Uncrewed but Confident: Forging New Rules of the Road for Drone De-escalation

On March 14, 2023, a U.S. Air Force Predator uncrewed aircraft was destroyed after it was struck by a Russian fighter jet in international airspace over the Black Sea. The actions conducted by the Russian fighter, including dumping fuel and flying in front of it, would have constituted a violation of the U.S.-Russia Incidents at Sea agreement — if the aircraft had been crewed.

In a time of growing international tension, with interstate warfare ongoing in Europe and 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. Given most militaries’ relative lack of experience in operating uncrewed systems, with novel benefits come novel risks: of accidents, miscalculation, and inadvertent conflict. The ongoing large-scale introduction of autonomous and uncrewed systems calls for the development of rule sets and confidence-building measures to mitigate the risks associated with their operational employment.

While successful confidence-building measures have been rare, there is promise in verifiable measures built on existing maritime and air safety rules. With this in mind, the United States and China should amend existing bilateral agreements to cover uncrewed systems and update multilateral maritime and air rules. The two countries should also work toward establishing standard signals and communications for uncrewed platforms, as well as defining potential exclusion zones for armed uncrewed systems, agreeing to not deploy nuclear weapons on uncrewed systems, and creating mechanisms to indicate loss of control of uncrewed systems. In the absence of U.S.-Chinese agreement, unilateral adoption of confidence-building principles by the United States (and also potentially its allies and partners) could be beneficial. However, scenario exercises indicate that confidence-building measures may be ineffective during full-blown crises, when uncrewed systems could provide stability benefits.

Confidence-Building Measures for New Technology

As states compete in their application of new technologies, norms governing their employment will inevitably 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. Confidence-building measures are a type of international agreement that can clarify and reinforce norms of military behavior. Traditionally, they have consisted of actions falling into categories such as exchanging information between parties, exchanging observers and conducting inspections, establishing “rules of the road” for military operations, and applying restraints on the operations of military forces. Pursuing such a process of political commitments, rather than a formal treaty, can allow for a more flexible approach to construction and diffusion of norms — most confidence-building measures do not require the same level of state commitment as a treaty.

There are certain considerations that policymakers should keep in mind when developing such measures. First, grafting new norms onto existing rules and behaviors can enhance their legitimacy, making them seem like outgrowths of existing rule sets and institutions. Framing a norm as a “best practice” with early adoption by a few key players can help to encourage adoption by others, with the threat of labeling those who do not sign up as “rogues.” Next, pursuing norms on a bilateral basis can have advantages. Adoption may be easier with only one other party involved, and transaction costs may be relatively low. However, pursuing norms on a multilateral basis among like-minded states can also be useful, creating islands of norm convergence that can then influence states outside the group.

Risks of Inadvertent Escalation

In recent years, analysts of AI and autonomous military systems have discussed new risks associated with these tools, such as the brittleness of algorithms, the potential for arms racing, and the risk that militaries might put too much trust in AI-driven systems. Another cause for concern is that, while accidental or inadvertent escalation into open conflict is somewhat rare in the historical record, there are several factors that put the United States and China at higher risk of such an event happening in the future.

First, if an air or maritime incident occurs between China and the United States, there are likely to be quite different perceptions of among the parties as to what caused the incident. If the incident happened within China’s “Nine Dash Line” or the Taiwan Strait, China may believe that it happened in “their” air or sea space, despite these areas’ 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 back to the initial forging of the ethos of the U.S. Navy in the Barbary Wars of the early 19th century.

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. U.S. efforts to restore these links have been frequently rebuffed, though there have been some recent signs of progress at the highest levels. Unlike Washington, Beijing does not necessarily see these links as tools whose utility may be at its greatest when bilateral relations are at their lowest. Instead, Chinese officials seems to perceive them as a reward to be given to the U.S. government for what they consider less hostile behavior and aim to dissuade U.S. military operations in China’s neighborhood by deliberately raising the perception of risk.

Perhaps the most important factor 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 so than in past eras of naval warfare, victory is likely to go the side that can attack effectively first. This is reinforced by history (and arithmetic) that shows that — given roughly equal technology and training — the larger navy almost always wins. The Chinese navy is — perhaps more so than any other major navy in the world — a missile-centered force. Further driving first-mover incentives, Chinese military doctrine writ large is focused on attacking suddenly to seize air and information dominance before an adversary like the United States can bring its other advantages to bear. For both sides, the incentives to quickly escalate by striking first could cause a crisis driven by accident and/or misunderstanding to spiral far more quickly than has been seen in previous international confrontations. This is on top of the even greater speed with which developments may unfold given the faster decision-making that may occur when conducted by AI systems — or due to their assistance to human decision-makers.

Key Considerations for Confidence

Considering the history of confidence-building measures, a few principles emerge that provide a useful framework for potential rules development. First, successful confidence-building agreements are rare and tough to get right. The process of negotiating such agreements can actually consume trust, and they are only as strong as the fundamental political will for compromise. There are numerous challenges to negotiating and implementing them, including the perceived cost of the measures — political, military, and otherwise — versus their likely benefits. Such challenges are particularly acute when the negotiating parties lack a shared understanding of the risks of conflict or its potential causes, and if asymmetries exist in military capability and investment. Challenges to implementation can include difficulty in maintaining the interest of parties, incomplete and vaguely drafted agreements, domestic political challenges, and deceptive or selective compliance with the agreements.

While confidence-building measures can be a useful coordination mechanism when states seek to avoid conflict, they are not an effective constraint against deliberate escalation. For example, India and China have multiple agreements in place to reduce border tensions, including restricting military buildups along their border. These agreements did not prevent a series of deadly clashes between Indian and Chinese troops beginning in 2020. However, they may have prevented greater fatalitiesthan would have been the case had the agreements not been in place. In the end, states may violate confidence-building measures if they see it as in their interests to do so.

Despite such challenges, air and maritime uncrewed platforms in the Indo-Pacific may be a promising area for developing new confidence-building measures. As a status quo power, the United States is likely to want to avoid unnecessary conflict that might upset stability in the Indo-Pacific, while still indicating a willingness to fight. On the Chinese side, their military doctrine expresses a strong interest in maintaining “effective control” during a crisis — for example, by not having an accidental war. Pre-existing U.S.-Chinese air and maritime safety agreements provide starting points for negotiation and iteration, and China’s growing nuclear arsenal may motivate both sides to seek greater strategic stability. Maritime confidence-building has often been productive — see the long history of maritime prize law and centuries of regulations to avoid collision — and the locus of U.S.-Chinese competition is a largely maritime theater.

Another consideration in developing successful confidence-building measures is to build on existing rule sets and agreements. As precedents, the U.S.-Soviet air and maritime safety agreement, current U.S.-Chinese air and maritime safety agreements, and the multinational Code for Unplanned Encounters agreement all lean heavily on existing rules of behavior and means of communication. International collision avoidance regulations, the U.N. Convention on the Law of the Sea, and international civil aviation guidelines are all referenced extensively within these agreements in order to provide well-thought-out details of expected behavior during interactions by vessels and aircraft.

When the level of trust between competitor states is low — such as between the United States and China — confidence-building measures are more likely to be effective if they include provisions for verification. As such, confidence-building measures governing uncrewed systems are likely to be most successful if they focus on information and behaviors that are verifiable, and not on non-verifiable items where a lack of existing trust may drive suspicion of violations — real or imagined. It bears considering that any approach to China will need to be a delicate one, as China’s record of support for verification provisions in arms control is decidedly a mixed one. China tends to believe that verification activity can provide an asymmetric advantage to the stronger party, posing a threat to its security interests. China also tends to see mutual national trust as a precondition for verification cooperation, rather than the typical Western preference for implementing confidence-building measures as a means of building that trust.

One last factor to consider is that a potential stability-instability paradox exists for uncrewed systems which could make them useful in a crisis. Specifically, the reduced emotional impact of losing uncrewed systems versus human lives could act as a pressure-release valve for countries looking to coerce the other in times of tension without escalating to war. Uncrewed systems are different than crewed assets precisely because they allowed decision-makers to mitigate risk in otherwise dangerous scenarios. This creates the paradox: there might be significant contestation — and even bloodless violence — because of the increased use of uncrewed systems, but the use of machines over human platforms may also create a kind of stability, allowing states to de-escalate even with the loss of valuable uncrewed systems. 

Options for Uncrewed System Confidence-Building Measures

For the United States and China, there are three existing agreements that could be updated to account for the introduction of uncrewed air and maritime platforms: the 2014 U.S.-China Memorandum of Understanding for safety of air and maritime encounters; a 2015 supplement to that memorandum covering air-to-air encounters; and the multinational Code for Unplanned Encounters at Sea.

In addition to updates to these agreements, the current international regulations for prevention of collisions at sea (COLREGs) require updating to cover uncrewed vessels. In the case of the 2014 U.S.-Chinese memorandum, there are several gaps. First, due to the definitions provided within the document, the agreement appears to exclude uncrewed warships. Next, while the agreement describes actions that each side’s commanders should avoid, such as shining lights at another vessel’s bridge or shining a laser at personnel, it contains no guidance for actions that should be avoided with respect to the other side’s uncrewed vessels — such as blinding sensors, interfering with their systems operation, or jamming communication and control channels. The agreement also states that both sides should not board or salvage the other’s military vessels or aircraft without prior explicit consent. The agreements also provide communication rules which require updating, as uncrewed platforms may not be equipped to conduct such communication over the usual frequencies. The U.S.-Chinese agreement on aircraft safety also requires revision, as its provisions for aircraft collision avoidance are ambiguous when applied to uncrewed aircraft and provide little guidance for how uncrewed aircraft should interact with each other.

The current multinational Code for Unplanned Encounters at Sea agreement should also be updated to account for uncrewed platforms. It, too, excludes uncrewed vessels as warships, and does not provide means for communication with or between uncrewed systems. The regulations for collision avoidance at sea will also require amendment for remotely operated and autonomous vessels to account for terminology, lights, shapes and sound signals, the role of vessels’ masters, responsibility of remote operators, and distress signals.

To close these gaps, the United States could make a unilateral declaration that extends U.S. compliance with the 2014 and 2015 U.S.-Chinese memorandums, as well as the Code, to uncrewed vessels and aircraft. Another variant of this approach could be to make a plurilateral declaration with other like-minded nations, providing greater momentum for norm adoption. This approach could provide a benefit for the United States and its partners — establishing new norms and pressuring China. Of note, this approach might also make China more suspicious, given the U.S. role as a first mover in the process.

Next, the United States could pursue a minimalist negotiated approach. This would involve working with China to make minor additions or revisions to current agreements to explicitly extend coverage to uncrewed vessels and aircraft. Both sides would then issue additional statements saying they would follow existing international guidance to the greatest extent practicable and await forthcoming revisions to those international rulesets.

As a third option, the United States could pursue a pro-active negotiated approach. Under this process, Washington would work to negotiate additional agreements covering uncrewed platforms, such as a ban on the carriage of nuclear weapons by uncrewed platforms. These would be accompanied by additional information exchange provisions, and provisions for communicating when an uncrewed platform has “gone rogue” and is no longer under effective control. Both sides could also develop provisions to define types of areas where uncrewed platforms would only be operated unarmed, in the interest of mitigating the risk of crisis escalation. While neither side is likely to agree on specific geographic areas where such restrictions would be in place on a permanent basis, they could be announced by either or both sides’ militaries in the event of a crisis or heightened tensions.


Given what has been observed in Ukraine and other recent conflicts, the likelihood of successful arms control measures for uncrewed systems is low. This means the focus of the American and allied defense establishments should be on confidence-building measures. The measures most likely to succeed would be those that build on existing rule sets and agreements, and that focus on provisions conducive to verification. In the end, policymakers should remember that the utility of these confidence-building measures could be limited in a full-blown U.S-Chinese crisis, when both sides might specifically use uncrewed systems as lower-stakes platforms that can serve as a stabilizing “half-rung” in an escalation ladder. But by implementing confidence-building measures at other, more normal times, policymakers on both sides may be able to reduce the risk of just such a crisis happening in the first place.

Retired Capt. Thomas Shugart, U.S. Navy, is a former submarine warfare officer, an adjunct senior fellow at the Center for a New American Security, and the founder of Archer Strategic Consulting.

Image: U.S. Navy

Read More

Thomas Shugart