The Report
On June 10, *New Scientist* published a report that changes the framing of the global autonomous weapons debate: fully autonomous drones have killed human soldiers on the battlefield for the first time.
The source is a senior figure in Ukraine's defense industry, disclosed by drone manufacturer Alexander Kokhanovskyy at a press event at the Ukrainian embassy.
The Incident
- Time: roughly two years ago (around 2024)
- Location: eastern Ukraine front line, near Bakhmut and Chasiv Yar
- Weapon: 10 quadcopter drones from the "Terminator" project
- Mode of operation: so-called "Terminator mode" — after launch, no human intervention at all; the AI model autonomously searched for targets, locked on, and attacked
- Result: several Russian soldiers and a truck
- Key fact: no video footage exists — because the drones ran fully autonomously, there was no recording. Crewed drones were later sent to the site to confirm the results.
- Major Danylo Polozhukhno, a senior officer in Ukraine's 21st Separate Unmanned Systems Regiment, stated explicitly: "We do not use fully autonomous drone systems to independently select and attack targets... Ukraine complies with international humanitarian law."
- Yet Kokhanovskyy himself is developing the next-generation system ALITA — 16 launch pads, 64 drones, 450 km/h, all controlled by 2 operators, expected ready by October 2026, aimed at intercepting Russian "Shahed" drones
- He says he is "very willing" (to relax restrictions).
- New Scientist original: https://www.newscientist.com/article/2529849-fully-autonomous-drones-have-killed-human-soldiers-for-the-first-time
- Policy-side coverage (China): https://www.ithome.com/0/961/749.htm
Kokhanovskyy's own words: "Once we launch it, we know that everything detected in that area will die. The drone has no connection — you can't see video... everything it sees will be destroyed."
A "Watershed Moment"
The report quotes scholars using the term "watershed moment." It matters not because of the technology — fully autonomous attack has long been technically feasible and was hardly news in academic circles — but because it is the first public case with a report, a disclosure, and specific execution details.
Until now, "fully autonomous weapons" in public discussion remained a hypothetical. A 2021 UN report noted that Turkish-made Kargu-2 quadcopters may have autonomously attacked humans in Libya the year before, but details remained vague.
Now there is, for the first time, a public case where someone is willing to step forward and say "I did this."
Ethics and Legal Status
Oxford professor Mariarosaria Taddeo's reaction: killing with AI deprives soldiers of dignity and removes the attacker's responsibility, and must be banned. "It is not just problematic — it is appalling."
UN Secretary-General Antonio Guterres has already clearly called for action: "Lethal autonomous weapons systems have no place in our world." But currently no official international ban exists on autonomous weapons that can kill without human intervention.
Notably, Ukraine's official stance is contradictory:
Why This Matters to the AI Industry
1. The "human-in-the-loop" principle is breached for the first time — AI governance communities have emphasized human-in-the-loop / human-on-the-loop principles for years; this disclosure challenges that principle in a battlefield context. 2. Technology mature vs. legal vacuum — the technology is ready; the bottleneck is law and ethics. Drone suppliers, the Ukrainian government, the UN, and academia show diverging policy intentions on this event. 3. It will accelerate international rulemaking on autonomous weapons — such disclosures will inevitably push the UN, the Red Cross, the Council on Foreign Relations, and similar institutions to accelerate new international norms. 4. This cannot be watched from the sidelines — on June 8, China's MIIT and SASAC issued a notice requiring humanoid robots and other key products to complete application validation and enter regular deployment by end of 2026; the boundary between civilian and military uses of embodied intelligence and autonomous systems will become increasingly sharp.
Takeaway for Practitioners
This is not a "distant future" scenario. It has already happened. It is a landmark event that AI governance practitioners, policy researchers, AI ethics scholars, and autonomous systems engineers must recalibrate their understanding around.