---
title: "NATO drones learn to strike without comms: Swedish onboard AI acts autonomously in EW zones"
description: "Swedish startup Scaleout Systems is testing onboard AI for NATO drones that can autonomously detect and engage targets without an operator's command, even under enemy communications jamming."
date: 2026-09-18T10:16:01.000Z
lang: en
url: https://xab.info/en/posts/nato-drones-autonomous-ai-swedish-startup-scaleout-systems
tags: [nato-drones, artificial-intelligence, scaleout-systems, sweden, autonomous-weapons, counter-electronic-warfare]
publisher: "XAB.info"
---

# NATO drones learn to strike without comms: Swedish onboard AI acts autonomously in EW zones

![NATO drone onboard AI: conceptual visualization of the Swedish autonomous computing module and sensors for operation in electronic warfare zones](https://xab.info/media/2026/09/19/drony-nato-avtonomnyy-ii-shvedskiy-startap-scaleout-systems/drony-nato-avtonomnyy-ii-shvedskiy-startap-scaleout-systems-1.webp)

## 🎯 Key Points

- Scaleout Systems adapted compact computer vision algorithms to the weak processors of drones and field command posts, abandoning bulky cloud-based models.
- Under the NATO DIANA program, the AI has been integrated into the low-cost suicide drone ALMA, developed jointly with BAE Systems Bofors.
- In January 2026 tests in Sweden, the drone autonomously detected threats and struck an armored engineering vehicle without an operator's command.
- June tests in Uppsala confirmed the ability to train the AI at field nodes without connectivity and to synchronize data afterward (federated learning).

In conditions where the adversary actively employs electronic warfare (EW) jamming, traditional drones that rely on a stable link with the operator lose their combat effectiveness. It is precisely this vulnerability that NATO is trying to close by testing the onboard artificial intelligence developed by Swedish startup Scaleout Systems. The system runs directly on the drone itself and enables it to operate autonomously — from target detection to delivering the strike.

### From trucks to defense

Scaleout Systems was founded by researchers from Uppsala University and initially developed artificial intelligence algorithms for freight transport. Following the start of Russia's full-scale invasion of Ukraine in 2022, the startup fully pivoted to the defense sector, adapting its computer vision expertise to the tasks of military reconnaissance and target engagement.

### Compact vision instead of the cloud

The key differentiator of Scaleout's solution is its rejection of bulky cloud-based models in the vein of OpenAI or Anthropic. The developers adapted compact computer vision algorithms to run on the weaker processors installed on drones, in tablets, and at field command posts. This allows the system to operate locally, without requiring a constant connection to external servers — a critical advantage within the reach of EW systems.

### Testing: from detection to autonomous strike

Under the NATO DIANA program, the developers integrated their AI into the ALMA project, a low-cost suicide drone being developed jointly with BAE Systems Bofors. According to January 2026 test results in Sweden, the drone independently detected, identified, and determined the coordinates of all potential threats in the observation zone. Based on predefined priorities, the algorithm selected the most valuable target — an armored engineering vehicle — and directed the explosive at it without a direct operator command.

### Federated learning and operation without comms

June tests at the airbase in Uppsala showed that field nodes are capable of continuing to train the AI even after losing connection with the central laboratory, and then synchronizing the accumulated data once the network is restored. In the long term, this latest "federated learning" system will enable instant exchange of intelligence data among all NATO member states, creating a common operational picture without a centralized node.

### Strategic significance

The combination of a low-cost carrier, onboard AI, and federated data exchange changes the logic of countering EW: instead of depending on a communications channel, the drone becomes an independent "eye and hand" on the battlefield. For the alliance, this is a step toward the mass deployment of autonomous strike systems capable of operating in conditions where traditional radio-channel control is impossible.

## 🔍 Fact-Check Verification

- [NATO drones to bypass EW: new AI will allow unmanned vehicles to operate fully autonomously](https://www.rbc.ua/ukr/news/droni-nato-obiydut-reb-noviy-shi-dozvolit-1789721106.html) - Заголовок напрямую подтверждает тезис об автономном действии дронов НАТО в условиях РЭБ за счет нового ИИ.
- [Swedish startup develops small AI model for autonomous target recognition and drone strikes](https://24tv.ua/tech/ru/iskusstvennyj-intellekt-dlja-dronov-pozvoljaet-avtonomno-provodit-ataki-protiv-celej_n3144506) - Подтверждает шведское происхождение стартапа, малую ИИ-модель и функцию автономного распознавания и атаки целей.
- [Drones, EW, AI and more: Ukraine and Lithuania launch defense support initiative ...](https://www.obozrevatel.com/politics-news/bespilotniki-reb-ii-i-ne-tolko-ukraina-i-litva-zapuskayut-initsiativu-podderzhki-oboronnyih-innovatsij.htm) - Даёт контекст более широкой инициативы поддержки оборонных инноваций (Украина, Литва, дроны, РЭБ, ИИ); не противоречит базовому нарративу, но раскрывает смежный, а не тот же сюжет.
- [NATO plans to deploy AI-powered drones](https://www.vedomosti.ru/politics/news/2026/08/14/1221208-nato-bespilotniki-ii) - Подтверждает намерение НАТО внедрять беспилотники с ИИ (материал от 14.08.2026).

## ❓ FAQ

### Q: What does Scaleout Systems' AI do on board the drone?
**A:** Compact computer vision algorithms run directly on the drone and enable it to autonomously detect, identify, and determine the coordinates of targets, and — based on predefined priorities — to engage the selected target without a direct operator command.

### Q: Why is this important under communications jamming (EW)?
**A:** Because the system runs locally on the drone's weak processor rather than through the cloud, it does not depend on a stable radio link with the operator and retains its combat effectiveness within an electronic warfare jamming zone.

### Q: In which NATO project is this system being tested?
**A:** Under the NATO DIANA program, the AI has been integrated into the ALMA project, a low-cost suicide drone being developed jointly with BAE Systems Bofors.

### Q: What is "federated learning" in this context?
**A:** It is a mechanism in which field nodes continue to train the AI even without a connection to the central laboratory, and then synchronize new data once the network is restored; in the long term, this will enable the exchange of intelligence data among NATO member states.