Scout AI
Last reviewed
Jun 3, 2026
Sources
7 citations
Review status
Source-backed
Revision
v1 · 1,801 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 3, 2026
Sources
7 citations
Review status
Source-backed
Revision
v1 · 1,801 words
Add missing citations, update stale details, or suggest a clearer explanation.
Scout AI is an American defense artificial intelligence company that builds foundation models to control uncrewed military systems. The company describes itself as a "frontier lab" for defense autonomy and is best known for Fury, a camera-only vision-language-action model designed to turn drones, ground vehicles, and other robots into autonomous agents that take orders in plain language. Founded in 2024 by Colby Adcock and Collin Otis and based in Silicon Valley, Scout AI raised an oversubscribed $100 million Series A in April 2026, which it and several trade outlets described as the largest Series A round in the U.S. defense technology sector to that point.[1][2][3]
The company sits squarely inside the post-2023 wave of defense-AI startups, alongside firms such as Anduril, Shield AI, and Palantir, that have argued the U.S. military needs large fleets of cheap, expendable autonomous systems coordinated by software rather than small numbers of exquisite, human-piloted platforms. Scout AI's particular bet is that the coordination layer should be a single robotics foundation model, trained the way large language models are trained, rather than the hand-coded autonomy stacks that have dominated military robotics.
Scout AI was founded in August 2024 and emerged from stealth on April 16, 2025.[4] It is headquartered in Silicon Valley; early announcements listed Sunnyvale, California.[4]
The chief executive and co-founder is Colby Adcock, a former technology private-equity investor who also sits on the board of the humanoid-robotics company Figure AI.[4] The chief technology officer and co-founder is Collin Otis, an autonomy engineer who was a founding engineer and director of autonomy and AI at the self-driving trucking company Kodiak Robotics, and earlier worked at Uber's self-driving unit, Uber ATG, as head of data science and chief of staff.[4] Otis has framed Scout's goal in unusually blunt terms, telling TechCrunch the company wants to "teach this thing to be an incredible military AGI."[1]
The team has leaned heavily on people with military backgrounds. Operations are led by Jay Adams, a retired U.S. Army captain, and the company has hosted serving officers as fellows, including an active-duty infantry officer who acts as a military liaison.[1] As of the April 2026 funding round, Scout AI employed 34 people across AI, robotics, and national security.[2]
Fury is Scout AI's core product: a defense-specific vision-language-action foundation model. The "vision-language-action" framing, borrowed from the broader embodied AI research community, means the model takes in camera imagery and a natural-language instruction and outputs the low-level control commands that actually drive a vehicle or aircraft.[4] Scout describes Fury as a fully learned, camera-only autonomy system, meaning it does not rely on expensive sensor suites such as lidar; it is built to run on high-rate commercial off-the-shelf hardware and to keep working in environments where communications and GPS are jammed or unavailable.[2][4]
The pitch is portability across very different machines. Because the model learns to map perception and intent to action rather than encoding rules for one specific airframe, Scout claims the same base model can be adapted to ground vehicles, aircraft, and other platforms. Otis has used the analogy of handing someone a video-game controller: with enough general competence baked in, he argued, "you could learn to fly that thing in minutes."[1] When it came out of stealth, Scout demonstrated two early prototypes running Fury autonomously at a proving ground in the Santa Cruz Mountains: a ground vehicle designated G01 and an aerial vehicle designated A01.[4]
Above the model sits a command-and-control product. In February 2026 Scout introduced what it then called the Fury Autonomous Vehicle Orchestrator, an "agentic interoperability layer" that lets a commander state mission intent through a C2 interface and then translates that intent into platform-specific instructions for a mixed fleet, fusing telemetry and video into a common operating picture and adjusting plans as conditions change.[5] By the April 2026 funding announcement the orchestration product was being marketed under the name Ox, described as a C2-based autonomous vehicle orchestrator that lets a single soldier direct multiple drones and ground vehicles through prompt-like commands.[2][3] In practice the two names describe the same idea: Ox is the control software, and Fury is the model running underneath it.
Scout's headline demonstration is a fully autonomous, end-to-end strike mission carried out on a military base in central California. In it, a commander issued a natural-language order and the orchestrator deployed and controlled a self-driving ground vehicle together with two armed drones to find, identify, and strike a target.[2][3] Scout emphasized that the demonstration ran on real hardware in relevant terrain rather than in simulation, with a human operator retaining oversight.[5]
That demonstration touches the most contested question in this field: how much lethal decision-making should be delegated to software. U.S. policy still requires "appropriate levels of human judgment" over the use of force, and Scout, like its competitors, stresses that a human stays in or on the loop. The company's marketing nonetheless makes plain that the technical capability for an AI system to run a strike from order to impact is the thing being built and sold, which is exactly why lethal autonomous weapons remain an active subject of arms-control debate.
Unusually for an AI company, a lot of Scout's work happens outdoors. The company runs a training operation it calls the Foundry at a U.S. military base in central California, where human drivers pilot all-terrain vehicles across hillside trails in long shifts.[1] Those sessions generate the driving data that trains and refines Fury, and a reinforcement learning loop logs every moment a safety driver takes over from the model, treating each takeover as a signal about where the model still falls short.[1] Scout told TechCrunch it began with civilian ATVs and spent roughly six weeks on real-vehicle training, supplemented by simulated off-road missions.[1] The approach echoes the data-collection and imitation-learning playbook that Otis worked on in commercial self-driving, now pointed at military ground vehicles.
Scout AI emerged from stealth in April 2025 with an oversubscribed $15 million seed round that had closed in January 2025. The seed was led by Align Ventures and Booz Allen Ventures, with participation from Draper Associates, Decisive Point, Perot Jain, Sigmas Group, Evolution VC, BVVC, FJ Labs, Gaingels, and others.[4]
On April 29, 2026, the company announced a $100 million Series A.[2] The round was co-led by Align Ventures and Draper Associates, the venture firm founded by Tim Draper, with participation from Decisive Point, Booz Allen Ventures, BVVC, Neman Ventures, Evolution VC Partners, Heraclitus Capital Management, Sigmas Group, Disruptive Founders Fund, and Vaughn Capital Partners.[2] Scout and multiple defense and venture outlets described the oversubscribed round as the largest Series A in U.S. defense technology to date.[2][3] Adcock framed the raise around the company's mission, saying in the announcement that "the most important frontier in AI is the physical world, and it should be pursued in service to the men and women who defend this country."[2] The funding is earmarked for scaling Fury and its multi-agent coordination capabilities.[2]
The table below summarizes Scout AI's disclosed financing.
| Round | Date announced | Amount | Lead investors |
|---|---|---|---|
| Seed | April 16, 2025 | $15 million | Align Ventures, Booz Allen Ventures[4] |
| Series A | April 29, 2026 | $100 million | Align Ventures, Draper Associates[2] |
Scout has built its government business quickly. In its first year it reported booking roughly $11 million in contracts tied to the U.S. defense establishment, including work with DARPA and the Army Applications Laboratory.[1][2] The federal agency that buys this work is the Department of Defense, which the Trump administration began branding as the "Department of War" under a September 2025 executive order; Scout's announcements use the newer name.[6]
The most concrete program is an Army uncrewed-systems autonomy award announced in August 2025. Under that 16-month effort, Scout agreed to integrate Fury onto Army-furnished Infantry Squad Vehicles to show that the model could meet the service's full autonomy requirements, with Textron Systems leading vehicle integration and Edge Case Research providing independent safety validation. The award carried a potential follow-on procurement value of up to $150 million.[7] Scout's technology has also been tested by the Army's 1st Cavalry Division at Fort Hood, Texas, as one of roughly 20 autonomy companies in that exercise, with the company targeting a fielded deployment around 2027.[1]
Adcock cast the Army award as a deliberate break from older approaches, saying Fury "is modeled on the same end-to-end learning approach that powers the most widely deployed and cost-effective autonomy systems in the world" and that the contract showed "the Army's intent to leapfrog legacy automation systems and incorporate AI."[7]
Scout AI is part of a broader shift in how the Pentagon thinks about autonomy, captured by initiatives such as the Defense Department's Replicator program, which aims to field thousands of attritable autonomous systems quickly. The strategic logic, shared across the sector, is that mass and software-driven coordination can offset an adversary's numbers and that one human should be able to command many machines rather than babysit one. Scout's specific contribution to that argument is the claim that a single learned foundation model, trained on real driving and flying data and wrapped in an agentic C2 layer, is a better path than bespoke autonomy code written platform by platform.
Whether Fury lives up to that claim at the scale of an actual battlefield remains unproven; the public record so far consists of controlled demonstrations, early prototypes, and a growing but still modest contract base. What is clear is that investors are treating defense autonomy as one of the hottest categories in venture capital, and that a $100 million Series A for a company barely two years old is a sign of how much money is now chasing the idea of an "AI brain" for uncrewed warfare.