Scout AI Secures $100M to Build War-Ready AI: Inside the Bootcamp

At a US military base in central California, four-seater all-terrain vehicles navigate hillside trails. This isn't a typical training exercise for personnel; it's a dedicated effort to train advanced AI models for deployment in complex conflict zones. This initiative is spearheaded by Scout AI, a burgeoning startup founded in 2024 by Coby Adcock and Collin Otis, positioning itself as a “frontier lab for defense.” The company recently announced a significant $100 million Series A funding round, led by Align Ventures and Draper Associates, following a $15 million seed round in January 2025. This substantial investment underscores the growing importance of AI in modern military applications.

The Rise of Autonomous Military AI

Scout AI granted GearTech an exclusive tour of its training operations at a military base, the location of which remains confidential. The core of their work revolves around developing an AI model, internally dubbed “Fury,” designed to operate and command military assets. Initially focused on logistical support, the long-term vision extends to fully autonomous weapons systems. CTO Collin Otis draws a compelling parallel between this process and traditional soldier training.

“They start when they’re 18 years old, and sometimes even after college, so you want to start with that base level of intelligence,” Otis explained to GearTech. “It’s useful to start with someone who’s already made an investment in learning and then focus on teaching this AI to become an incredible military AGI, rather than just a broadly intelligent AGI?” This approach highlights the need for specialized AI tailored to the unique demands of the battlefield.

Securing Contracts and Military Integration

Scout AI has already secured military technology development contracts totaling $11 million from key organizations like DARPA, the Army Applications Laboratory, and other Department of Defense entities. The company is also one of 20 autonomy companies participating in the US Army’s 1st Cavalry Division’s training cycle at Ft. Hood in Texas. The expectation is that successful technologies will be deployed when the unit next deploys in 2027, demonstrating a clear path to real-world application. This integration signifies a growing trust in AI-driven solutions within the military.

Testing Ground: The California Military Base

The practical testing of Scout AI’s technology takes place on the hilly terrain of the military base. A dedicated operations team, comprised of former soldiers, rigorously puts the autonomous ATVs through simulated missions. This hands-on approach ensures the AI is tested under realistic conditions.

While autonomous vehicles are becoming increasingly common in urban environments, operating in unstructured, off-road settings presents a significantly greater challenge. Otis, previously an executive at autonomous trucking company Kodiak, founded Scout AI after realizing that existing systems lacked the intelligence required to navigate the unpredictable nature of a war zone. The need for robust, adaptable AI became strikingly clear.

An autonomous ground vehicle controlled by Scout AI’s Fury model. Image Credit: Scout Ai / Scout AI

A New Approach to Autonomy: Vision Language Action Models

Scout AI is leveraging a cutting-edge autonomy technology: Vision Language Action (VLA) models. These models, built upon Large Language Models (LLMs), are designed to control robots. First introduced by Google DeepMind in 2023, VLAs have quickly gained traction in the robotics startup space, influencing companies like Physical Intelligence and Figure.AI, the latter led by Adcock’s brother, Brett.

Adcock, a board member at Figure, believes that the success of Figure has demonstrated the potential of bringing broader intelligence to the military’s growing fleet of autonomous vehicles. He connected with Otis, who was advising Figure, and together they began applying the latest advancements in AI to military solutions. This collaboration highlights the synergy between civilian robotics innovation and defense applications.

The Power of VLAs: Bridging the Gap Between Intelligence and Action

“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 explained. “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.” This analogy illustrates the efficiency and adaptability of VLA models.

GearTech had the opportunity to drive one of Scout AI’s ATVs on the challenging terrain, experiencing firsthand the steep hills, loose sand, and confusing intersections. Even with limited ATV driving experience, the task was manageable. This demonstrates the level of general intelligence Scout AI aims to instill in its models. The company has been training these models using ATVs for just six weeks, building upon initial training with civilian vehicles.

Riding in the autonomously controlled ATV revealed a noticeable difference – faster acceleration and a more assertive navigation style. The operations team pointed out how the vehicles hug the right on wider trails but stay centered on narrow ones, mirroring the behavior of their human trainers. The vehicle also exhibited a cautious pause when encountering ambiguous situations, slowing down to assess its next move during a 6.5 km loop.

While VLAs are still relatively new and haven’t been deployed in operational settings, Stuart Young, a former DARPA program manager specializing in ground vehicle autonomy, believes “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.” Like other autonomy companies, Scout AI’s full autonomy stack incorporates deterministic systems and other AI flavors to enhance its agents’ capabilities.

Young recently left DARPA to join Field, having managed the RACER program, which aimed to create high-speed, autonomous off-road vehicles – a program that helped seed this space, similar to how DARPA’s Grand Challenge propelled the development of self-driving cars. Two companies spun out of RACER, Field AI and Overland AI, and Scout AI also participated as a later addition.

Initial Applications: Automated Resupply and Logistics

According to Scout AI executives and military technologists, the first practical applications of ground autonomy will likely be in automated resupply operations. This includes tasks like carrying water or ammunition to remote observation posts, or following crewed trucks in convoys with six to ten autonomous vehicles, freeing up human labor for more critical tasks. Brian Mathwich, an active duty infantry officer currently working at Scout AI, recalled a recent exercise in Alaska where he led a resupply convoy in complete darkness and wished for the assistance of autonomous vehicles.

Image Credit: Scout AI / Scout AI

Adding Intelligence to the Army’s Motorpool

Scout AI positions itself primarily as a software company, focusing on building an intelligence layer for existing military machines. The company doesn’t intend to manufacture the autonomous vehicles themselves but rather to enhance their capabilities.

Adcock anticipates that the startup’s first widely adopted product will be “Ox,” its command and control software, packaged with ruggedized computer hardware (GPUs, communications, cameras). This software will allow individual soldiers to orchestrate multiple drones and autonomous ground vehicles using simple, prompt-like commands: “Go to this waypoint and watch for enemy forces.”

The company’s training range, “Foundry,” located at the military base, is crucial for refining this software. Drivers spend eight-hour shifts operating the ATVs, and the system logs instances where human intervention is required. This data is then used to improve the AI model. The base commander has even authorized the ATVs to participate in security patrols.

Scout AI hypothesizes that VLAs, combined with limited real-world data and simulation training, can deliver a fully capable driving agent. While the vehicle performs well on trails, it’s not yet ready for fully off-road operation.

Expanding to Drones and Munitions

Scout AI is also experimenting with drones for reconnaissance and as potential weapons platforms, equipping them with intelligence powered by vision language models. The company is developing a system where groups of munition drones would operate under the command of a larger “quarterback” platform, providing increased computing power. In a potential mission scenario, these drones would search a designated area for hidden enemy tanks and engage them, potentially without human intervention. Otis argues that this approach could be more precise than traditional indirect artillery fire.

While autonomous weapons are a contentious topic in defense technology, experts point out that the concept isn’t new – heat-seeking missiles and mines have been in use for decades. The key question, according to Jay Adams, a retired U.S. Army Captain leading Scout AI’s operations team, is how these weapons are controlled.

Adams notes that the company’s munitions drones can be programmed to only attack targets within a specific geographic area or require human confirmation before engaging. He also suggests that autonomous weapons platforms are less likely to fire due to fear, unlike a young soldier.

VLAs also offer the potential for improved targeting accuracy. Scout AI claims its models are pre-trained on military data to prepare them for scenarios like encountering an enemy tank during a resupply mission. Lt. Col Nick Rinaldi, who oversees Scout AI’s work for the Army Applications Laboratory, acknowledges that automated targeting is challenging and unlikely to be widely used outside of controlled environments in the near future, but believes the reasoning capabilities of VLAs make them a promising technology to explore.

Adams emphasizes that drones capable of independently identifying targets are crucial for future warfare. While Russia’s invasion of Ukraine has sparked interest in drone warfare, he believes relying on humans to operate individual UAVs isn’t scalable enough to counter a large number of low-cost unmanned systems.

A Mission to Embrace Military Collaboration

Image Credit: Scout AI / Scout AI

Like many defense startups, Scout AI openly embraces its mission and is critical of companies hesitant to collaborate with the government. Google, for example, reportedly withdrew from a Pentagon contest to develop control systems for autonomous drone swarms, a capability Scout AI is also pursuing.

“The AI people don’t want to work with the military,” Otis told GearTech, referencing Anthropic’s dispute with the Pentagon over its terms of service. “None of them are open to running agents on one-way attack drones, or running agents on missile systems.”

Despite this reluctance from some, Scout AI is currently utilizing existing LLMs as the foundation for its agents, though the company declined to disclose which specific models. Otis stated that they have agreements with “very well known hyperscalers” to provide the pre-trained intelligence for Scout AI’s foundation model. He also refrained from commenting on whether they use open-weight models, such as those offered by Chinese companies, which are often used to reduce AI inference costs.

Scout AI plans to address this by developing its own model from scratch in the coming years, with a significant portion of its funding allocated to training and compute costs. Otis believes that Scout AI could potentially surpass existing leaders in AGI due to its model’s constant interaction with the real world.

“There’s an argument in the AGI community that you can only get so intelligent by reading the internet, and most intelligence comes with interacting in the world,” Otis said.

Is Adcock competing with his brother’s humanoid robot army at Figure? Not necessarily, according to Otis, but “we can get to scale much faster because our customer has assets,” he said, referring to the Pentagon.