The battlefield in Ukraine has become the epicentre of a global artificial intelligence (AI) arms race. In a remarkably short period, AI has transitioned from a background tool used for logistics and translation into a frontline combatant. AI innovation offers tactical advantages and compensates for troop shortages, drastically reshaping modern warfare. However, the speed of development creates a dangerous vacuum where lethal technology outpaces legal frameworks designed to govern it.
Ukraine as a High-Tech Proving Ground
Today, the war in Ukraine is defined by drone swarms, uncrewed boats and the world’s first uncrewed ground vehicle (UGV) battalions. These armed, partially autonomous robots have already seen combat success, even reportedly taking enemy soldiers prisoner. The UGVs may enter high-risk zones where human soldiers cannot, often armed with Kalashnikov machine guns, to observe and identify threats. While these systems are currently designed to identify rather than independently fire, the margin for error is slim. Experts say that the intensified use of AI in warfare has increased death rates and will continue to accelerate the pace and scope of modern warfare.
The US military and private sector leaders, such as former Google CEO Eric Schmidt, have described Ukraine as a “laboratory” for testing AI-led drone prototypes and new weaponry. While Russia also utilises AI for its uncrewed aerial vehicles (UAVs), its development is hampered by international sanctions. As Ukraine faces a significant disadvantage in manpower, it has been forced to invest heavily in software and unmanned systems to maximise autonomy on the frontlines. In addition to Ukraine and Russia, AI has also been deployed for lethal purposes in Palestine and Iran.
Evolution: From Logistics to Lethality
Historically, military AI was mainly used for translation, logistics and the maintenance of planes, cars and weapons. Today, AI is making weapons autonomous. As remotely controlled drones become standard, they are increasingly vulnerable to jamming, which severs the link between the pilot and the machine. To bypass this, developers are granting AI the “last-mile” autonomy, which is the ability for a drone to lock onto and hit a target independently once the human signal is lost. Software designers have also developed applications allowing flight control of a drone network to be passed between human pilots or for drones to be organised into sequenced swarm attacks with minimal human oversight. These capabilities are used by land brigades as well as air defence, intelligence and deep-strike units.
Beyond the hardware, soldiers are beginning to use AI systems to consult on tactical decisions. These systems are not yet widespread, but they are clear indicators of this shift.
An Innovation Race
In traditional Western military research, innovation was a top-down, slow-moving process. Today, breakthroughs are happening in the private sector and on the battlefield. To keep pace, Western militaries are attempting to adopt bottom-up innovation models, focusing on rapid experimentation and adaptation.
Because this race is largely occurring outside of traditional military settings, Silicon Valley and commercial tech hubs move at a speed that bureaucracies struggle to match. While NATO has established initiatives like the Defence Innovation Accelerator for the North Atlantic (DIANA) and the NATO Innovation Fund (NIF) to bridge this gap, these are primarily economic and technical solutions. Although Western militaries are attempting to adapt to this new form of innovation through specific military projects, new institutions, and civil-military collaborations, the technology race is outpacing legal frameworks.
The Principle of Human Accountability
The deployment of UGVs and autonomous drones raises a crucial question of International Humanitarian Law (IHL): If an autonomous system commits a crime, who is held accountable? Is it the programmer responsible for a coding error? The commander who deployed the unit? Or is it the AI itself? This is a serious concern of human rights and protection of civilians in armed conflict.
The solution has been to keep a “human-in-the-loop”. Recent scholarship suggests we should move from discussing “human control” to “human agency”. While “control” implies the human is a master in total command, “agency” acknowledges that AI and humans now share cognitive tasks. When a soldier must react to an algorithm’s target identification in milliseconds, the machine’s suggestions become impossible to override. This way, the human’s behaviour is channelled by the software. Militaries tend to use the comfortable label of “human-machine teaming” to suggest humans remain in charge, but this is a dangerous simplification that risks making the long chain of responsibility invisible.
Conclusion: A Call for Dynamic Regulation
This rapid development in AI technology will not halt. The tactical benefits are too great for any nation to ignore. The use of UGVs and AI-powered drone swarms will only increase. There is a need for a new “Digital Geneva Convention” or updated legal frameworks that are as dynamic as the software they seek to regulate. Technology has the potential to enhance our collective security, but only if we establish the legal safeguards necessary to protect our humanity.
Recommended readings:
Chivers, C. J. (2026): https://www.nytimes.com/2025/12/31/magazine/ukraine-ai-drones-war-russia.html
Csernatoni, Broeders, Andersen, Hoijtink, Bode, Lindsay & Schwarz (2025): https://link.springer.com/article/10.1007/s11023-025-09741-0
Goncharuk, 2024: https://icds.ee/wp-content/uploads/dlm_uploads/2024/09/Layout-AI-in-Defence-of-Ukraine.pdf