At a U.S. military installation in central California, four-wheeled all-terrain vehicles navigate rugged hillside paths. While this appears to be a standard training drill, it serves a different purpose: teaching artificial intelligence models how to function in combat zones. These autonomous military ATVs are operated by Scout AI, a company founded in 2024 by Coby Adcock and Collin Otis, which describes itself as a "frontier lab for defense." On Wednesday, the startup announced it had secured a $100 million Series A funding round, spearheaded by Align Ventures and Draper Associates, building on a $15 million seed round from January 2025. Scout is developing an AI system called "Fury" designed to control and direct military equipment, initially focused on logistics but eventually intended for autonomous weaponry. Chief Technology Officer Collin Otis compares this effort—which builds upon existing large language models—to training soldiers. "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."

Scout has already obtained $11 million in military technology development contracts from organizations such as DARPA, the Army Applications Laboratory, and other Department of Defense clients. It ranks among 20 autonomy companies whose technology is being utilized by the U.S. Army’s 1st Cavalry Division during routine training at Fort Hood in Texas, with plans for the unit to deploy proven products when it next heads out in 2027. For Scout’s internal evaluations, the real test occurs on the base's uneven terrain. There, the company’s operations team—led by former soldiers—pushes the vehicles through simulated missions. Although autonomous cars are becoming more common in urban settings, they operate within structured environments governed by rules. Navigating autonomously on unmarked trails or off-road presents a far greater challenge. Otis, a former executive at autonomous trucking firm Kodiak, said he was inspired to launch Scout after realizing the system he helped build there lacked the intelligence needed to function in unpredictable war zones.

**A Fresh Approach to Autonomy**

Scout is adopting a newer autonomy technology: Vision Language Action models, or VLAs, which are rooted in LLMs and used to control robots. First introduced by Google DeepMind in 2023, this technology has spawned robotics startups like Physical Intelligence and Figure AI, the humanoid robot company led by Adcock’s brother, Brett. Adcock sits on Figure’s board and says that experience convinced him of the potential to bring broader intelligence to the military’s expanding fleet of autonomous vehicles. His brother connected him with Otis, who was advising Figure, and together they began applying cutting-edge AI to military challenges. "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."

I had the chance to drive one of Scout’s ATVs across the bumpy trails, and the terrain proved demanding: steep inclines, loose sand on curves, disappearing paths, and confusing junctions. Though I’m not an experienced ATV driver, I managed reasonably well on my first attempt. That kind of general intelligence is what the company aims to embed in its models, which it has been training using these ATVs for just six weeks after initially using civilian vehicles. I also rode in the ATV under autonomous control and noticed a distinct difference—it accelerates faster than a human driver who might consider passenger comfort. The operations team highlighted how the vehicles hug the right side on wider trails but stay centered on narrow ones, mimicking their training drivers. When uncertain, they abruptly slow down to assess their next move, a behavior that occurred several times during a 6.5-kilometer loop before returning to base. Although VLAs are so new that no company has deployed them operationally yet, "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," said Stuart Young, a former DARPA program manager who worked on ground vehicle autonomy. Like other autonomy companies, Scout’s full autonomy stack also incorporates deterministic systems and other AI variants to round out its agents’ capabilities. Young left DARPA this month to join Field after overseeing a program called RACER, which challenged companies to create high-speed, autonomous off-road vehicles, helping seed this field much like the organization’s Grand Challenge boosted self-driving cars. Two competitors in this space, Field AI and Overland AI, emerged from that program, and Scout joined later as an additional participant. According to Scout executives and military technologists, the initial applications of ground autonomy will involve automated resupply: carrying water or ammunition to remote observation posts, or in convoys where a crewed truck might be followed by six to ten autonomous vehicles, freeing up human labor for more critical tasks. 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 to assist.

**Adding Intelligence to the Army’s Motorpool**

Scout views itself primarily as a software company, creating an intelligence layer for military machines. It doesn’t plan to manufacture autonomous vehicles but rather to build on top of them. Adcock expects the startup’s first widely adopted product to be one called "Ox," the company’s command and control software, bundled on hardened computer hardware (GPUs, communications, cameras). This system is designed to let individual soldiers orchestrate multiple drones and autonomous ground vehicles using prompt-like commands: "Go to this waypoint and watch for enemy forces."

However, making that software functional requires training on real vehicles. That’s where Foundry comes in—the company’s name for its training range at the military base. There, drivers spend eight-hour shifts putting the ATVs through their paces, then use a reinforcement learning system to log instances where they had to take over, which is then used to improve the model. The base commander has even asked the company’s ATV to assist with security patrols. One hypothesis Scout is testing is that VLAs, combined with training data from simulations, will enable this relatively limited dataset to produce a fully capable driving agent. While the vehicle seems comfortable on trails, for example, it isn’t ready to operate fully off-road yet. Scout is also experimenting with drones for reconnaissance and as weapons, equipping them with intelligence through vision language models, a multi-modal LLM variant. The company is working on a system where groups of munition drones fly alongside a larger "quarterback" platform that provides more computing power to command them. In one mission, the drones would search a geographic area for hidden enemy tanks and attack them, potentially without human intervention. Otis argues that the alternative in this scenario might be indirect artillery fire, which is less precise than drone strikes. While autonomous weapons are a contentious topic in defense tech politics, experts note the concept is not new: heat-seeking missiles and mines have been in use for decades. The key question for technologists is how the weapons are controlled, according to Jay Adams, a retired U.S. military officer. He notes that Scout’s munition drones can be programmed to only attack threats within a specific geographic area, or only with human confirmation. He also says autonomous weapons platforms are unlikely to fire out of fear, as a young soldier might. VLAs also offer promise for improved targeting. Scout says its models are pretrained on a specific set of military data to prepare them for scenarios like encountering an enemy tank during a resupply mission. Lt. Col. Nick Rinaldi, who oversees Scout’s work for the Army Applications Laboratory, says that while automated targeting is challenging 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 believes that drones capable of identifying their own targets are crucial for future warfare: while Russia’s invasion of Ukraine has sparked intense interest in drone warfare, he thinks having humans operate individual UAVs doesn’t scale enough for the U.S. to counter a large number of low-cost unmanned systems should they threaten American forces.

**A Mission to Counter Anti-Military Vibes**

Like many defense startups, Scout wears its mission on its sleeve, and executives freely criticize companies reluctant to hand their technology over to the government. Google, for instance, reportedly withdrew from a Pentagon competition to develop control systems for autonomous drone swarms—a capability Scout is also working on. "None of them are open to running agents on one-way attack drones, or running agents on missile systems."

Nevertheless, Scout is actually using existing LLMs as the foundation for building its agents, though it declined to specify which ones. Otis says it has agreements with "very well known hyperscalers" to provide the pretrained intelligence for Scout’s foundation model. Otis also declined to comment on whether it uses open-weight models, such as those offered by Chinese companies. Many companies reliant on AI inference build on these models to operate at lower cost compared to models from frontier labs like Anthropic or OpenAI. Scout expects to address this by building its own model from scratch in the years ahead, and the founders say much of its capital will go toward those training and compute costs. Indeed, Otis wonders if Scout will 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. Does that mean Adcock is competing with his brother’s army of humanoid robots at Figure? No, Otis says, but "we can get to scale much faster because our customer has assets," he said, referring to the Pentagon.