Ukraine Harnesses Vast Military Data to Train AI Models — Reuters

UAV "Chaklun". Photo: chaklun.in.ua

As warfare increasingly incorporates artificial intelligence, Ukraine is leveraging a valuable asset: millions of hours of drone footage to train AI systems for battlefield decision-making.

Oleksandr Dmytriev, founder of OCHI, a Ukrainian non-profit that centralizes and analyzes video from over 15,000 frontline drones, revealed that since 2022, the system has amassed two million hours—equivalent to 228 years—of drone footage.

“This is AI fodder: give it two million hours of video, and it becomes something supernatural,” Dmytriev explained. The system processes an average of 5–6 terabytes of new combat footage daily, providing essential data for AI training on combat tactics, target identification, and weapons system optimization. Dmytriev added, “Essentially, this is experience turned into mathematics. AI can learn trajectories and angles for maximum weapon efficiency.”

OCHI has drawn interest from Ukraine’s foreign allies, with ongoing discussions about its potential applications.

Kateryna Bondar, a researcher at the Wadhwani Center for Artificial Intelligence, emphasized the importance of dataset size and image quality, as AI learns to identify targets by shape and color. However, Bondar noted that U.S. priorities lie in preparing AI systems for conflicts in the Pacific, focusing on potential confrontations with China rather than Europe’s terrain.

AI technologies are already in widespread use in the conflict. Thousands of drones utilize AI for autonomous targeting, and Ukraine employs similar systems for mine clearance and developing drone swarms. These swarms operate through interconnected cloud systems, executing commands collectively.

Meanwhile, Russia claims to use AI in combat as well, notably in Lancet attack drones for target recognition.

Read also: The Netherlands, partners invest over EUR 200M in Ukraine’s digital security