A NATO-backed startup is adapting artificial intelligence so that small military drones can detect, identify and attack targets without direct human commands, according to a detailed report by Ars Technica. The company, Scaleout Systems, demonstrated an AI-enabled loitering munition that autonomously prioritized what it deemed the highest-value target on a battlefield and flew to it to drop an explosive — a sign of how quickly autonomous targeting is moving from research papers into European defense programs.
Andreas Hellander, cofounder and CEO of Scaleout Systems, told Ars that the technology is designed to give NATO allies an edge by operationalizing sensor data at the edge of the battlefield. For more context on this story, see our ongoing latest AI developments.
From Truck Telemetry to the Battlefield
Scaleout Systems was founded in 2018 by researchers from Uppsala University in Sweden, and its original focus was far from the arms industry: training and deploying machine learning models directly on hardware inside commercial trucks and other vehicles. That changed after Russia launched its full-scale invasion of Ukraine in 2022, when the company pivoted toward defense applications.
"With the war in Ukraine and a shifting world, we realized that this technology can be very important to operationalize edge data and sensor data for machine learning to make sure that NATO allies have found that strategic advantage," Hellander told Ars.
The company was selected in 2025 to join NATO's Defence Innovator Accelerator for the North Atlantic (DIANA) Challenge Program, where it has worked on a project called Federated Aerial Intelligence for Recon. The effort adapts machine learning models for the edge computing hardware found on drones, drone pilot tablets, and field command posts — hardware that, as Hellander told Ars, ranges from small embedded devices to powerful edge workstations.
Federated Learning at the Front Line
Scaleout's approach differs from the frontier-model playbook of OpenAI or Anthropic. Instead of large centralized models, the company uses leaner machine learning models that run computer vision tasks directly on drone hardware. The drones help operators with target identification while their cameras and sensors collect battlefield data, then intermittently share selective updates with computing nodes at local platoon or company headquarters — without transmitting sensitive raw data.
Those headquarters nodes retrain the models on aggregated data from multiple sources and push updated capabilities back out to edge devices when conditions allow. The benefit is adaptability: "Models might have been trained in a desert environment, and if we try to deploy them in an urban environment, they're not going to perform well," Hellander told Ars. He argued that models which keep learning and improving from battlefield sensor data during a single operation represent a "sustainable advantage."
In June, Scaleout tested the concept at a Swedish Air Force base in Uppsala, where the Swedish military already holds a license for the company's software platform. The demonstration showed that a forward-deployed computing node could keep running its own AI inference and active-learning processes even after losing contact with a central node in the company's lab, syncing its learned updates once the connection returned. Hellander told Ars this federated strategy could eventually scale across entire regions, noting, "In principle, you can unlock collaboration between NATO member states."
ALMA: An Autonomous Loitering Munition
The most consequential application is Scaleout's participation in the Affordable Loitering Modular Ammunition (ALMA) project, headed by BAE Systems Bofors, which aims to develop a low-cost autonomous kamikaze drone. The ALMA concept was first publicly demonstrated at a Winter Demo 2026 event in Sweden in January.
During that demonstration, a drone autonomously "detect[ed], identif[ied] and geolocate[d] all potential spotted threats with the use of AI," according to a Scaleout Systems presentation cited by Ars. "All data is handled by dedicated onboard computing, allowing the system to perform in real-time without any external processing."
The drone then automatically prioritized the highest-value target as defined by its mission — in this case, an armored engineering vehicle — flew to it, and dropped an explosive. Ars reported that a human operator could still control and direct the drone, but that the drone carried out the mission on its own without direct human commands. Hellander told Ars the company has also built a functioning surveillance and reconnaissance concept demonstrated publicly in Sweden.
Why Edge AI Matters in Modern War
The push toward onboard AI reflects the realities of current conflicts. Electronic warfare and enemy jamming frequently interfere with communications, which is one reason a growing number of Ukrainian military drones already incorporate onboard AI into cheap kamikaze designs. Ars also noted that relying on centralized data centers to run AI models looks riskier at a time when large data centers have been targeted and destroyed during the war between the US and Iran.
Local AI models that operate independently — without continuous contact with a distant server — keep working when links go down, which is precisely the scenario military planners worry about most.
The Wider Debate Over Autonomous Weapons
The ALMA demonstration lands in the middle of an unresolved international debate over lethal autonomous weapons, where the key question is how much human oversight must remain in the loop when machines select and engage targets. In Scaleout's public demonstrations, a human operator retained the ability to control and direct the drone, and the company frames its technology as decision support for NATO militaries. What the January demo showed, though, is that the full mission cycle — detect, identify, prioritize, strike — can now run on a low-cost drone's onboard computing without external processing.
As Ars Technica's report makes clear, the building blocks of autonomous attack capability are no longer confined to major powers' classified programs. They are being assembled by startups, accelerated through NATO innovation channels, and demonstrated in public — a development that arms-control advocates and military planners alike will be watching closely.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
