Scout AI, a defense-focused startup founded in 2024 by Coby Adcock and Collin Otis, announced Wednesday that it has raised $100 million in Series A funding to develop artificial intelligence models for military operations. The round, led by Align Ventures and Draper Associates, follows a $15 million seed round in January 2025. The company invited TechCrunch for an exclusive visit to its training operations at a U.S. military base in central California, which it asked not to be named.
At the base, four-seater all-terrain vehicles roam hillside trails in a training exercise designed not for human drivers, but to teach AI how to operate in conflict zones. The company calls itself a “frontier lab for defense” and is building an AI model named “Fury” to control military assets, initially for logistics but eventually for autonomous weapons.
Also read: Thinking Machines Lab unveils AI that listens while it talks, mimicking natural conversation
Training AI like a soldier
Scout AI’s approach draws on existing large language models, but the company adapts them for military environments. CTO Collin Otis, a former executive at autonomous trucking company Kodiak, compares the process to training human soldiers.
“They start when they’re 18 years old, and sometimes they even start after college, so you want to start with that base level of intelligence,” Otis told TechCrunch. “It’s useful to start with someone who’s already made an investment and then say, hey, what do I have to do to teach this thing to be an incredible military AGI, versus just being a broadly intelligent AGI?”
Also read: GM lays off 600 IT workers in deliberate shift toward AI-native talent
Scout has secured $11 million in military technology development contracts from organizations including DARPA, the Army Applications Laboratory, and other Department of Defense customers. The company is one of 20 autonomy firms whose technology is being used by the U.S. Army’s 1st Cavalry Division during training at Fort Hood, Texas. Products that prove themselves may be deployed with the unit when it next deploys in 2027.
Why off-road autonomy is harder than city driving
While autonomous cars operate in structured environments with clear rules, off-road navigation presents fundamentally different challenges. Unmarked trails, loose sand, steep hills, and confusing intersections require a level of general intelligence that traditional autonomy systems lack.
Otis said he was motivated to start Scout after realizing the system he helped build at Kodiak was not intelligent enough to operate in an unpredictable war zone. The company is turning to Vision Language Action models, or VLAs, a technology first released by Google DeepMind in 2023. VLAs are based on LLMs and are used to control robots. They have seeded robotics startups like Physical Intelligence and Figure.AI, the humanoid robot company led by Adcock’s brother, Brett.
“If I handed you the controller of a drone right now and I strapped a headset on you, you could learn to fly that thing in minutes,” Otis said. “You’re actually just learning how to connect your prior knowledge to these couple little joysticks. It’s not a big leap. That’s the way to think about VLAs and why they’re such an unlock.”
Inside the bootcamp: driving and learning
TechCrunch’s reporter drove one of Scout’s ATVs on the rutty trails, finding the terrain challenging with steep hills, loose sand on turns, and disappearing tracks. The experience illustrated the kind of general intelligence the company wants in its models, which have been trained via these ATVs for six weeks after initial work on civilian vehicles.
Riding in the ATV under autonomous control revealed differences from human driving: it accelerates faster, hugs the right on wider trails but stays in the middle of narrow ones, and occasionally slows down to think over its next move. These behaviors reflect training data from human drivers.
Stuart Young, a former DARPA program manager who worked on ground vehicle autonomy, said the technology is “good enough to be doing that experimentation in the field with soldiers to figure out how to most be effective to US forces.” Young left DARPA this month to join Field AI after managing a program called RACER, which helped seed the autonomous off-road vehicle space.
From logistics to lethal applications
The first applications of ground autonomy, according to Scout executives and military technologists, will be automated resupply: carrying water or ammunition to distant observation posts, or in convoys where a crewed truck is followed by autonomous vehicles. Brian Mathwich, an active duty infantry officer serving as a military fellow at Scout, recalled a recent exercise in Alaska where he led a resupply convoy in total darkness and wished for autonomous vehicles.
Scout sees itself primarily as a software company, building an intelligence layer for military machines rather than manufacturing vehicles. Its first product, called “Ox,” is command and control software bundled on hardened computer hardware that allows soldiers to orchestrate multiple drones and ground vehicles with prompt-like commands such as: “Go to this waypoint and watch for enemy forces.”
Scout is also developing drone systems for reconnaissance and as weapons, using vision language models to give them intelligence. In one mission concept, groups of munition drones would fly with a larger “quarterback” platform providing compute resources. The drones would search a geographic area for hidden enemy tanks and attack them, possibly without human intervention.
The autonomous weapons debate
Autonomous weapons remain a flash point in defense technology politics, but experts note the concept is not new: heat-seeking missiles and mines have been in use for decades. Jay Adams, a retired U.S. Army Captain who leads Scout’s operations team, said the question is how weapons are controlled. Scout’s munition drones can be programmed to only attack threats in a specific geographic area, or only with human confirmation. Adams also noted that autonomous weapons platforms are unlikely to fire because they are scared, the way a young soldier might be.
VLAs offer promise for better targeting. Lt. Col Nick Rinaldi, who supervises Scout’s work for the Army Applications Laboratory, said that while automated targeting is hard and unlikely to be used outside constrained environments in the near term, the potential of VLAs to reason about threats makes them a promising technology to investigate.
Adams said the promise of drones that can identify their own targets is key to future warfare. While Russia’s invasion of Ukraine has generated intense interest in drone warfare, he believes having humans operate individual UAVs does not scale enough for the U.S. to face a large number of low-cost unmanned systems.
Funding, compute costs, and the path to AGI
Scout is using existing LLMs as the base for its agents, though it declined to name which ones. Otis said the company has agreements with “very well known hyperscalers” to provide pretrained intelligence for its foundation model. He declined to comment on whether Scout uses open-weight models from Chinese companies.
Scout expects to build its own model from the ground up in the years ahead, and the founders say much of its capital will go into training and compute costs. Otis suggested Scout may beat existing leaders to AGI because its model will constantly interact with the real world.
“There’s an argument in the AGI community along the lines that you can only get so intelligent by reading the internet, and most intelligence comes with interacting in the world,” Otis said.
When asked whether Adcock is competing with his brother’s humanoid robots at Figure, Otis said no, but added: “We can get to scale much faster because our customer has assets,” referring to the Pentagon.
Conclusion
Scout AI’s $100 million raise and exclusive access to military training facilities underscore the accelerating convergence of AI and defense. The company’s use of Vision Language Action models for off-road autonomy represents a significant technical bet, and its contracts with DARPA and the Army suggest institutional interest. However, questions about autonomous weapons, the reliability of VLA-based systems in combat, and the ethics of delegating lethal decisions to machines remain unresolved. Scout’s progress over the next year, particularly as the 1st Cavalry Division evaluates its technology, will offer the clearest signal of whether this approach can deliver on its promise.
FAQs
Q1: What is Scout AI and what does it do?
Scout AI is a defense startup founded in 2024 that develops artificial intelligence models for military vehicles and drones. Its Fury model is designed to operate autonomous ground vehicles and aerial drones for logistics, reconnaissance, and potentially weapons deployment.
Q2: How much funding has Scout AI raised and who invested?
Scout AI raised a $100 million Series A round led by Align Ventures and Draper Associates, following a $15 million seed round in January 2025. The company has also secured $11 million in military technology development contracts from DARPA, the Army Applications Laboratory, and other DoD customers.
Q3: What technology does Scout use for its autonomous systems?
Scout uses Vision Language Action models, or VLAs, which are based on large language models and adapted for robot control. The company trains its models using real-world driving data from ATVs and simulation, combined with reinforcement learning.
Q4: Will Scout’s technology be used for autonomous weapons?
Yes, Scout is developing systems that could allow drones to identify and attack targets without human intervention. However, the company says its weapons can be constrained to specific geographic areas or require human confirmation. The first applications are expected to be logistics and resupply missions.
Q5: How is Scout different from other autonomous vehicle companies?
Scout focuses specifically on military off-road environments, which present greater challenges than structured urban settings. The company also builds its technology as a software layer for existing military vehicles rather than manufacturing its own hardware.
Be the first to comment