Almost one year ago, we invested in the first round of Scout AI, a defense technology company advancing ground autonomy in military applications. Ground vehicles remain mostly human-controlled and vulnerable. Scout aims to change this by enabling them to understand their surroundings, follow natural language commands, and make on-the-spot decisions, even without GPS or direct communication (for a demonstration of Scout click here).
In a recent California military base demo, Scout’s Fury Orchestrator, a 100B+ parameter AI model, controlled a self-driving off-road vehicle and two lethal drones that located and destroyed a hidden truck, showcasing end-to-end autonomy from natural-language orders. Using AI instead of fragile LiDAR systems, which are the norm for autonomous vehicles but vulnerable to weather conditions and irregular terrain, Scout works with a range of hardware and cameras to build more reliable, scalable battlefield robots, training a VLA (Visual-Language-Action) model in order to transform uncrewed platforms into intelligent and autonomous agents. With 4 existing DoD contracts, U.S. Army and DARPA support, and a pending government contract bid focused on multi-drone coordination, Scout is advancing next-generation defense autonomy.
Building on Scout’s progress, it is clear that the defense landscape is evolving rapidly. To illustrate this, I wanted to share some highlights from a comprehensive article in the MIT Technology Review, focused on how the rules for military action are being rewritten.
NATO’s Hedgehog Exercise
Last spring, 3,000 British soldiers of the 4th Light Brigade, the “Black Rats”, deployed to eastern Estonia for NATO’s Hedgehog exercise, supported by Project ASGARD, an automated intelligence network assembled in just four months. When a reconnaissance drone identified enemy armor, it sent images and locations directly to artillery or drones, bypassing the traditional command chain. Soldiers used Samsung smartphones to select strike options, enabling near-instant, decisive action. As a result, the detection-to-decision kill chain is now under a minute, increasing lethality tenfold and demonstrating how AI and drones are shaping the battlefield.
The system deployed during NATO’s exercise codenamed “Hedgehog” changed the pace of battlefield decision-making. This fast, sensor-driven decision chain enabled by AI compresses the time from detection to strike and simplifies human interfaces.
The Human Bottleneck
However, the biggest bottleneck is not technology but people. A million drones are great, but we are going to need a million people. Today, governments insist humans remain in the loop for lethal decisions. Helsing, (valued at ~$12b), is developing drones that use object recognition with operator approval and operate autonomously only in the final terminal guidance phase. Research and development is pushing one-to-many control, with a single operator supervising more than 10 drones, thereby reducing granular oversight.
Helsing’s full vision spans multiple domains. It includes Loft Orbital satellites for worldwide asset classification, HF-1 and HX-2 drones that combine reconnaissance and missile functions, mini-subs drones that dive to 3,000 feet for 90 days autonomously, and Europa, a 4.5-ton pilotless fighter jet carrying hundreds of pounds of weaponry into defended airspace under remote control. Altra software orchestrates saturation attacks, coordinating cross-border barrages of missiles, drones, and artillery to breach defenses. The U.S. Navy is pursuing similar capabilities for Taiwan. Governments know that further automation risks a race with no winners, yet manpower shortages leave few alternatives.
Scale, Autonomy, and the Path Forward
European officials recognize this challenge. As stated by Sven Weizenegger, CEO of the German Cyber Innovation Hub, “If you don’t have the people, then you can’t control so many drones. Therefore you need swarming technologies in place, autonomous systems.” Helsing’s distributed factories, with one producing 1,000 drones per month, support wartime scaling. Germany’s €300m order for 12,000 HX-2s is a test case, with some suggesting stockpiles of 100,000 to 200,000 drones per country as a deterrent.
Final words
Autonomous systems, artificial intelligence, and the networks that connect them represent the future of modern warfare. In an increasingly volatile geopolitical environment, maintaining technological leadership in defense is no longer optional—it is essential.