Battlefield AI is being designed to keep working when connectivity fails

Battlefield AI systems are increasingly being designed to continue operating when communications or central computing facilities become unavailable. The challenge affects drones, sensors, AI-assisted targeting and decision-making systems that depend on reliable compute and connectivity.

Andreas Hellander, CEO and co-founder of Scaleout, said resilient systems should combine central coordination with bounded local capability. A disconnected node can continue local inference using its last approved model, while caching detections and telemetry until a connection returns. If an aggregation point becomes unavailable, clients can be reassigned and training can continue with reachable systems.

Centralised infrastructure provides shared compute, operational coordination and places to enforce security, model approval and command policy. It can also concentrate risk when facilities or links are disrupted. Distributed systems reduce latency and allow local functions to continue, but may rely on partial or outdated information.

Hellander described a hybrid design as the usual goal. Retraining should remain controlled and separately approved, while platforms use pre-approved failsafes after losing a link. Federated learning can reduce the movement of sensitive footage by exchanging protected model updates, but those updates still require security, validation and auditing.

Scaleout has worked with NATO, most recently the Swedish Air Force, and defense companies including BAE Systems. The company uses commercially available edge hardware such as NVIDIA Jetson modules and ruggedised GPU servers.