technologies.1 China has since formulated and actively implemented the New Generation Artificial Intelligence Development Plan (新一代人工智能发展规划), released in July 2017, which articulated the objective for China to “lead the world” in AI by 2030.2 Increasingly, U.S. concerns over China’s emergence as an AI powerhouse—and would-be “AI superpower”3—have influenced calls for an American strategy for AI, towards which critical progress has occurred with the launch of the American AI initiative in February 2019.4 Increasingly, AI seems to be front and center in today’s great power competition.5

This tendency towards U.S.-China AI rivalry is not inherently problematic, insofar as competition can act as an impetus to spur advances in technologies that can create significant advantages and opportunities to enhance human well-being. Moreover, competition in this context is not purely bilateral, nor is it the defining feature of this relationship, given the opportunities for continued cooperation between the U.S. and China. Thus far, some of the leading players in AI have been tech companies with global presence and interests. Chinese and American companies and universities have often engaged and collaborated, including new partnerships announced in 2018 between MIT and Chinese AI start-ups iFlytek and Sensetime.6 Beyond the United States and China, a range of nations worldwide, from Russia to Japan and the European Union, are launching their own AI policies and strategies.7 In some respects, smaller states and non-state actors might even have an advantage in AI, if proving more agile and creative in taking advantage of its applications.8 However, against the backdrop of a severe, and seemingly worsening, U.S.-China security dilemma, which is exacerbated by mistrust and misperception,9 there are risks that this tendency towards competition could exacerbate arms racing dynamics in the military applications of AI, perhaps resulting in troubling compromises on the safety and surety of these technologies. Indeed, as the United States and Chinese militaries seek new opportunities to achieve a disruptive, or even decisive, advantage, their apparent enthusiasm for the potential of AI in national security and defense might, at worst, result in new threats to strategic stability. This paper explores the impetus, objectives, and indicators for advances in China’s quest to advance an agenda for military “intelligentization” (智能化) in comparative perspective, while highlighting likely risks and potential challenges that may adversely impact future strategic stability in this complex, consequential relationship going forward.10

Chinese Military Innovation and “Intelligentization”

China’s reaction to the U.S. Third Offset Strategy has shaped the prioritization of military innovation by the People’s Liberation Army (PLA) under the leadership of Xi Jinping. In recent history, the PLA has often focused on ‘learning without fighting,’11 through a close study of U.S. ways of warfare and attempts to draw lessons from other militaries’ experiences, seeking to compensate for its own lack of recent combat experience.12 Unsurprisingly, PLA academics and strategists have closely tracked the Third Offset since its inception, particularly given the potential implications of this strategy for the future military balance.13 Indeed, U.S. defense leaders decided to undertake the Third Offset strategy in reaction to concerns that great power rivals, namely China and Russia, were starting to emerge as true near-peer competitors, including through their development of better battle networks, i.e., command, control, communications, computers, intelligence, surveillance, and reconnaissance (C4ISR) systems, along with what U.S. military planners have characterized as Anti-Access/Area Denial (A2/AD) capabilities.14 Consequently, the Department of Defense has sought new opportunities to regain advantage at the operational level of warfare, drawing upon the legacy of prior “offsets”—including the “first offset” of nuclear weapons and the “second offset,” which involved the introduction of stealth, precision strike, and battle networks.15 At the core of this strategy was a focus on the potential of emerging capabilities—including learning systems, human-machine collaboration and combat teaming, and network-enabled and cyber-hardened autonomous weapons.16 These U.S. concepts and initiatives have evidently intensified China’s own quest to advance new directions in its military modernization.

Xi Jinping has highlighted the imperative of an “innovation-driven” strategy for China’s development, while calling upon the PLA to advance military innovation.17 This strategic approach is motivated by an assessment of trends in the global Revolution in Military Affairs (RMA),18 characterized by advances in long-range, precise, smart, stealthy, and unmanned weapons, as highlighted in the defense white paper on “China’s Military Strategy.”19 The PLA’s commitment to military innovation includes not only technological advancement but also the formulation of new theories, concepts of operations, and organizational approaches.20 These ambitions must be recognized as the continuation of a decadal process of modernization, through which the PLA has sought to transform itself from a force that was once scarcely mechanized to one that is becoming truly “informatized” (信息化)—and ever more capable, in the words of Xi Jinping, of “fighting and winning” wars. Indeed, the “China Dream” of “national rejuvenation” includes the objective of a powerful military (强军目标) capable of defending China’s core and national interests, which are expanding globally in their scope and scale.21 Presently, the PLA is undertaking historic reforms and reorganization that could increase its capabilities for joint operations. The PLA is also concentrating on the development of “new-type forces” and “new concept” weapons, from the construction of the Strategic Support Force, which has integrated and consolidated space, cyber, electronic, and psychological warfare capabilities,22 to the advancement of hypersonic weapons systems.23

Xi Jinping’s call to “Accelerate the development of military intelligentization” in his remarks during the 19th Party Congress in October 2017, has placed official imprimatur upon the leveraging of AI to enhance future military power.24 This concept, which extends and builds upon the PLA’s prior strategy of informatization, could emerge as a new principle guiding the future trajectory of Chinese military modernization, which is increasingly advancing a range of applications of AI and related technologies. As characterized by a number of Chinese defense academics and strategists, this concept does not merely imply the militarization of AI,25 but also seems to involve and require related advances in theories and training for future operations, as well as the related materials and supporting infrastructure.26 According to a PLA scholar from the National University of Defense Technology, “military intelligentization refers to the overall operational description of the forces system of systems consisting of people, weapons equipment, and ways of combat.” In practice, such a system of systems would involve not only intelligent weaponry but also concepts of human-machine integration (人机一体) and intelligence leading (智能主导).27 In practice, the PLA’s agenda for intelligentization may prove quite expansive, extending across all concepts in which AI might have military relevance in enabling and enhancing war-fighting capabilities, from logistics to early warning and intelligence, military wargaming, and command decision-making.28

The PLA’s traditional emphasis on asymmetric capabilities, often referred to as “assassin’s mace” or “trump card” weapons (杀手锏), could influence its approach to these emerging technologies. For instance, the Chinese military and defense industry have concentrated on the development of swarms of drones and even of ships, recognizing their potential to overwhelm the defenses of an adversary’s valuable weapons platforms.29 At present, a majority of these capabilities remains at various stages in research, development, and experimentation, and it is difficult to evaluate their sophistication based on appearance or announcements alone.30 Looking forward, Lieutenant General Liu Guozhi, director of the Central Military Commission’s Science and Technology Commission, has emphasized the imperative of advancing military intelligentization, arguing, “this is a rare strategic opportunity for our nation to achieve innovation leapfrogging.”31 At present, however, many of the technologies in question remain relatively immature, pending future testing and verification. In this regard, the realization of this agenda for military innovation will play out for years to come. It is too soon to evaluate with confidence whether the PLA will prove successful in achieving these aspirations to surpass or leapfrog ahead of the U.S. military.

Initial Progress and Challenges

The PLA is actively supporting a range of projects and research activities involving military applications of artificial intelligence and related technologies. Within the PLA, the Central Military Commission (CMC) Science and Technology Commission (S&TC) is taking the lead in advancing military-technological innovation, including through a plan focused on “frontier innovation” (前沿创新). Concurrently, the CMC Equipment Development Department (EDD) has created an AI Expert Group that may guide and direct the advancement of AI-enabled armaments. There are a variety of funding mechanisms through which the PLA is supporting dual-use and military research across the defense industry, technology companies, and universities. For instance, a fund focused on naval innovation has funded the development of AI-enabled image processing and target recognition technologies, including to detect ships in satellite imagery, while the EDD has also supported research on the use of machine learning, and specifically deep learning, for the analysis of underwater acoustic signals.32 It is worth noting that research on these and other applications of AI often reflects a history that extends beyond the recent emphasis on and enthusiasm for these technologies, even dating back to the 863 Plan that supported early AI-related research from the late 1980s onward.33 Gradually, a more robust and diverse innovation ecosystem is emerging across PLA research institutes, the Chinese defense industry, and a growing number of private companies, from smaller start-ups, such as Yunzhou-Tech, to ‘national champions’ like Baidu, which is clearly at the forefront of China’s AI revolution.34

In the near future, certain of the promising applications of these technologies will involve enabling capabilities that reflect more ‘qualitative’ augmentation of existing systems, often in ways that may be difficult to discern readily from observable indicators. For instance, there has also been apparent progress in research involving the AI-enabled mining and processing of data from remote sensing satellites, which may already be in use, at least to a limited degree, by the PLA Rocket Force and Strategic Support Force.35 Recognizing the criticality of data fusion and intelligence to future operational advantage, PLA researchers are also exploring options to leverage AI technologies to train on and learn from target features acquired by a range of sensors.36 For instance, the application of machine learning to improve automatic target recognition (ATR), including with the use of convolutional neural networks in Synthetic Aperture Radar (SAR) image target recognition, is recognized as an opportunity to build upon prior research to enhance the precision in targeting.37,38 As virtual domains emerge as critical frontiers of military competition, the PLA’s interest in “cognitive electronic warfare,” influenced in part by concern over U.S. efforts, reflects a recognition of the potential impact of applications of machine learning to enable more adaptive countermeasures to seize dominance in the electromagnetic spectrum.39 So too, given the threats of such countermeasures, Chinese researchers are also focusing on ways in which AI can bolster the resilience of future communications networks against an adversary’s attempted jamming and interference.40 Considering new threats and confrontation in cyberspace, the potential of AI to enhance cyber security, defense, and perhaps also offensive operations may also emerge as a major priority for China under the PLA Strategic Support Force.

Anticipating a trend towards future “unmanned” warfare, the PLA is rapidly advancing the development of new capabilities for greater ‘intelligence’ and autonomy across all domains.41 The PLA Navy is deploying and experimenting with a range of intelligent/autonomous surface vessels and underwater vehicles.42 Notably, the Haiyi (海翼) or “Sea Wing,” an underwater glider designed by the Chinese Academy of Sciences (CAS) Shenyang Institute for Automation, used so far for scientific missions in the South China Sea,43 has also been highlighted as having potential military utility, including due to its low acoustic signature.44 The Haiyi could be leveraged to enable the detection of foreign submarines, thus potentially enhancing to enhance PLA anti-submarine warfare capabilities.45 In addition, the Jinghai (精海), an ‘intelligent’ vessel with the reported capability to navigate autonomously, appears to be in use with the PLAN and might support maritime sensing and domain awareness.46 As Chinese defense industry players, such as the China Shipbuilding Industry Corporation (CSIC), and even private sector contenders, such as Yunzhou-Tech, which is recognized for its significant contributions to military-civil fusion, develop a wide array of models and designs,47 China has also established the world’s largest facility for the testing of such vessels, including those optimized for combat missions, at the Wanshan Marine Test Site in Zhuhai.48 The Chinese defense industry is also developing a range of USVs that could be employed for combat in support of the PLAN.49 Reportedly, the PLA is also developing AI-enabled submarines to advance Chinese capabilities in undersea warfare, through a classified military program known as the 912 Project, which was disclosed in English-language reporting, perhaps a deliberate signaling of potential future capabilities.50 Although fully autonomous submarines may remain a long-term objective, the introduction of AI technologies for decision support on submarines could prove more feasible in the meantime.51

The PLA’s apparent enthusiasm for the disruptive potential of swarming capabilities has been manifest in multiple demonstrations featured prominently in official media. Notably, the China Electronics Technology Group (CETC), a leading state-owned defense conglomerate, has tested swarms of 67, 119, and then 200 fixed-wing UAVs, which have engaged demonstrated complex formations, setting this latest record in May 2018.52 (Ehang, a UAV company, has even demonstrated a swarm of 1,374 drones in total, breaking the Guinness World Record.53) The PLA’s National University of Defense Technology (NUDT) has also developed and demonstrated swarms of UAVs through its Academy of Intelligence Sciences.54 Meanwhile, the PLA Air Force has organized a swarm challenge for the development of swarms with greater degrees of autonomy, capable of collaboration and coordination.55 These intelligent unmanned systems could serve as an asymmetric means through which to target high-value U.S. weapons platforms, including fighter jets or aircraft carriers. Of note, China’s Military Museum includes in an exhibit on future warfare a depiction of a UAV swarm combat system (无人机蜂群作战系统) with swarms used for reconnaissance, jamming, and “swarm assault” (群打击) targeting an aircraft carrier, in conjunction with manned aircraft.56 Chinese official reporting has also highlighted the successful testing of a ‘shark swarm’ of drone vessels, developed by Yunzhou-Tech, that could be acquired by the PLA Navy for future combat scenarios.57 While swarms today range from dozens to hundreds in size typically, continued advances in commercial technologies, including cheap drones and more advanced techniques for additive manufacturing, could enable the scaling up of such swarming capabilities, perhaps even to include thousands of drones.58 Of course, size alone is not a reliable indicator of capability in this context, and the true level of autonomy and sophistication of these swarms is difficult to determine by observation alone.

Looking forward, the PLA anticipates that the advent of AI in warfare could augment future command decision-making, potentially enabling new opportunities to achieve decision superiority relative to an adversary.59 In an authoritative commentary, the CMC Joint Staff Department has called for the PLA to accelerate its construction of a joint operations command system through progress toward intelligentized command decision-making that takes advantage of the potential of AI, as well as big data, cloud computing, and other advanced technologies.60 The Joint Staff Department’s commentary highlighted that the victory of Google’s AlphaGo in that “human-machine war” demonstrated the tremendous potential of AI in operational command, military war-gaming, and support to decision-making.61 The introduction of AI to augment a commander’s capabilities is expected to help compensate for human shortcomings and to enable speed and superiority in future decision-making.62 According to PLA strategists, this introduction of AI into the realm of command decision-making may prove inevitable, and future warfare will involve the combination of intelligentized combat and command platforms,63 with “intelligence dominance” (制智权) emerging as the core mechanism for victory.64 In the foreseeable future, the PLA might leverage AI technologies to enhance the decision-making of fighter pilots or the commanders of submarines. According to credible reporting, there is a project underway to update the computer systems on PLAN nuclear submarines with an AI decision-support system in order reduce commanding officers’ workload and mental burden.65,66 Through these efforts, AI may take on certain “thinking” functions on nuclear submarines, which could include, at a basic level, interpreting and answering signals picked up by sonar, such as through the use of convolutional neural networks.67 As the PLA looks to promote innovation in ‘actual combat’ (实战) training, the use of AI to enhance the realism and sophistication of training methods may also emerge as a priority.68

Pursuant to national strategy of military-civil fusion (军民融合), China seeks to create and leverage critical synergies among academic, defense, and commercial developments in AI technologies. Increasingly, a range of incentives and partnerships have been introduced to advance this strategy, influenced not only by a long history in China’s thinking on mobilizing resources for technological development but also by a close study of the successes of American defense innovation. While this agenda is increasingly expansive in scope and scale, a key metric of success will be China’s capability to mobilize leading companies and universities in support of this agenda, taking advantage of their capabilities to promote dual-use technological developments. For instance, Baidu and CETC’s 28th Research Institute have established the Joint Laboratory for Intelligent Command and Control Technologies, which will focus on increasing the level of ‘intelligentization’ in command information systems through incorporating big data, artificial intelligence, and cloud computing.69 Notably, Tsinghua University, often characterized as ‘China’s MIT,’ is strongly and institutionally committed to military-civil fusion and to supporting the advancement of Chinese military applications of AI. Tsinghua Vice President You Zheng has highlighted research underway on project on future human-machine cooperative (combat) operations.70 Tsinghua is also building the “High-End Laboratory for Military (Artificial) Intelligence” with support from the PLA’s Central Military Commission,71 along with the Military-Civil Fusion National Defense Peak Technologies Laboratory, which will create a platform for the pursuit of dual-use applications of emerging technologies.72 In the latest meeting of a commission on the topic, Xi Jinping called for the PLA to undertake more competitive approaches to procurement.73

Of course, the PLA’s capability to operationalize this agenda of military intelligentization will be shaped and may be constrained by challenges of talent recruitment, new directions in training, and organizational adaptation.74 That is, the development and deployment of AI require the recruitment of researchers and more proficient personnel into the Chinese defense industry and military. For instance, CETC, which has launched its own plan dedicated to AI development, has recruited for researchers with experience in artificial intelligence and machine learning.75 The PLA’s current reforms to its system for civilian personnel, which could facilitate the hiring of civilian scientists into military positions with benefits on par with the civil service, may contribute to its future human capital ecosystem. For instance, the PLA Strategic Support Force is currently recruiting a number of researchers with a background in AI, including for positions in “aerospace artificial intelligence.”76 Of course, the ability to attract capable candidates will be inherently challenging, given the apparent shortfall in AI talent in China at present and competition with a dynamic tech sector. However, PLA universities are expanding their own educational opportunities in AI. For instance, the National University of Defense Technology (NUDT) has launched the Academy of Intelligent Sciences and evidently constitutes a key center of excellence for AI education and development within the PLA.77 As AI is introduced into Chinese universities as a first-level discipline, and as the Ministry of Education implements a new plan to expand the offering of AI as a major and AI research institutes, this pipeline may further expand.78

As the PLA introduces more complex automated, autonomous, and otherwise ‘intelligentized’ technologies, the challenges for training could increase, including due to requirements for greater technical proficiency. However, there may also be some cases in which the PLA’s current lack of experience and apparent shortcomings may motivate a more rapid adoption of ‘unmanned’ technologies. Notably, the CH-7 stealth UAV, revealed at the Zhuhai Airshow in the fall of 2018, has provoked speculation that the PLA Navy may be the first to introduce unmanned carrier aviation.79 Such a potential leapfrogging that could reflect a response to and recognition of its lack of experience in manned carrier aviation.80 The PLA’s apparent enthusiasm for unmanned technologies, which has been less visibly hindered by a preference for manned aviation that other militaries seem to display, is reflected in portrayals of PLA UAV pilots, who are often lionized in PLA media.81 Although there has been an expectation that authoritarian regimes may neglect the essential human element in future warfare, the PLA academics and strategists have, in fact, emphasized the importance of human intelligence and called for advances in intelligentized training in order to prepare personnel for the demands of future intelligentized operations,82 including to improve human-machine cooperative (人机协同) decision-making capabilities.83 However, certain PLA thinkers also anticipate that warfare may evolve and training must adapt from humans “in the loop” to “on the loop” to “outside the loop,” guided by a focus on “human-machine interactive thinking” (智能化人机互动思维).84 Although human-machine cooperation may be the trend in intelligentized operations in the near future, some PLA thinkers expect higher levels of autonomy will be an inevitable feature of the future battlefield.85 Consequently, ‘algorithmic advantage’ is expected to become a decisive factor for future victory.86

New Risks and Security Dilemmas

Are the United States and China engaged in an “AI arms race” at present? It is clear that the U.S. and Chinese militaries share a strong interest and are undertaking increasing investments in the military applications of these technologies. Although attempts to predict and anticipate the future of warfare are imperfect at best, there is consensus that artificial intelligence and autonomy will have a range of impactful applications that could, in the aggregate, constitute a new AI-enabled “Revolution in Military Affairs” (‘AI-RMA’), potentially in conjunction with other emerging technologies, ranging from hypersonics to biotechnology and directed energy weapons. If that does indeed prove to be the case, then whichever military first masters these technologies—and the powerful convergences among them—could achieve dominance in a new generation of warfare. In this regard, concerns over an “arms race” in AI cannot be disentangled from the extant dynamics of intense military rivalry among great powers. Although multiple militaries around the world do appear to anticipate that today’s advances in AI could prove transformative, perhaps even revolutionary,87 the inherent uncertainties and challenges in the trajectory of these technologies’ development and operationalization should not be discounted amidst the hype. Moreover, the notion of an “arms race” in AI is inadequate and incomplete as a concept at best, in light of the complexity of these technologies and the multiplicity of their applications. However, today’s advances in AI technology could exacerbate current security dilemmas among great powers, particularly considering that these technologies often contribute to qualitative enhancements in the capabilities of potential adversaries in ways that are difficult to evaluate through observation alone. The resulting uncertainties about the military balance could contribute to overreaction to and overestimation of competitors’ advances, spurring further investments and increasing competitions.

Indeed, rapid advances in AI could disrupt and destabilize the existing global balance of power, though not necessarily for the reasons that have captured the popular imagination.88 There seems to be less danger that “killer robots” or “superintelligence” could be created as a result of such competition and more reasons for concern that advances in AI could present new risks to strategic stability, in conjunction with parallel technological developments. If military competition continues to intensify, there are reasons for concern that a “new round of arms racing” could indeed emerge.89 Nonetheless, there are also ways in which the introduction of AI, including to enhance Chinese early warning capabilities and situational awareness, could prove stabilizing, potentially increasing China’s confidence in its capability for rapid response to a nuclear attack.90 However, Chinese defense academics and strategists, including Professor Zhu Qichao of the PLA’s National University of Defense Technology (NUDT), have also highlighted the acute risks that the advent of AI in warfare might present, given the greater speed, precision, and potential for errors.91 Indeed, AI is recognized as a “double-edged sword” that could introduce new risks to national security, military security, and societal security, as a recent Chinese white paper on AI safety/security highlighted.92 The PLA’s pursuit of AI-enabled capabilities could exacerbate issues of failures in complex systems, adversary attempts to attack or otherwise undermine autonomous systems, or unanticipated interactions between adversarial systems.93 For instance, the PLA is also likely to look to AI to enhance its offensive and defensive cyber capabilities, including under the aegis of the Strategic Support Force. The trend towards integrating AI with cyber capabilities to achieve an advantage could intensify escalatory dynamics in this contested domain.

Despite fears that advances in AI are “summoning the demon,”94 the present limitations of AI, rather than its potential power or questions of controllability, could become more problematic in the near term for China and the United States alike. In some cases, errors and issues might arise in more routine applications, even when the algorithm in question is not used directly in a weapons system or to make a life-and-death decision. Even military applications of AI in intelligence and to augment command decision-making could raise risks of mistakes that might contribute to crisis instability or exacerbate the risks of escalation, potentially creating new avenues for misperception or miscalculation in leadership decision-making. At this stage of development, AI remains far from intelligent, tending to make mistakes no human would make. Such errors can be unpredictable or difficult to mitigate. In certain cases, the results can be amusing or nonsensical, as in the case of adversarial examples.95 In a military context, however, there could be adverse consequences and new operational risks.96 There will be a higher likelihood of errors or unexpected emergent behavior as the level of complexity increases and if a situation exceeds the expected parameters, such as in the transition from virtual training to real-world environments. Consequently, major militaries should take a proactive approach to evaluating and mitigating the potential risks introduced by advances in military applications of AI. It is and will remain in their interest to do so, since the United States, China, and Russia still share at least a basic commitment to strategic stability and recognize the undesirability of inadvertent escalation.

Going forward, the United States and China must also recognize the new threats resulting from the likely proliferation of AI-enabled capabilities. Thus far, developments have been propelled forward primarily by commercial technologies, and the companies at the forefront of the field have taken a very open approach to releasing their results, including open-sourcing some of the key tools, algorithms, and infrastructure. Given that the barriers to entry have decreased as a result, there may be more opportunities for non-state actors, including criminals and perhaps even terrorist organizations, to explore options to leverage these technologies for malicious purposes.97 This trend may converge with that of the cheapness and availability of commercial drones, which will only increase with advances in additive manufacturing. Even terrorist organizations, notably ISIS, are starting to employ drones for battlefield reconnaissance and to convey explosives.98 The successes of Chinese company DJI, which commands the majority of the world’s market for commercial drones,99 has only rendered these capabilities more and more readily accessible.100 Concurrently, while AI technologies become ever more pervasive in our societies and economies, new threat vectors may arise, from the potential weaponization of self-driving cars to the manipulation of smart devices. While these “AI superpowers” may reap the greatest benefits from advances in these technologies, it is undeniable that many players and potential malicious actors could develop the capability to exploit them. As American and Chinese policymakers turn their attention to questions of AI safety and security, the shared nature of these threats should be an impetus to explore opportunities for cooperation on risk mitigation.


Elsa B. Kania is an adjunct senior fellow with the Technology and National Security Program at the Center for a New American Security (CNAS) and a PhD student at Harvard University.


1 National Science and Technology Council, “The National Artificial Intelligence Research and Development Strategic Plan,” Executive Office of the President of the United States, Washington, D.C., October 2016,
2 “State Council Notice on the Issuance of the New Generation AI Development Plan” [国务院关于印发新一代人工智能发展规划的通知], July 20, 2017,
3 On the notion of the U.S. and China as “AI superpowers,” see: Kai-Fu Lee, AI Superpowers: China, Silicon Valley, and the New World Order, Houghton Mifflin Harcourt, 2018.
4 See: “Executive Order on Maintaining American Leadership in Artificial Intelligence,” February 11, 2019,
5 Michael Horowitz, Elsa B. Kania, Gregory C. Allen, and Paul Scharre, “Strategic Competition in an Era of Artificial Intelligence,” Center for a New American Security, July 25, 2018,
6 Given that the companies in question contribute to the Chinese government’s surveillance capabilities, the ethics of these partnerships may merit further debate and consideration.
“MIT and SenseTime announce effort to advance artificial intelligence research,” MIT News, February 28, 2018,
“CSAIL launches new five-year collaboration with iFlyTek,” MIT News, June 15, 2018,
7 For a quick overview of various policies and strategies around the world, see, for instance: Tim Dutton, “Artificial Intelligence Strategies,”
8 For a great discussion of the potential success of smaller states in this technology in the course of its diffusion, see: Michael Horowitz, “The Algorithms of August,” Foreign Policy, September 12, 2018,
9 For a discussion of how ‘memes’ about China can achieve great traction, see: Alastair Iain Johnston, “How new and assertive is China’s new assertiveness?” International Security 37, no. 4 (2013): 7-48.
For an earlier discussion of the security dilemma, see: Thomas J. Christensen, “China, the US-Japan alliance, and the security dilemma in East Asia,” International Security 23, no. 4 (1999): 49-80.
For the classic work on the security dilemma, see: Robert Jervis, “Cooperation under the security dilemma,” World Politics 30, no. 2 (1978): 167-214.
10 For the author’s prior reflection on the issues, see: Elsa B. Kania, “Great Power Competition and the AI Revolution: A Range of Risks to Military and Strategic Stability,” Lawfare, September 19, 2017,
11 This phrasing reflects the author’s characterization of the PLA’s approach, and it is also the title of a future project, including papers that will be forthcoming in early or mid-2019.
12 For an earlier assessment of this tendency, see: Andrew Scobell, David Lai, and Roy Kamphausen, Chinese Lessons From Other People’s Wars, Strategic Studies Institute Book, November 2011,
13 For one of a number of articles on the topic, see: Wang Peng [王鹏], Shao Dan [邵丹], “Whom the New U.S. “Offset Strategy” Seeks to Offset” [美国新“抵消战略”欲抵消谁], China Youth Daily [中国青年报], January 15, 2016,
14 For prior discussions of the challenge of anti-access/area denial, see: Andrew F. Krepinevich,, Barry D. Watts, and Robert O. Work. Meeting the Anti-Access and Area Denial Challenge. Washington, DC: Center for Strategic and Budgetary Assessments, 2003.
15 For an excellent assessment of these issues, see: Robert O. Work and Shawn Brimley, “Preparing for War in the Robotic Age,” Center for a New American Security, January 22, 2014,
16 For authoritative discussion of and commentary upon these issues, see Bob Work’s remarks upon several occasions, including: Bob Work, “The Third U.S. Offset Strategy and its Implications for Partners and Allies,” Deputy Secretary of Defense Speech, Willard Hotel, Washington, D.C., January 28, 2015,
Cheryl Pellerin, “Work: Human-Machine Teaming Represents Defense Technology Future,” DoD News, November 8, 2015,
17 See the official strategy released on innovation-driven development, “The CCP Central Commission and State Council Release the “National Innovation-Driven Development Strategy Outline” [中共中央 国务院印发《国家创新驱动发展战略纲要], Xinhua, May 19, 2016,
See also Xi Jinping’s remarks on this approach in the context of military modernization: “Xi Jinping: Comprehensively Advance an Innovation-Driven Development Strategy, Promote New Leaps in National Defense and Military Construction” [习近平:全面实施创新驱动发展战略 推动国防和军队建设实现新跨越], Xinhua, March 13, 2016,
18 For an early and reasonably authoritative discussion of these trends, see: Xiao Tianliang [肖天亮], “Seize the Reform Initiative in the Tide of Military Transformation” [顺应军事变革潮流把握改革主动], PLA Daily, January 5, 2016,
19 “Xi Jinping: Accurately Grasp the New Trend in Global Military Developments and Keep Pace with the Times, Strongly Advancing Military Innovation” [习近平:准确把握世界军事发展新趋势 与时俱进大力推进军事创新], Xinhua, August 30, 2014,
Academy of Military Science Military Strategy Research Department [军事科学院军事战略研究部], eds., The Science of Military Strategy [战略学], Military Science Press [军事科学出版社], 2013, 97-98.
State Council Information Office [中华人民共和国国务院新闻办公室], “China’s Military Strategy” (full text) [中国的军事战略(全文)], May 26, 2015,
20 See, for instance: “Scientific and Technological Innovation, a Powerful Engine for a World-Class Military” [科技创新,迈向世界一流军队的强大引擎],  Xinhua, September 14, 2017,
21 Zhou Junjie [周俊杰],“Profoundly grasp the rich connotation of the Party’s strong military objectives in the new era” [深刻把握党在新时代的强军目标的丰富内涵], PLA Daily, February 23, 2018,
22 Elsa Kania and John Costello, “The Strategic Support Force and the Future of Chinese Information Operations,” Cyber Defense Review, Spring 2018,
23 For context, see: “Prepared Statement of Mark A. Stokes Executive Director Project 2049 an
Advanced Weapons Development, February 23, 2017,
See: Lora Saalman, “China’s Integration of Neural Networks into Hypersonic Glide Vehicles,” in “AI, China, Russia, and the Global Order: Technological, Political, Global, and Creative Perspectives,” A Strategic Multilayer Assessment (SMA) Periodic Publication, December 2018,
24 “Xi Jinping’s Report at the Chinese Communist Party 19th National Congress” [习近平在中国共产党第十九次全国代表大会上的报告], Xinhua, October 27, 2017,
25 “Experts: Military Intelligentization Is Not Merely Artificial Intelligence” [ 专家:军事智能化绝不仅仅是人工智能], December 6, 2017,
26 Zhao Ming [赵明] “Take the Fast Train of Military Intelligentization Development” [搭上军事智能化发展的快车], PLA Daily, November 14, 2017,
27 “Experts: Military Intelligentization Is Not Merely Artificial Intelligence” [专家:军事智能化绝不仅仅是人工智能], December 6, 2017,
28 Military Science Editorial Department [中国军事科学 编辑部], “A Summary of the Workshop on the Game between AlphaGo and Lee Sedol and the Intelligentization of Military Command Decision-Making” [围棋人机大战与军事指挥决策智能化研讨会观点综述], China Military Science [中国军事科学], April 2, 2016. CMC Joint Staff Department [中央军委联合参谋部], “Accelerate the Construction of a Joint Operations Command System with Our Nation’s Characteristics—Thoroughly Study Chairman Xi’s Important Sayings When Inspecting the CMC Joint Operations Command Center [加快构建具有我军特色的联合作战指挥体系—— 深入学习贯彻习主席视察军委联指中心时的重要讲话], Seeking Truth, August 15, 2016,
29 For an account of China’s first world record in swarming, see: “Our Country Breaks World Records for the Number of Fixed-Wing UAVs Swarm Flying” [我国打破世界固定翼无人机集群飞行飞机数量纪录], China Military Online, November 6, 2016,
30 See, for instance: Andrea Gilli and Mauro Gilli, “Is China Ahead of America in Next-Generation Unmanned Aircraft?,” War on the Rocks, November 13, 2018,
31 “Experts: Military Intelligentization Is Not Merely Artificial Intelligence” [ 专家:军事智能化绝不仅仅是人工智能], December 6, 2017,
32 The sources in question are available upon request.
33 This is contemporaneous with the U.S. Strategic Computing Initiative and Japan’s Fifth Generation Computing.
34 “China Electronics Science and Technology Group and Baidu Company established the “Joint Laboratory for Intelligent Command and Control Technology” to promote military-civil fusion in the field of new technologies” [中国电科28所与百度公司成立“智能指挥控制技术联合实验室”推动军民融合向新技术领域纵深迈进], January 23, 2018,
35 “Remote Sensing Big Data Intelligent Processing and Mining Theories, Methodologies, and Major Applications” [遥感大数据智能处理与挖掘理论方法及重大应用] from the 2018 National Science and Technology Progress Award Nomination Content [2018年度国家科学技术进步奖提名公示内容]
36 See, for instance: “2017 Target Recognition and Artificial Intelligence Forum” [2017目标识别与人工智能高峰论坛], April 22, 2017,
37 For one of a number of publications on these topics, see: Tian Zhuangzhuang, Zhan Ronghui, Hu Jiemin, et al., “SAR ATR based on convolutional neural networks,” Journal of Radars, 2016, 5(3): 320–325. DOI: 10.12000/JR16037.
38 “2017 Target Recognition and Artificial Intelligence Forum” [2017目标识别与人工智能高峰论坛], April 22, 2017,
39 “Academician Wang Shafei: AI and Electromagnetic Spectrum Warfare” [王沙飞院士:人工智能与电磁频谱战], February 22, 2018,
40 “Anti-Jamming Academic Forum: Applications of Artificial Intelligence with Future Communication Networks & Machine Learning in Communications and Reconnaissance” [抗干扰学术论坛:人工智能与未来通信网络&机器学习在通信和侦察中的应用], November 9, 2017,
41 For a more extensive discussion of these issues, see: Elsa Kania, “Chinese Advances in Unmanned Systems and the Military Applications of Artificial Intelligence—the PLA’s Trajectory towards Unmanned, “Intelligentized” Warfare,” Testimony before the U.S.-China Economic and Security Review Commission, February 23, 2017,
Again, of course, the extent to which any given unmanned system is actually ‘intelligent’ or autonomous cannot be observed from appearance alone.
42 For another analysis of the capabilities under development, see: Lyle J. Goldstein, “Meet the HN-1, China’s New AI-Powered Underwater Drone,” The National Interest, July 15, 2018,
43 “Sea Wing Series of Underwater Gliders Achieves the Largest Model of Swarms Simultaneously Observing” [“海翼”系列水下滑翔机实现最大规模集群同步观测], Shenyang Institute of Automation, August 24, 2017,
44 See “Ordinance Industry Science and Technology, 2017, Issue 19” [《兵工科技》2016年第19期杂志], September 22, 2016,
45  Lyle Goldstein, “America May Soon Find Itself in an Underwater War with China,” The National Interest, July 24, 2017,
46 ““Jinghai Series” Unmanned Sensing Boat Debuts” [精海号”无人测量艇亮相], November 4, 2015,
47 Kelvin Wong, “IMDEX 2017: China’s Yunzhou-Tech showcases latest USVs,” IHS Jane’s International Defence Review, May 18, 2017,
48 “Zhuhai opens the world’s largest unmanned vessel offshore test site” [珠海开建世界最大无人船海上测试场], Xinhua, February 12, 2018,
For a more detailed English-language accounting of these plans, see: Jeffrey Lin and P.W. Singer, “China is building the world’s largest facility for robot ship research,” Eastern Arsenal, February 20, 2018,
49 For instance, see this recent model: “AAD 2018: China’s CSOC Unveils ‘JARI’ Unmanned Surface Combatant—USV,” September 23, 2018,
50 Stephen Chen, “China military develops robotic submarines to launch a new era of sea power,” South China Morning Post, July 22, 2018,
51 See my prior analysis on the topic: Elsa Kania, “Chinese Sub Commanders May Get AI Help for Decision-Making,” Defense One, February 12, 2018,
52 “China launches record-breaking drone swarm,” Xinhua, June 11, 2017,
“200 UAV Swarm Flight: China once again refreshed the fixed-wing UAV swarm flight record” [200架无人机集群飞行:我国再次刷新固定翼无人机集群飞行纪], May 15, 2018,
53 “Flight of imagination: Chinese firm breaks record with 1,374 dancing drones,” Reuters, May 2, 2018,
54 “NUDT Academy of Intelligent Sciences Experiments with UAV Swarm Autonomous Operations” [国防科大智能科学学院试验无人机集群自主作战], China Military Online, December 12, 2017,
55 “Advance Notice of the Air Force Equipment Department’s Holding the “Unmanned Struggle” Intelligent UAV Swarm System Challenge” [空军装备部关于举办“无人争锋”智能无人机集群系统挑战赛的预通知], Sina, September 29, 2017,
56 I observed this display during my August 2017 visit to the Military Museum in Beijing.
57 Again, the reporting upon these advances in English is significant, indicating that their success is intended to be made known to foreign audiences. See: “Unmanned ‘shark swarm’ to be used in sea battles, military patrols,” Global Times, June 6, 2018,
58 For an early and influential assessment of the future of swarming: Paul Scharre, “Robotics on the Battlefield Part II: The Coming Swarm,” Center for a New American Security, October 15, 2014,
59 Military Science Editorial Department [中国军事科学 编辑部], “A Summary of the Workshop on the Game between AlphaGo and Lee Sedol and the Intelligentization of Military Command Decision-Making” [围棋人机大战与军事指挥决策智能化研讨会观点综述], China Military Science [中国军事科学], April 2, 2016.
CMC Joint Staff Department [中央军委联合参谋部], “Accelerate the Construction of a Joint Operations Command System with Our Nation’s Characteristics—Thoroughly Study Chairman Xi’s Important Sayings When Inspecting the CMC Joint Operations Command Center [加快构建具有我军特色的联合作战指挥体系—— 深入学习贯彻习主席视察军委联指中心时的重要讲话], Seeking Truth, August 15, 2016,
60 CMC Joint Staff Department [中央军委联合参谋部], “Accelerate the Construction of a Joint Operations Command System with Our Nation’s Characteristics—Thoroughly Study Chairman Xi’s Important Sayings When Inspecting the CMC Joint Operations Command Center [加快构建具有我军特色的联合作战指挥体系—— 深入学习贯彻习主席视察军委联指中心时的重要讲话], Qiushi [求是], August 15, 2016,
61 Ibid.
62 Chen Yufei [陈玉飞] and Zhou Tao [周涛], “Will Artificial Intelligence Replace Commanders?” [人工智能能代替指挥员吗?], PLA Daily, June 8, 2017,
63 Guo Ruobing [郭若冰], Si Guangya [司光亚],  “Facing New Challenges to Military Command in the Era of Intelligentization” [接近智能化时代军事指挥面临的挑战], China Military Science, July 2016.
64 Wang Yunxian [王云宪] and Li Dawei [袁大伟], “Intelligentized Operations Require Intellientized Training” [智能化作战呼唤智能化训练], China Military Network, August 23, 2018,
65 For my prior analysis on the topic, see: Elsa Kania, “Chinese Sub Commanders May Get AI Help for Decision-Making,” Defense One, February 12, 2018,
66 For English-language reporting on the topic, see: Stephen Chen, “China’s plan to use artificial intelligence to boost the thinking skills of nuclear submarine commanders,” South China Morning Post, February 4, 2018,
67 Ibid.
68 “Advance Innovation in the Development of New-Era Military Training” [推进新时代军事训练创新发展], China Military Network, October 30, 2018,
69 “China Electronics Science and Technology Corporation and Baidu Company established the “Joint Laboratory for Intelligent Command and Control Technology” [中国电科28所与百度公司成立“智能指挥控制技术联合实验室”推动军民融合], January 23, 2018,
70 “The Ministry of Education Convened a Press Conference to Interpret the “Artificial Intelligence Innovation Action Plan for Colleges and Universities,” etc. [教育部举行新闻发布会解读《高等学校人工智能创新行动计划》等], Ministry of Education Website, June 8, 2018,
71 Ibid.
72 ‘Tsinghua Starts on the Establishment of the Military–Civil Fusion National Defense Peak Technologies Laboratory’ [清华启动筹建军民融合国防尖端技术实验室], China Education Report, 26 June 2017,
73 “Xi urges deeper military-civilian integration,” Xinhua, October 16, 2018,
74 Michael C. Horowitz, The Diffusion of Military Power: Causes and consequences for international politics, Princeton University Press, 2010.
75 See, for instance, this recruitment notice:
76 The data in question is available upon request.
77 “NUDT Academy of Intelligent Sciences Experiments with UAV Swarm Autonomous Operations” [国防科大智能科学学院试验无人机集群自主作战], China Military Online, December 12, 2017,
78 For a quick analysis on the topic: Elsa Kania, “China’s AI talent ‘arms race,” The Strategist, April 23, 2018,
Ministry of Education, “Artificial Intelligence Innovation Action Plan for Institutions of Higher Learning” [高等学校人工智能创新行动计划], April 4, 2018,
79 For English-language reporting on the topic, see: Patrick Tucker, “China’s Beating the US to Market on Combat Drones, By Copying US Technology,” Defense One, November 6, 2018,
For further details, see also: “Stealth UAV CH-7 First Debuted at Zhuhai Air Show” [隐身无人飞翼彩虹7号 首次亮相珠海航展], Observer Network, November 5, 2018,
80 For historical perspective on these challenges, see: Eliot A. Cohen, “Change and transformation in military affairs,” Journal of Strategic Studies 27, no. 3 (2004): 395-407.
See also: Norman Polmar, Aircraft Carriers: A History of Carrier Aviation and Its Influence on World Events, Volume I: 1909-1945. Vol. 1. Potomac Books, Inc., 2006.
81 For a more detailed discussion of the humans behind PLA “unmanned” systems, see: Elsa B. Kania, “The Human Factor in the “Unmanned” Systems of the People’s Liberation Army,” The Strategy Bridge, December 13, 2017,
82 Wang Yunxian [王云宪] and Li Dawei [袁大伟], “Intelligentized Operations Require Intellientized Training” [智能化作战呼唤智能化训练], China Military Network, August 23, 2018,
83 Wang Xingzi [王行自] and Nie Xiaoli [聂晓丽], “Advance the Development of Military Training Intelligentization” [推动军事训练智能化发展], February 2, 2018,  China Social Sciences Network,
84 Wang Yunxian [王云宪] and Li Dawei [袁大伟], “Intelligentized Operations Require Intellientized Training” [智能化作战呼唤智能化训练].
85 “The Typical Style of Intelligentized Operations: Human-Machine Cooperation!” [智能化作战的典型方式:人机协同!], PLA Daily, October 18, 2017.
86 Li Minghai [李明海]  “Where is the winning mechanism of intelligent warfare?” [智能化战争的制胜机理变在哪里], PLA Daily, January 15, 2019.
87 There have even been predictions that AI could change not only the character but also the nature of warfare. Frank Hoffman, “Will War’s Nature Change in the Seventh Military Revolution?” Parameters, Winter 2017,
88 For a more detailed consideration of the potential impact of AI on the balance of power, see: Michael Horowitz, “Artificial Intelligence, International Competition, and the Balance of Power,” Texas National Security Review, May 2018,
89 This white paper by a Chinese think-tank discusses a range of safety and security issues that include the risks of arms racing dynamics: China Institute of Information and Communications (CAICT), “AI Security White Paper” [人工智能安全白皮书], September 2018,
90 For an authoritative assessment of this issue, see: Lora Saalman, “Fear of false negatives: AI and China’s nuclear posture,” Bulletin of the Atomic Scientists, April 24, 2018,
91 Zhu Qichao [朱启超], “AI Intervenes in Military Affairs or Assaults Humanity’s Ethical Bottom Line,” [人工智能介入军事或冲击人类道德底线], The Paper, June 23, 2017,
92 China Institute of Information and Communications (CAICT), “AI Security White Paper” [人工智能安全白皮书], September 2018,
93 For an insightful discussion of these issues, see also: Paul Scharre, “Autonomous Weapons and Operational Risk,” Center for a New American Security, February 2016.
94 “Elon Musk Compares Building Artificial Intelligence To “Summoning The Demon,” TechCrunch, October 26, 2014,
95 “Attacking Machine Learning with Adversarial Examples,” Open AI, February 24, 2016,
96 For more on these issues, see: Paul Scharre, “Autonomous Weapons and Operational Risk,” Center for a New American Security, February 29, 2016,
97 For an early and authoritative assessment, see: “The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation,”
98 Ben Watson, “The Drones of ISIS,” Defense One, January 12, 2017,
99 “DJI is taking over every inch of the drone market,” The Verge, January 23, 2018,
100 “Isis use of hobby drones as weapons tests Chinese makers,” Financial Times, December 10, 2017,
overlay image