Tetiana Bubynets collapsed on the asphalt after a Russian drone exploded at a gas station in the southern Ukrainian city of Zaporizhzhia.

Despite the violent shockwave and blast of drone shrapnel produced by the attack on July 6, the 18-year-old accounting student had no visible wounds except for a nosebleed.

She was rushed to hospital and pronounced dead shortly after, alongside two other civilians, aged 41 and 48.

One of her lecturers, Tetiana Melikhova of Zaporizhzhia National University, said Bubynets was an only child and that her mother “lost the meaning of life” after she died.

“She couldn’t speak. At the funeral, she just lay on the coffin; she couldn’t let her child go,” Melikhova told Al Jazeera.

The strike that killed Bubynets was the first recorded case of Russia’s use of a fully autonomous attack drone powered by artificial intelligence (AI), according to a New York Times analysis.

While drone strikes have hammered Ukraine since Russia launched its full-scale invasion in February 2022, Ukrainian officials who examined the wreckage found the unmanned vehicle to be unlike others.

In place of an antenna, it featured an AI-powered minicomputer sold commercially by US technology company Nvidia, which rendered it capable of directing itself to the type of target it was programmed to spot – likely propane tanks in this case – lock on to it and proceed to strike.

AI-powered drones are the tip of the iceberg as Russia and Ukraine race towards boosting tech infrastructure aimed at coordinating reconnaissance and attack, according to analysts.

“It’s not about a drone having AI in it, and then seeing something and striking,” Marina Miron, researcher at the defence studies department at King’s College London, said. “Rather, it’s the drone seeing something, communicating to a software complex which knows how to map the GPS coordinates on a live battlefield map, and notifying the closest artillery unit.”

Referred to as reconnaissance-strike architecture, these unified battlefield management systems act like a central command. They track the location of friendly forces and identify enemy positions.

AI acts as the accelerator, instantly processing massive amounts of data collected through sensors in drones or any other military products, including interceptor missiles or artillery units.

The tactical advantage on the battlefield is given by the system’s ability to compress the kill chain – meaning coordinating and distributing the intelligence to military assets on the ground to find, track and destroy targets in a matter of minutes.

In such an ecosystem, AI-powered automated drones are just pieces of a broader puzzle.

Ukraine is known to be developing its system, which it named DELTA. Following operational trials in late 2025, Russia has been rolling out its version, the Svod Tactical Situational Awareness Complex, to front-line units.

Both sides claim that once the battlefield data is processed, a decision support system (DSS) presents a military commander with options to choose from, keeping a human in the decision-making loop, said Miron.

“In how far that is true is, of course, impossible to find out,” she added.

Acknowledging full AI autonomy would provoke ethical and legal questions.

Even if humans were to retain final decision-making authority, research has shown AI systems subtly shape choices, predetermining outcomes.

“Under time and pressure, commanders often do as AI suggests,” said Miron, whose research focuses on military AI and remotely piloted and autonomous systems.

Towards full automation

Coordinated AI-powered military systems are the next frontier of modern warfare, said Serhiy Beskrestnov, a Ukrainian Ministry of Defence adviser on loitering munitions, known by his callsign Flash.

“That’s what it will all come to,” he posted on his Telegram channel on July 1. “Navigation, search for a target, and attacks will be fully autonomous.”

Drones with some autonomous capabilities are already being deployed at scale.

Heorhii Volkov, commander of Ukraine’s Yasni Ochi drone strike unit, said this is currently mostly used in the final phase of long-distance strikes, where first-person view (FPV) drones are likely to lose radio connection due to electronic warfare or the curvature of the Earth.

The use of AI in so-called last-mile operations solves this battlefield hitch. If the signal is lost, an integrated AI-enabled onboard module takes full control of the unmanned vehicle, autonomously locks on the target and carries the strike to completion.

Yaroslav Azhnyuk, founder of Ukrainian technology company The Fourth Law, which develops AI-powered autonomy software for unmanned vehicles, said Russia and Ukraine were working to deploy a wider range of autonomous systems.

“This is not simply a race to demonstrate a fully autonomous drone. It is a race to make autonomy reliable, affordable and deployable at scale,” Azhnyuk told Al Jazeera. “Electronic warfare, increasingly long engagement distances and the growing number of drones are making human-controlled systems less sufficient. Autonomy solves exactly these constraints.”

Azhnyuk believes autonomy will become a game-changer over the next six to 12 months.

“The decisive advantage will belong to the side that can deploy reliable autonomy across tens or hundreds of thousands of systems, not just demonstrate it in prototypes,” he said.

Kateryna Bondar, senior fellow at the Center for Strategic and International Studies (CSIS), noted the dangers of AI’s reliance on algorithms.

“The system doesn’t understand what it’s doing; it’s just following the algorithm and the rules which people created,” said Bondar, who works at the CSIS’s Wadhwani AI Center. “AI has to be trained to recognise targets and engage these targets,” but writing code that distinguishes between a bus and a tank is easier than between a Russian and a Ukrainian soldier.

Additionally, cost-effective FPV drones have low-quality cameras and their resolution can be lowered further by adverse atmospheric conditions.

Military units using autonomous systems therefore “delegate the decision to a very imperfect technology that cannot distinguish between a civilian and a military person, combatant or non-combatant”, Bondar said.

“This is why there is a conversation around who makes a decision and who has responsibility for mistakes and errors.”