A portrait of a soldier from the “Taifun” unmanned aerial vehicle unit holding a new model ‘Marsianin’ attack drone on April 7, 2026 in Kharkiv region, Ukraine.
Getty Images
When Vladimir Putin launched his invasion of Ukraine in February 2022, he expected a quick victory rather than a prolonged war. Now, well into its fifth year, Russian soldiers struggle to advance through waves of Ukrainian drones across the fields of the Donbas. The character of the war has changed, as cheap unmanned systems increasingly dominate the battlefield, and the next contest is over autonomy.
Early in the war, Ukraine turned to first-person-view drones for reconnaissance before adapting them to drop explosives. Over time, the drones grew in range and capability. Ukraine also built a defense ecosystem capable of producing and procuring millions of drones.
For its part, Russia adapted to Ukraine’s growing use of drones along the front by deploying electronic warfare systems at scale. Drones that rely on radio communications are vulnerable to jamming, disrupting the signals between the operator and the aircraft.
Starlink satellite communications have been a key advantage for Kyiv, providing reliable connectivity on the battlefield. However, their use is restricted to Ukrainian territory and not available for operations inside mainland Russia.
Bryan Pickens, a former U.S. Army Green Beret who fought in Ukraine, told me Russia pushes waves of infantry forward while holding key assets deep behind the front, protected by heavy electronic warfare. In the past, valuable targets were often located roughly 100 kilometers behind the front, a zone Ukraine lacked the ability to strike at scale until 2026.
“Anything radio-controlled is largely obsolete in those environments,” Pickens said. “Autonomous systems or satellite communications enabled systems are the future.”
Pickens added that AI navigation and autonomous targeting have made strikes against Russian forces more effective. “Technology needed to adapt,” he said.
Cheap AI Is Changing The Battlefield
Among the drones Ukraine is using against Russian forces is the Hornet, an AI-enabled fixed-wing strike drone produced by Perennial Autonomy, a U.S. defense technology company founded by former Google CEO Eric Schmidt. The system reportedly costs less than $10,000 per unit and has become a key component of Ukraine’s campaign against Russian logistics.
Other Ukrainian mid-range drones, including Darts, also use AI-enabled systems for guidance and targeting.
“The only effective drones are those that can fly autonomously on pre-planned routes and require only minimal human input in the final stage,” said Pickens. He compared manually controlled first-person-view drones to muskets, while autonomous drones are the equivalent of rifled rifles. “We’re moving toward a world where one operator programs 100 drones,” he added.
These AI-enabled drones have already prompted Russian countermeasures.
Danylo Tsvok, head of the Defense AI Center A1, established under Ukraine’s Ministry of Defense in March to help build an AI-driven military, told me, “The enemy has begun camouflaging its equipment to deceive our drones’ computer vision systems.” Ukraine responded by retraining its models on new data, specifically images of equipment using the new camouflage patterns.
“The future of warfare is a competition of operating systems,” said Tsvok. “The side whose system understands the battlefield better, draws conclusions faster and recommends action wins.”
Autonomous technologies help speed up the kill chain, improving target identification, while increasing strike effectiveness.
“Obi,” a special forces operator from New Zealand serving in Ukraine’s 3rd Separate Special Purpose Regiment, told me that Russian units try to conceal vehicles in trees, bushes, and under camouflage, but AI can identify targets that human operators might miss. “Once the target is locked, the operator can click exactly where they want the drone to hit.”
Obi added, “Flying a Hornet is nothing like flying a traditional FPV drone, where precise stick control is required throughout the flight. With these newer systems, someone who is comfortable using a computer can learn very quickly.”
Still, for Ukraine, the challenge is not just developing AI systems but making them widely available and standardized. “Units don’t all have the same systems,” said Pickens. “There’s no universal operating environment.”
Earlier in the war, many companies rushed to develop autonomous systems and sold upgrades to frontline units that often failed in combat, eroding trust in the technology.
Yaroslav Azhnyuk, founder of The Fourth Law, one of Ukraine’s leading developers of autonomous defense technologies, is trying to solve both problems.
“Units that have tried other last-mile autonomy solutions often tell us they don’t need autonomous targeting because previous systems didn’t work well,” said Azhnyuk. “We then ask them to try our system. After a short demonstration, many immediately ask for large quantities because they can see the difference.”
Among the systems the company offers is the TFL-1 autonomy module. The cheapest version, available in Ukraine for as little as $150, offers frontline units a cost-effective way to add AI targeting capabilities to their drones. The FPV modules are built around Raspberry Pi Zero computers, which cost around $15 each. “It’s important that autonomous systems are priced proportionally to the drone itself,” added Azhnyuk.
Lypunis-10-TFL-1-T, a night-operating strike drone featuring Level 1 autonomy.
Photo: Iryna Supruniuk
“Our module replaces the standard camera with a camera integrated with an onboard computer,” said Azhnyuk. “It then interfaces with the rest of the flight system.”
Azhnyuk said roughly 200 companies in Ukraine have attempted to solve the challenge of last-mile targeting, but only a handful have done so effectively. “One is ours,” he said. “The other is Eric Schmidt’s group of companies operating under the Perennial Autonomy brand.”
Azhnyuk added that Perennial Autonomy’s systems are fully integrated, while The Fourth Law’s are modular, allowing customers either to install the company’s autonomy modules on existing drones or purchase drones with the modules already integrated.
When units find technology that performs well at an acceptable cost, adoption can spread quickly across the front. “In one brigade, we increased mission success rates from roughly 20% to 71%,” said Azhnyuk. “At the same time, we expanded the effective kill zone from approximately three kilometers to ten kilometers while increasing the cost of each drone by only about 10%, from roughly $500 to $550.”
Yaroslav Azhnyuk, founder and CEO of The Fourth Law.
Photo: Iryna Supruniuk
For frontline units, the increase in effectiveness for a modest increase in cost makes a compelling case for adopting last-mile autonomy. “Ukraine’s procurement system buys those products not because they’re Ukrainian, but because they’re the best in action and effectiveness,” said Azhnyuk. “These are life-and-death decisions.”
Using the added TFL module, once a Ukrainian operator identifies and locks onto a target, the drone no longer depends on the control link during the final phase of the attack. At that point, onboard autonomy takes over, making control link jamming largely irrelevant, Azhnyuk said.
The Race To Scale Autonomy
With the technology proven, the next challenge is scaling it.
Still, the technology deployed on Ukraine’s battlefield is not fully autonomous. Deborah Fairlamb, co-founder of Green Flag Ventures, a venture capital fund investing in Ukrainian startups, told me she considers these systems AI-enabled.
Azhnyuk argues that the “real game changer in current and future warfare is massively scalable individual autonomy,” with swarming representing just one application of that capability.
He describes five levels of autonomy: last-mile strike, last-mile bombing, target detection, navigation, and autonomous takeoff and landing. A drone with all five capabilities is, in his view, fully autonomous. Connect enough of these drones through a coordination layer, and they become a swarm.
“You achieve scalable individual autonomy, and then unlock effects that are much more powerful and much more broadly applicable than swarming itself,” Azhnyuk said.
The next challenge, he argues, is measuring autonomous systems objectively. Just as large language models are evaluated against standardized benchmarks, drone autonomy should be tested across thousands of scenarios involving target recognition, last-mile guidance, navigation, camouflage, terrain, and environmental conditions.
“This needs to be approached scientifically, not by AI companies claiming their systems work 100 percent of the time,” he said.
Taken together, Ukraine is seeking to field autonomous systems that are cheap, effective, and easy to use, while establishing clear benchmarks to measure their performance. In a war of attrition dominated by unmanned systems, greater autonomy will enable drones to strike deeper behind enemy lines with greater accuracy, expanding the no man’s land between opposing forces. It could also help Ukraine sustain its defense over the long term.
Leave a comment