NATO-backed startup adapts edge AI for autonomous attack drones

Scaleout Systems, a NATO-backed startup originally founded in 2018 by Uppsala University researchers, transitioned its focus from commercial vehicles to defense applications following Russia's 2022 invasion of Ukraine. [1] The company provides decentralized machine learning technology designed to operate on drones and military bases. [1] Scaleout's target recognition and engagement capabilities were recently demonstrated on an autonomous attack drone as part of the BAE Systems Bofors-led ALMA project. [1]

Instead of relying on large frontier AI models from developers like OpenAI or Anthropic, Scaleout uses leaner computer vision models. [1] CEO Andreas Hellander stated that these machine learning models are explicitly designed to fit on forward-deployed edge hardware. [1] This deployment hardware ranges from small embedded devices directly on drones to more powerful edge workstations located at forward bases. [1] The company was also recently selected to participate in NATO's Defence Innovator Accelerator program. [1]

The deployment of federated learning for autonomous drone strikes indicates a shift toward localized edge computing in battlefield environments where reliance on cloud-based frontier models is impractical. Scaleout’s pivot highlights how the ongoing war in Ukraine is accelerating European defense integration with decentralized AI systems. By focusing on smaller computer vision models that can run on constrained hardware, the ALMA project demonstrates an alternative to the massive parameters typically required for advanced object recognition by companies like OpenAI and Anthropic. What the sources don't address: How much human oversight remains in the decision loop before an autonomous drone is permitted to engage a detected target.