The system, developed with assistance from several NATO member countries' defence technology programmes, does not autonomously authorise strikes — a human operator makes the final decision in each case. What it does is compress the time between target identification and strike authorisation from an average of 22 minutes under the previous process to between 90 seconds and four minutes, and increase the volume of targets that can be processed simultaneously by a factor of approximately twelve.
The International Committee of the Red Cross has said the system raises "serious questions" about meaningful human control, on the grounds that a human operator reviewing an AI-generated target recommendation in 90 seconds cannot conduct the legal assessment required by international humanitarian law — proportionality, distinction between combatants and civilians, military necessity — in any substantive sense. Ukraine's military has contested this characterisation, arguing that the AI's classification process incorporates IHL criteria and that human review adds a layer of judgment rather than replacing one.
The debate matters beyond this conflict because the system's architecture will be replicated. Every military with the technical capacity to build it is studying Ukraine's experience in real time. The norms established by how this system is used — and whether any accountability follows from misuse — will shape the development of autonomous weapons systems for decades.
"The laws of war were written for humans making decisions at human speed," said ICRC legal adviser Cordula Droege. "We have not yet written the laws for humans ratifying decisions at machine speed. Those are not the same thing."