“We try anything that can kill more Russians.” New Ukraine AI drones require just 30-min training

Ukrainian forces have cut drone operator training using AI-guided systems that have boosted target accuracy up to 80%, a recent CSIS report finds.
"We try anything that can kill more Russians." New Ukraine AI drones require just 30-min training
A Ukrainian soldier with a drone on the frontline. Photo: General Staff via Facebook
“We try anything that can kill more Russians.” New Ukraine AI drones require just 30-min training

Ukraine has successfully enhanced its drone capabilities by retraining publicly available AI models on real-world combat data, increasing the odds of hitting Russian targets “three- or four-fold,” according to a new report from the Center for Strategic and International Studies (CSIS).

“By removing the need for constant manual control and stable communications … drones enabled with autonomous navigation raise the target engagement success rate from 10-20% to around 70-80%,” writes Ukrainian-American scholar Kateryna Bondar, a former advisor to Kyiv, in the report released on 6 March.

The AI systems allow drones to autonomously navigate the final 100 to 1,000 meters to their targets—a significant advancement that emerged after Russia tried and apparently abandoned similar capabilities on its widely used Lancet drones just 13 months ago. Starting this fall, reports of AI-guided Ukrainian drones began surfacing, marking a major technological leap for a country fighting continuous bombardment.

“We’re very far from killer robots,” Bondar told Breaking Defense in an exclusive interview. But in contrast to the more cautious bureaucracy of the West, she said, “the Ukrainians are more open to testing and trying anything and everything that can kill more Russians.”

AI gains autonomy, but humans stay in control

Ukraine’s success hinges on creating specialized datasets tailored to specific sectors of the front line and drone types.

“The frontline is very long … and the situation is very different on different parts of the front,” Bondar explained.

She also noted that conditions vary dramatically from Bakhmut’s static trenches to areas where Russian forces continue to advance with “wave after wave” of troops.

This targeted approach to data training is paired with rapid operator preparation. Training for these autonomous systems now takes as little as 30 minutes to one day, compared to the extensive flight hours previously required. This accelerated learning curve has dramatically expanded the pool of qualified operators and overall operational readiness.

Ukraine preserves its technological advantage through robust encryption of onboard AI software, making it difficult for adversaries to reverse engineer. While hardware designs can be replicated in weeks, sophisticated encryption significantly delays the development of equivalent systems, establishing a strategic deterrent against immediate duplication.

Despite increasing autonomy, human oversight remains central—particularly for engagement decisions. The current human-in-the-loop approach ensures operators can override autonomous functions, keeping critical ethical and strategic judgments under human control while laying the groundwork for broader autonomy in the future.

Ukraine ditched mega-platforms for targeted drone tech

A key insight from Ukraine’s approach is their focus on numerous small, specialized AI systems rather than ambitious mega-projects.

“They also started with this idea of creating a huge mega-platform which would cover everything and do everything,” Bondar said. “They slowly realized … the current level of AI development allows you to train models on very small datasets.”

She emphasized the practical advantages: “These small models are easier to train and update. They’re way cheaper.”

To address the initial “zoo” of technologies where soldiers sometimes used “10 different software systems” simultaneously, the Ukrainian government created “one universal military dataset” with standardized tags and categories housed in a government-run system called Delta.

Despite comprising less than 0.5 percent of Ukraine’s current drone fleet—approximately 10,000 out of nearly 2 million contracted in 2024—these AI-guided systems have proven so effective that Kyiv plans to have AI guidance in roughly half of all drones purchased in 2025.

If Ukraine achieves this goal, the resulting deployment could represent a twelve-fold increase in combat effectiveness—potentially decisive in a war where drones have replaced artillery as the main cause of casualties.

Read more: 

You could close this page. Or you could join our community and help us produce more materials like this.  We keep our reporting open and accessible to everyone because we believe in the power of free information. This is why our small, cost-effective team depends on the support of readers like you to bring deliver timely news, quality analysis, and on-the-ground reports about Russia's war against Ukraine and Ukraine's struggle to build a democratic society. A little bit goes a long way: for as little as the cost of one cup of coffee a month, you can help build bridges between Ukraine and the rest of the world, plus become a co-creator and vote for topics we should cover next. Become a patron or see other ways to support. Become a Patron!

To suggest a correction or clarification, write to us here

You can also highlight the text and press Ctrl + Enter

Please leave your suggestions or corrections here



    Euromaidan Press

    We are an independent media outlet that relies solely on advertising revenue to sustain itself. We do not endorse or promote any products or services for financial gain. Therefore, we kindly ask for your support by disabling your ad blocker. Your assistance helps us continue providing quality content. Thank you!