Zelenskyy warns AI could lead decisions on the battlefield next year
Drones have transformed the way wars are fought, and artificial intelligence is making them faster and more capable, according to Yaroslav Azhnyuk. Azhnyuk runs The Fourth Law and Odd Systems, Ukraine-based defense firms that he says are among the top companies producing AI-powered drones in the country. “When the other side has orders of magnitudes higher capabilities, it gives you an asymmetric advantage,” he said. “It can make you 10,000 times more effective.”
Dozens of Ukrainian military units already use the firms’ autonomous drone, the Lupynis-10-TFL-1, according to Azhnyuk. Russia is also deploying AI-enabled drones against Ukraine, and the resulting high-tech arms race has drawn concern from Ukrainian President Volodymyr Zelenskyy, who has warned that AI could lead decisions on the battlefield as soon as next year.
Azhnyuk, 37, is a computer scientist from a family of scholars and scientists who spent years working in Silicon Valley. He founded Petcube, which makes smartphone-powered cameras that use artificial intelligence to watch pets while they are alone. “We sold a million of these devices that are quite popular in the U.S.,” he said. He met NPR in a conference room in Kyiv and asked that the location not be further disclosed for security reasons.
After Russia launched its full-scale invasion in 2022, Azhnyuk — by then living in Ukraine — pivoted to defense. “I was like, OK, I’ve got to move on from Petcube and do something to protect my country,” he said. “And so now we’re making not cameras that fling treats but cameras that fling explosives at people who came to occupy our land.” Those cameras are part of an image recognition system incorporated into first-person-view, or FPV, drones. “We are leaders in autonomous drones and thermal cameras,” he said.
The autonomous drone combines a small, affordable FPV drone with an autonomous terminal guidance unit, meaning an onboard computer and AI technology take control during the last 500 meters, or 0.3 miles, of the flight. Azhnyuk said the system works differently from large AI models. “But it’s very different from, say, the kind of AI that sits on those data centers that power Claude,” he said, “or even the kind of AI that does self-driving for your Tesla.”
The technology uses simple tools to recognize images and software that, in Azhnyuk’s words, allows the drone “either to navigate autonomously without a human or to choose a target autonomously or to fly toward the target autonomously.”
The Fourth Law named the drone after Anatoliy Lupynis, a 20th-century Ukrainian poet and political activist who endured decades of Soviet imprisonment and psychiatric torture. The company named all of its weapons after Ukrainian intellectuals and artists killed or persecuted by Russians in the last century. “It is important for us to help to resurrect those people,” Azhnyuk said. “Their ideas are still with us. And now they’re fighting back.”
The Fourth Law’s teams test the drones daily. NPR met one team in an overgrown field outside Kyiv, where members had set up a fold-out table near a copse of trees and a hornets’ nest. The company asked that the location and the identities of those who work there not be disclosed for security reasons; the team members use call signs. Stab, the head of the test and integration team, pointed to a drone. “So you see here is our camera, plus the microcomputer inside this 3-D model,” he said. “All of the mind is here.” The microcomputer runs a deep-learning neural network that uses hours of combat footage to recognize targets — “vehicles, light-armored vehicles, tanks, and flying objects,” Stab said.
A flight begins with a pilot, who also uses a call sign, Professor, guiding the drone with a control stick. In a demonstration, Professor selected a target — a dilapidated car an employee drove along a dirt road — and watched a video monitor until a red flash appeared. “That means it has locked,” Stab said. Professor released the control stick, and the drone flew itself to the car, its microcomputer running algorithms to keep it on course. “See, Professor is doing nothing, and the drone is doing everything itself,” Stab said. On the battlefield the drone would carry an explosive to hit its target; the test drone carried none.
Stab said AI drones are useful because they are hard to jam. He also insisted safeguards are in place so The Fourth Law’s drones will not go rogue. “Humans select the target,” he said.
Back in the conference room, Azhnyuk said he acknowledged the concern over destructive weapons but argued it is not specific to AI. “It has been with humanity since the invention of fire or steel,” he said. “Any kind of tool can be used for good or for bad. And in the end, it’s the human that decides how to use that tool.” He also rejected the idea that his company’s drones are killer robots. “They’re very far from a Terminator kind of scenario, right?” he said. “These are just sort of calculators. Like they’re no more likely for a Terminator uprising than a warehouse with artillery shells.”
Azhnyuk acknowledged that the technology is developing quickly. “It’s not unthinkable,” he said, “to imagine that some years in the future there will be drones that will be able to make their own decisions on designating targets or not.” For now, he said, Ukraine has little choice but to compete. “Our adversaries are using the very same technologies, and if they use AI for decision-making in an offensive scenario, and if we refuse to use AI in decision-making in a defensive scenario, we’ve lost, right?” he said. “Because their systems are capable of making decisions in microseconds, and human control takes minutes, hours and sometimes days.” If Ukraine loses, he said, “there’s no way for us to argue about the ethics of AI.” The most important thing, he said, is don’t die.