Ukraine Symposium – The Continuing Autonomous Arms Race

As Russia’s invasion of Ukraine drags into a grueling war of attrition, both sides are intensifying efforts to achieve a battlefield breakthrough with technological innovation. The invasion has sparked a rapid proliferation of low-cost drones and robotic systems that have fundamentally reshaped modern warfare, increasing reliance on autonomous and remote-controlled technologies.

The conflict has evolved into a hybrid battlefield where unmanned systems increasingly support conventional forces, transforming the nature of combat. Ukrainian territory has become a testing ground for the future of warfare, where emerging technologies and tactics are rapidly shaping the conflicts of tomorrow. In 2023, Ukraine claimed it was already deploying autonomous drones without human oversight that hit Russian targets. In the words of Mick Ryan, a former major general in the Australian Army, “Undoubtedly, the last three years have been the most innovative period for remotely operated, semi-autonomous, and autonomous systems since they were first introduced during the Second World War.”

The true cost of this war laboratory is often borne by civilians, who may suffer the most devastating consequences of relentless innovation on the battlefield. One of the biggest challenges for regulators in Ukraine and the West is developing autonomous systems that comply with international law while facing an enemy that operates without restrictions. Yet the pace of battlefield innovation is moving so quickly that regulators have a hard time legislating to keep up with the realities on the ground.

Drone Warfare in Ukraine

In a November 2023 interview with The Economist, Ukraine’s former commander-in-chief, Valerii Zaluzhnyi, compared the battlefield to the great conflicts of a century ago. “Just like in the First World War, we have reached a level of technology that puts us into a stalemate,” he stated, emphasizing the challenges of modern attritional warfare. Zaluzhnyi concluded that breaking the deadlock would require a significant technological leap that also involved unmanned and robotic systems, admitting, “There will most likely be no deep and beautiful breakthrough.”

At the start of the war, volunteer organizations like Dzyga’s Paw—comprised of former tech workers—scrambled to enhance soldiers’ situational awareness and improve artillery coordination. They focused on integrating drones with live video feeds, gradually advancing battlefield operations to provide troops with a clearer and more comprehensive view of their surroundings. One Ukrainian drone pilot recalled that in the early days of the war, he could deploy a drone from virtually anywhere, but now doing so requires extensive planning and preparation, as Russian forces have adapted and developed their own sophisticated countermeasures. These drones, which often cost just a few hundred dollars and are readily available, have played a significant role in neutralizing the traditional dominance of tanks on the battlefield. The Economist recently reported that in the battle for Velyka Novosilka, for example, armored vehicles played little role in the fighting. Captain Ivan Sekach of Ukraine’s 110th Brigade recalled, “One of our tanks moved near the frontlines, and ten drones attacked, setting it alight almost immediately.” Instead of tanks, the infantry fought the battle, with small Russian groups of three to five advancing in waves, often backed by their own tactical drones, artillery, and mortars.

The Effects of and Response to Drones

Ukraine uses drones as a vital substitute for its artillery shortages, helping to hold off Russian advances. Consequentially, Russian soldiers harbor a deep hatred for Ukrainian drone pilots, who have inflicted extensive casualties and significantly disrupted their operations on the battlefield. The precision and effectiveness of Ukraine’s drone strikes have made these pilots a top target, with Russian forces expending considerable effort to locate and neutralize them. In October 2024, in the Ukrainian-occupied Kursk region, Russian forces captured and executed nine Ukrainian drone operators, an act reminiscent of the brutal treatment reserved for flamethrower operators and snipers in the First and Second World Wars.

As drones became increasingly vital to the war, so did the development of countermeasures. Electronic jamming emerged as a significant challenge, making it harder to conduct drone operations effectively. To circumvent these disruptions, both sides began integrating fiber-optic technology into drones, ensuring reliable control even in heavily jammed environments. Additionally, efforts intensified to develop autonomous systems capable of operating independently of external signals.

The autonomous arms race extends beyond aerial drones, encompassing a variety of unmanned systems on the battlefield. In December 2024, Ukraine successfully carried out a groundbreaking all-drone, multi-domain attack on Russian positions near Kharkiv. This operation, conducted near the village of Lyptsi, was unique in that it relied entirely on weaponized uncrewed ground vehicles (UGVs) and first-person view (FPV) attack drones, all without any crewed platforms or soldiers on the ground. The Ukrainian National Guard confirmed the attack, with a spokesperson stating that Ukraine deployed “dozens” of unmanned systems equipped with machine guns and munitions to engage Russian forces. While these likely were remote-piloted systems, Ukraine is starting to make good on its earlier wishes to replace soldiers with robotic systems in the most dangerous combat scenarios.

Ukraine has additionally made significant advances in naval uncrewed systems. Ukrainian naval drones, both surface and underwater, have proven highly effective in the Black Sea, leading to the destruction of several Russian warships. Kyiv has ambitious plans to build “the world’s first fleet of naval drones” to further challenge Russian dominance at sea. The Russian Navy, in turn, is seeking to learn from Ukraine’s experience, as it intends to unveil its own naval drone lineup that will be apparently designed by artificial intelligence and machine learning tools.

Looking Ahead

In early 2024, Ukraine set an ambitious goal to produce one million drones within the year. However, as the war progressed and the demand for uncrewed systems grew, Ukraine revised its target to two million by March. In October, President Volodymyr Zelensky announced that Ukraine’s defense industry had significantly scaled up its capabilities, reaching an annual production capacity of up to four million drones. Two months later, Mykhailo Fedorov, Minister of Digital Transformation of Ukraine, predicted that 2025 “will significantly increase the percentage of autonomous drones with targeting. We might see the first real drone swarm uses, though not on a massive scale. The first steps will happen.”

Meanwhile, Russia is reportedly matching this pace of production. Vladimir Putin also recently directed the Russian government and its leading bank to strengthen artificial intelligence (AI) cooperation with China, signaling deeper technological collaboration between the two States. In late January 2025, the Russian government once again reiterated its main goals for military AI, which include automatic processing and analysis of intelligence data, improving information support for combat operations, and increasing the ability to predict threats and the course of conflicts. Russia considers itself a viable contender in the new arms race that the Kremlin views as “global AI competition.”

Kyiv’s Ambassador to the European Union, Vsevolod Chentsov, also stated that Russia has begun using AI to make it more difficult to jam their Shahed drones. Earlier Shaheds relied on GPS for navigation. The newer models use 4G data modems with Ukrainian SIM cards and Chinese satellite navigation antennas, allowing them to navigate via Ukrainian cell towers. This upgrade apparently improves their accuracy and helps them evade Ukrainian electronic warfare defenses.

The Russian defense industry, seeking to produce systems that can fly to targets on their own once such targets have been confirmed by operators and its technological startups, has unveiled at least half a dozen tactical drones that are reportedly guided by limited artificial intelligence. To match Ukraine’s successful drone interception of Russian unmanned aerial vehicles (UAV), the Russian military is developing its own UAV interceptors of Ukrainian drones that automatically detect and capture targets for neutralization.

Russia is also using drones to hunt civilians in Kherson, leading to allegations of a “human safari.” Locals suspect that Russian forces may be training their drone pilots by targeting Ukrainian civilians. However, with the gradual introduction of autonomous drones, Russia may also be training their AI models in autonomous drones. Moreover, to rush AI-enabled military systems to the front, the Russian government is betting on its many civilian AI developments becoming used for military purposes.

Training these AI models requires extensive battlefield data, of which both Ukraine and Russia have plenty. Oleksandr Dmitriev, founder of a Ukrainian non-profit that centralizes video feeds from over 15,000 frontline drone crews, told Reuters that since 2022, his system has collected two million hours—equivalent to 228 years—of battlefield footage. This footage trains the model on “combat tactics, spotting targets, and assessing the effectiveness of weapons systems.” According to the latest estimates, adding AI to FPV drones has increased Ukraine’s strike rate from 30-50 percent to around 80 percent.

Risk of Asymmetric Compliance and the Need for Human Control

What happens if one side develops more restrained autonomous systems while the opposing side does not respect international law? For instance, what if Ukraine trains its AI models in accordance with applicable frameworks and norms and Russian forces begin dressing as civilians to avoid detection by Ukrainian AI drones? Given the lack of armored vehicles and military-grade systems at certain parts of the front, Russia has already conducted assaults on the frontline using civilian cars.

These are some of the key issues that activists like those leading the campaign to “Stop Killer Robots” have yet to address. Even if some States agree to halt the development of lethal autonomous weapons, China and Russia certainly will not stop their own efforts given the ongoing technological arms race that advancements in the Russo-Ukrainian war exemplify. The “Stop Killer Robots” campaign urged States to push for new international law on autonomous weapons by 2026. However, with this “stop date” approaching fast, it seems unlikely that the technological developments evident in Ukraine will cease.

As one Ukrainian drone pilot warned, “The stakes are monumental. If we lag in pioneering these autonomous drones, adversaries like the Russians might outpace us, and they do not care how many innocent people they kill.” Nevertheless, ensuring human involvement in the targeting process is key. As another Ukrainian drone pilot noted, “A pilot must identify the target first and then allow the AI-guided homing system to strike the target in cases where technical limitations prevent manual targeting.”

This is also the concept behind Russia’s many drone developments involving target lock and automated target tracking. With human operators having the final say when it comes to target designation, the Russian military—at least for now—puts a human firmly in the loop for the final kill decision. At the same time, Russian industry claims on a regular basis to develop drones with greater autonomous capability that rely less and less on humans for guidance and target selection.

In an interview with the authors of this post, the commander of an aerial reconnaissance unit for Ukraine’s 128th Separate Mountain Assault Brigade emphasized caution in AI-controlled targeting. He noted, “The use of artificial intelligence for targeting should be limited to guiding munitions to pre-programmed coordinates in an intelligent manner. It is probably unwise to entrust human lives to an algorithm, as there is always a risk that something could go wrong.” He nevertheless acknowledged AI’s advantages in intelligence gathering: “When it comes to detection, reconnaissance, surveillance, and control systems, AI could be more effective than humans, who may be hindered by fatigue, lack of experience, or other human factors.” Still, such accuracy also depends on vast datasets for proper and correct training and identification of potential targets, something Ukraine has publicly acknowledged it already has.

According to Mick Ryan, most States have strict guidelines requiring human involvement in targeting decisions. However, the sustainability of these policies in the face of adversaries like China and Russia is unclear. The use of autonomous targeting could follow the path of social media, which has rapidly evolved beyond regulation and control, with no way to rein it back in. “My sense is that we will go the same way,” Ryan noted.

Gaps in International Law

There is no comprehensive legal framework governing the use of autonomous weapon systems as such in combat. While international humanitarian law provides a durable framework for managing and constraining the emergence of autonomous weapons, binding, platform-specific regulations remain elusive. Instead, many States are adopting non-binding protocols to enhance the responsible use of military-grade AI. These often aim to ensure “meaningful human control” or “appropriate human judgment” when such systems are used.

The incentives for States to adhere to such protocols largely stem from public perceptions of legitimacy, which play a crucial role in the sustainable use of force abroad. Democratic States are particularly motivated to adopt responsible AI guidelines to maintain public trust and accountability.

Despite ongoing discussions on ethical considerations, a complete ban on fully autonomous drones remains highly unlikely. As Paul Lushenko argued in an interview, “barring some unforeseen crisis, such as that depicted in the short film Slaughterbots, a ban on fully autonomous drones is unlikely because these capabilities are thought to provide countries a competitive edge during future war.” The need to outpace the enemy guides both sides in the Ukraine conflict to adopt the latest technologies and modifications to numerous weapons systems, with ethics often taking the backseat to the raw quest for survival or domination.

There are many risks associated with fully autonomous systems, such as the possibility of hackers taking control of drones and turning them against their operators, adding a new layer of complexity to the battlefield. Yet it is very difficult to limit how belligerents use military autonomy. Many such inventions rely on commercial technologies that are readily available and are therefore hard to restrict. These include, for example, the use of Raspberry Pi and Orange Pi microcomputers for drone operations.

Concluding Thoughts

Prior international debates on restricting military autonomy largely considered military-grade developments. They did not address civilian-made drones like those used in Ukraine, which involve numerous volunteer efforts on both sides assembling staggering quantities of tactical drones from Chinese parts. The recent unveiling of the Chinese DeepSeek AI model further obscures the debate on restricting or limiting potential autonomy. DeepSeek can enable advanced algorithmic calculations on a personal computer or a smartphone, thereby enabling combatants and drone operators (like those fighting in Ukraine) to use massive datasets directly in military engagements, with apparently few limitations.

Given that a comprehensive ban is improbable, the next best solution lies in States adopting unilateral constraints informed by evolving international legal norms. This could serve as a more feasible approach to regulating autonomous weapons in the absence of binding treaties.

At the same time, with the Russo-Ukrainian War as a backdrop, there seems to be no shortage of rapid advancements and public announcements regarding yet another advanced autonomous capability that will be used in combat. For Ukraine in particular, 2025 is the year it should unveil more advanced autonomous capabilities and “real” drone swarms powered by AI. Of course, as described above, Russia aims not to be far behind. This war-powered technology race does not appear to be losing steam, and what happens on the battlefields of Ukraine can potentially define how belligerents use military autonomy in other armed conflicts.

***

Samuel Bendett is an Adjunct Senior Fellow at the Center for a New American Security’s Technology and National Security Program.

David Kirichenko is an Associate Research Fellow at the Henry Jackson Society, a London-based think tank.

Photo credit: ArmyInform