Two years ago, a quadcopter drone flew autonomously to Ukraine's front line, switched on its AI targeting system, and killed Russian soldiers with no human pulling the trigger. Alexander Kokhanovskyy, CEO of the Ukrainian drone maker Aero Center, dropped that detail in a recent interview with New Scientist at an event hosted by the Ukrainian embassy in London.
The Single Test With No Video Feed
Kokhanovskyy described the drone as preprogrammed to fly to a specified area before activating what the team called a "Terminator mode" — an AI-powered targeting loop that would seek out and attack any target in that zone. No video feed streamed back. No operator confirmed the kills in real time. After the mission, human-piloted drones flew reconnaissance over the area and found "a couple" of dead Russian soldiers. That was the evidence: dead bodies, no kill chain recording.
Aero Center has not released any sensor logs, flight telemetry, or target classification data. We're taking the CEO's word — and the word of the scout drone pilots who counted the casualties.
What This Means for Autonomous Weapons
This claim, if it holds, marks a genuine first: a fully autonomous drone — one that chose and engaged targets without a human in the loop — for lethal effect in active conflict. The era of AI deciding who lives and dies in combat isn't hypothetical anymore; it's a one-off field test with real consequences. Every major military is watching. Open questions remain: how did the AI distinguish combatants from civilians? What rules of engagement were baked in? Without a video feed, we can't audit the logic that led to those kills.
Kokhanovskyy didn't offer answers, and Aero Center likely won't. But the next test won't be a one-off. Engineers building autonomous targeting systems now have a real-world data point — and a moral ledger that just got heavier.
Source: Ukraine's one-time test used fully autonomous drones to kill Russian soldiers
Domain: arstechnica.com
Comments load interactively on the live page.