Autonomy and International Stability

Executive Summary

In a time of growing international tension, with interstate warfare ongoing in Europe and openly threatened over Taiwan, military forces worldwide are looking increasingly to autonomous and uncrewed systems to provide new capabilities, field forces more quickly and at reduced costs, and reduce risks to military personnel.¹ Given most militaries’ relative lack of experience in operating uncrewed systems, with novel benefits come novel risks: of accidents, miscalculation, and inadvertent conflict.

During previous times of military innovation and international tension, even adversarial nations agreed on rule sets that standardized and managed interactions, increased communications, and mitigated the negative impacts of novel weapons and platforms. The ongoing large-scale introduction of autonomous and uncrewed systems, particularly in a time of increased great power tension, calls for the development of rule sets and confidence-building measures (CBMs) to mitigate the risks associated with their operational employment.²

This report carefully considers the history of CBMs and finds there is promise in verifiable measures built on existing maritime and air safety rules, even if successful CBMs have been rare in practice. As such, this report endeavors to develop realistic and mutually beneficial rule sets and CBMs for U.S. and international policymakers to consider. The report’s recommendations include amending U.S.-China agreements to cover uncrewed systems and pushing to update multilateral maritime and air rules, as well as establishing standard signals and communications for uncrewed platforms. It also recommends defining potential exclusion zones for armed uncrewed systems, agreeing not to deploy nuclear weapons on these systems, and creating mechanisms to indicate loss of control of them. In the absence of U.S.-China agreement, unilateral adoption of CBM principles by the United States (and potentially its allies and partners) could be beneficial.³

Introduction

This report begins with an overview of behavioral norms and their relation to CBMs, and then examines the risks of inadvertent U.S.-China military escalation that CBMs could help forestall. It then provides a research framework for U.S.-China CBM development and provides key questions to consider in doing so. The report provides research findings on CBMs and how they might be applied to uncrewed systems within the U.S.-China strategic competition, and analyzes gaps in the existing CBM framework that could be addressed to provide coverage for them. It then lays out specific options for uncrewed system CBM development and provides the results of a tabletop exercise (TTX) that assessed the potential effectiveness of those options within the context of a hypothetical crisis. Finally, it provides recommendations for uncrewed system CBM development and implementation, as well as specific CBM provisions and wording.

The paper endeavors to answer the following questions about potential uncrewed system CBMs:

• Which types of autonomous systems are of greatest relevance for potential CBMs?
• What information should be shared between participating nations? How might that information be shared?
• How might states share results from accidents or near-accidents caused by uncrewed systems?
• To what extent may accepted behavior for crewed systems and platforms need to be adapted for uncrewed systems?
• How might uncrewed systems communicate their level of autonomy, such as through standardized signals or marking?
• Should there be specific geographic areas within which the use of uncrewed systems (or certain categories of uncrewed systems, such as armed ones) is mutually curtailed or prohibited?
• Should autonomy of nuclear-capable systems be uniquely restricted?
• What unilateral declaratory statements could be useful as precursors to full international rule set implementation?

Confidence-Building Measures for a New Technological Competition

As states interact with each other in their application of new and changing technologies—such as uncrewed systems—norms of behavior inevitably will form and evolve over time. Norms can be reinforced or clarified by communication between states, through unilateral declarations or bilateral or multilateral agreements. But norms themselves are not established merely by fiat or declaration; rather, they grow out of a particular social context and consist of a shared understanding of appropriate behavior.⁴ Norms are related to, but not the same as, laws (including treaty law). The law can serve as a basis for norms, and norms can be formulated into law.⁵

CBMs are a type of international agreement that can clarify and reinforce norms of military behavior. Traditionally, CBMs have consisted of actions in the following categories:

• Exchanging information between parties,
• Exchanging observers and/or conducting inspections,
• Establishing “rules of the road” for military operations, and
• Applying restraints on the operations and readiness of military forces.⁶

CBMs tend to follow a process: they start with easier areas where agreement is possible; then, as cooperation becomes established, expand the range of agreement to more parties, or to more contested issues.⁷ Pursuing a process of political commitments like these, rather than a formal treaty, can allow for a more flexible approach to construction and diffusion of norms. Most CBMs do not require the same level of state commitment as a treaty (including, for some states, formal domestic approval through treaty-ratifying processes).⁸

As policymakers consider the best processes for constructing and adopting CBMs, they should keep in mind some key considerations related to norms. First, grafting new norms onto existing rules and behaviors can enhance their legitimacy, making them seem as outgrowths of existing rule sets and institutions.⁹ Also, critical to norm emergence are “norm entrepreneurs” who frame the problem: naming, interpreting, and dramatizing it and then, on that basis, proposing a norm to address it. Being first in framing the problem is important, as these frames can be hard to dislodge once established.¹⁰ Of note, this first-mover advantage could encourage the United States and its partners to push forward with CBMs in this area, rather than let its strategic competitors establish the framing of an important field of technological advance like uncrewed systems.

Once a norm is proposed, framing it as a “best practice”—with early adoption by a few key players—can help to encourage adoption by others, with the threat of labeling as “rogues” those who do not sign up. This is another incentive for the United States and like-minded nations to move first in promoting CBMs for uncrewed systems.¹¹

Lastly, pursuing norms on a bilateral basis can have clear advantages, as adoption may be easier with only one other party involved, and transaction costs may be relatively low. Two high-status norm adopters may also inspire more parties to sign on over time. However, pursuing norms on a multilateral basis among like-minded states also can be useful, creating “islands” of norm convergence that can influence states outside the group.¹²

Risks of Inadvertent Escalation as a Potential Road to U.S.-China Conflict

This report’s primary impetus for proposing CBMs for uncrewed systems is to avoid dangerous incidents that might inflame tensions between the United States and China. In recent years, multiple analysts of artificial intelligence (AI) and autonomous military systems have posited new risks associated with those systems—such as the brittleness of algorithms, the potential for arms racing and threats to nuclear stability, and the risk that militaries might put too much trust in AI-driven systems. This study aims to propose specific means to achieve mitigation of these risks.¹³

While the risk of accidental escalation may be apparent in close encounters between U.S. and People’s Republic of China (PRC) ships and aircraft, few major wars have broken out in this way—though there have been some close calls, such as the Cuban Missile Crisis.¹⁴ Regardless, while accidental or inadvertent escalation into open conflict may be somewhat rare in the historical record, several factors may result in a significantly higher risk between China and the United States in the years to come.¹⁵

First, given that a U.S.-China air or maritime incident seems most likely to occur at a flashpoint in China’s near abroad due to China’s expansive territorial claims, there are likely to be quite different U.S. and Chinese perceptions of what caused the incident. The PRC is likely to believe—actual international law notwithstanding—that the incident happened in “their” air or sea space, such as within China’s “Nine Dash Line” or within the Taiwan Strait (despite its status as international waters). From the U.S. side, a challenge to freedom of navigation and overflight that goes unmet would be a betrayal of generations of American commitment to the concept, going all the way back to the initial forging of the ethos of the U.S. Navy in the Barbary Wars of the early 19th century.

Additionally, unlike when the United States and Soviet Union maintained Cold War–era military communication links that were useful for defusing an accident- or misperception-driven crisis, the United States and China have not been using similar links. And this is despite an increasing number of close encounters.¹⁶ U.S. efforts to restore these links have been frequently rebuffed, though there have been some recent signs of progress at the highest levels.¹⁷ It seems that China perceives links with the U.S. military as a reward to be given to the U.S. government for what the PRC sees as less hostile behavior, rather than as a tool whose utility may be at its greatest when bilateral relations are at their lowest.

Last, and perhaps most important with respect to the possibility of rapid military escalation, is a piling up of first-mover military advantages that may drive escalation ferociously if one side believes it has been attacked—or even is about to be attacked. Decades of war at sea in the missile age show that, even more than in past eras of naval warfare, victory is likely to go the side that can attack effectively first.¹⁸ This is reinforced by history (and arithmetic) that demonstrates that in naval warfare—given roughly equal technology and training—the larger navy almost always wins; and a smaller missile-age navy that gets off a surprise first salvo may make itself the larger one in a matter of minutes.¹⁹ When it comes to naval warfare the People’s Liberation Army (PLA) Navy is—perhaps more so than any other major navy in the world—a missile-centered force. Further driving first-mover incentives outside of naval warfare, PLA doctrine writ large is focused on attacking suddenly (and likely by surprise) to seize air and information dominance before an adversary like the United States can bring its other advantages to bear. For example, the PLA has an entire service branch dedicated to missile warfare, the PLA Rocket Force, that is arguably the crown jewel of the Chinese military. Central to the PLA’s plans for any major conflict, the Rocket Force’s ballistic and hypersonic missiles likely would be significantly more effective if they could catch adversary aircraft on the ground and non-dispersed, target enemy ships in port and at their piers, and strike fixed command centers while they are still full of key decision-making personnel.²⁰ For both sides, the incentives to escalate quickly by striking first—while still able to do so, given both sides’ likely focus on paralyzing enemy command and control—could cause a crisis driven by accident or misunderstanding to spiral far more quickly than in previous international confrontations. ²¹ And this does not even consider the even greater speed with which developments may unfold given the more rapid decision-making that may occur when conducted by AI systems or due to their assistance to human decision makers.²²

Framing the Problem

This report examines how CBMs could impact the operational employment of U.S. and Chinese uncrewed systems to reduce escalation risks. The geopolitical framing of this study and proposed rule sets are grounded in the competition between the United States and China in the Indo-Pacific—an air- and maritime-dominated theater where both players have nuclear weapons that significantly increase the potential consequences of unintended escalation. This competition also includes two nations at the forefront of AI and uncrewed system development.²³

The report focuses on the operation of uncrewed air and maritime platforms, as they are most likely to interact with one other in the international commons. Such uncrewed systems could be autonomous or remotely piloted. The paper also covers loitering munitions, as the distinction between some such munitions and other uncrewed platforms continues to narrow over time.²⁴ It considers potential interactions between crewed and uncrewed platforms, as those seem at least as likely to occur—and may have greater stakes with personnel at risk in the crewed platforms—as uncrewed-on-uncrewed interactions.

This report considers what CBMs could be created and take effect over a period of roughly five to fifteen years. This timeline is keyed to what seems likely to be a time of increasing tension and geopolitical risk in the mid- to late-2020s.²⁵ On the short end of the timeline, a goal given the likely pace of negotiation and norm development could be to provide amendments of existing agreements in the interest of quickly putting in place CBMs.²⁶ A 15-year limit (out to 2039) provides a reasonable time to possibly obtain a more comprehensive rule set that incorporates more years of experience operating uncrewed systems. Technological and geopolitical conditions also are more likely to be difficult to predict much beyond that time frame, given the accelerating pace of autonomy and uncrewed system development.

This report assumes the following will color the geostrategic environment between 2029 and 2039:

The United States and China will continue to be engaged in a strategic rivalry across the Indo-Pacific region, driven by factors such as disputed Chinese territorial claims against U.S. allies and partners (including the continued existence of Taiwan as a separate state), disagreements over the freedom of air and maritime navigation, and U.S. and international criticism of China for ongoing human rights abuses—the mere enunciation of which China considers to be a violation its sovereignty.²⁷

The rivalry between the United States and China will not break out into a state of war, but at the same time the United States will remain engaged enough in the region that U.S.-PRC interactions remain a potential flashpoint for such conflict.

Uncrewed system development will continue on an upward trajectory, with no international agreements reached to ban or significantly curtail the use of such systems in international air and sea spaces.

Given some of the advantages that AI integration may confer on uncrewed systems (e.g., reduced reaction times, a greater ability to operate with denied communications, etc.), an increasing proportion of them become autonomous rather than remotely operated, introducing AI-associated risks and uncertainty.

Research Insights

Based on this paper’s review of the literature and history of CBMs, as well as the trendlines of uncrewed system development and employment, the following insights may provide a background for potential uncrewed system rule set development.

Successful CBM agreements are rare and tough to get right.

The very process of negotiating and executing CBM agreements can consume trust, and they are only as strong as the fundamental political will for compromise.²⁸ Challenges to negotiation include the perceived cost of CBMs—political, military, and otherwise—versus their likely benefits. These challenges are particularly acute when the negotiating parties lack a shared understanding of the risks of conflict or its potential causes, and include sensitivities to who proposes what and when, asymmetries in military capability and investment, difficulty reaching agreement on areas of applicability and participants, and asymmetries of impact.²⁹ Challenges to implementation can include difficulty in maintaining the interest of parties, incomplete and vaguely drafted agreements, misalignment or changing of national motivations between the negotiation and implementation phases, domestic political challenges, and deceptive or selective compliance with agreements.³⁰ Several examples exist of CBMs that have been negotiated and agreed to, then violated later when their provisions became inconvenient to one or both sides of the agreement. Soviet Navy warships conducted “shouldering” actions against U.S. Navy ships in the 1980s, more than a decade after the largely-successful U.S.-Soviet Incidents at Sea (INCSEA) agreement was signed between the two parties in 1972.³¹ On multiple occasions, Chinese PLA Navy warships risked collision and conducted close approaches to U.S. Navy ships in violation of the International Maritime Organization’s (IMO’s) International Regulations for Preventing Collisions at Sea (COLREGS), despite an existing U.S.-China maritime safety Memorandum of Agreement (MOU) that should prohibit such action.³² While such violations may not have rendered their associated CBMs useless, they do point to their limits. The INCSEA agreement reduced the number and severity of collisions and near-collisions between the U.S. and Soviet navies, but did not eliminate these incidents entirely.³³ In a crisis, violations could be perceived as justification for compensatory escalation, with a possible net negative result in terms of escalation risk.³⁴