From Ars Technica · · 2m
AI drones learn to hunt targets on their own
NATO-backed startup deploys lean AI models that let drones identify and attack enemies without relying on distant servers.
In brief
NATO-backed startup deploys lean AI models that let drones identify and attack enemies without relying on distant servers. Small, efficient AI models running directly on drone hardware let militaries operate independently from centralized servers, a critical advantage when communications fail or infrastructure gets…
Swedish startup pivots to defense
Scaleout Systems, founded by Uppsala University researchers in 2018, originally deployed machine learning on commercial vehicles. Russia's 2022 invasion of Ukraine prompted the company to focus on military applications for NATO allies.
Small models, big advantage
Instead of using frontier AI from OpenAI or Anthropic, Scaleout harnesses leaner models that run computer vision tasks directly on drone hardware and forward-deployed devices without needing distant servers.
NATO accelerator program backs project
Scaleout joined NATO's Defence Innovator Accelerator in 2025 to work on the Federated Aerial Intelligence for Recon project, adapting machine learning for edge computing hardware in drones, tablets, and field command posts.
Models learn and improve in real time
AI models trained in one environment perform poorly in another. Scaleout's approach retrains models on battlefield data collected from multiple sources, releasing updated versions that improve throughout a single operation.
“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.”
Independence from centralized servers
Drones with onboard AI operate without relying on distant data centers, a critical advantage when communications fail, enemy jamming interferes, or centralized infrastructure gets destroyed.
Autonomous target detection demonstrated
During Sweden's Winter Demo 2026 event in January, a drone autonomously detected, identified, and geolocated threats using onboard AI. It prioritized the highest-value target and attacked without direct human commands.
“detect, identify and geolocate all potential spotted threats with the use of AI”
Federated learning scales across nations
In June, Scaleout tested its technology at a Swedish Air Force base, showing how local AI models can learn independently and then share updates with central nodes when reconnected.
NATO collaboration unlocked
This federated learning strategy allows decentralized AI networks to learn from aggregated data across geographic regions. It could eventually enable collaboration between NATO member states on shared military AI systems.
“In principle, you can unlock collaboration between NATO member states”
The headline
Small, efficient AI models running directly on drone hardware let militaries operate independently from centralized servers, a critical advantage when communications fail or infrastructure gets destroyed.
