Coby Adcock’s Scout AI raises $100 million to train its models for war. We visited its bootcamp.

At a military facility in central California, all-terrain vehicles designed for rugged environments are being put through their paces, but this exercise is not for human operators; it’s part of an initiative to refine AI models for operations in conflict zones. Scout AI, a startup established in 2024 by Coby Adcock and Collin Otis, describes itself as a "frontier lab for defense." Recently, the company announced it has successfully raised $100 million in a Series A funding round, led by Align Ventures and Draper Associates, following a $15 million seed round earlier this year. Scout invited TechCrunch to get an exclusive look at its training operations at an undisclosed military base. The company is in the process of developing an AI model known as "Fury," intended to manage and operate military resources, initially for logistical purposes and eventually for autonomous weaponry. Collin Otis, the company's CTO, draws parallels between training AI and preparing soldiers, highlighting the importance of starting with a foundational level of intelligence. Scout has secured contracts for military technology development amounting to $11 million from various organizations, including DARPA and the Army Applications Laboratory. Among a select group of 20 autonomy firms, Scout’s technology is being integrated into the training protocols of the US Army’s 1st Cavalry Division, which is set to deploy again in 2027, with expectations that successful innovations will accompany them. During internal testing, Scout’s team is rigorously evaluating the vehicles on the challenging terrains of the base. With a staff that includes former military personnel, the team is conducting simulated missions that put the autonomous vehicles to the test in unpredictable environments. While autonomous vehicles are becoming more common in urban areas, operating off-road presents a unique set of challenges. Otis, who previously worked with autonomous trucking ventures, founded Scout after realizing that existing systems lacked the necessary intelligence for unpredictable combat scenarios. The startup is leveraging advanced Vision Language Action (VLA) models, which are based on large language models (LLMs), to enhance robotic control. The technology, introduced by Google DeepMind in 2023, has already sparked innovation among robotics startups. Adcock believes this approach can significantly enhance the military's fleet of autonomous vehicles. He and Otis began collaborating on applying cutting-edge AI for defense solutions, inspired by advancements in robotics. The aim is to develop models that possess general intelligence, capable of navigating complex terrains and performing tasks efficiently. In a recent demonstration, TechCrunch experienced a ride in one of Scout's ATVs, noting its capability to handle steep inclines and rugged paths with ease. The autonomous mode proved to be faster and more precise than a human operator, showcasing the potential of these vehicles to replicate human-like decision-making under pressure. While the VLA technology is still in its infancy, experts believe it holds promise for future applications in military operations, particularly for automating resupply tasks. Scout envisions a future where autonomous vehicles can transport essential supplies, freeing human soldiers to focus on more critical missions. Adcock and his team are primarily focused on software development, intending to create an intelligence layer for military machinery rather than manufacturing the vehicles themselves. The company anticipates that its first major product, named "Ox," will facilitate command and control operations for soldiers managing both drones and ground vehicles. Training on real vehicles is essential for the success of this software, which is being tested in a controlled training environment known as Foundry. Additionally, Scout is experimenting with drones for reconnaissance and potential offensive operations, aiming to create a network of autonomous munitions that can independently identify and engage targets. As the landscape of warfare evolves, the integration of AI and autonomous technologies in military applications raises critical discussions about ethics and operational control. Scout’s initiatives highlight the ongoing challenges and opportunities in developing intelligent systems for defense, with the aim of enhancing the capabilities of US forces in future conflicts.