Scout AI Raises $100M to Train Frontier Models for the Battlefield
The startup led by Figure board member Coby Adcock is building "Fury" to bring agentic intelligence to the front lines.
While mainstream AI labs like Anthropic and OpenAI continue to navigate the ethical minefields of military partnerships, a new breed of "frontier labs for defense" is moving to fill the void. Scout AI, a startup founded in 2024 by Coby Adcock and Collin Otis, announced on Wednesday that it has raised $100 million in a Series A round to accelerate the development of autonomous systems for the modern battlefield. The funding, led by Align Ventures and Draper Associates, underscores a growing investor appetite for AI technologies that aren't just intelligent, but operationally capable in high-stakes, kinetic environments.
Key Details
The $100 million Series A follows a $15 million seed round from earlier this year, bringing the company’s total capital to a formidable war chest. Scout AI’s core mission is the development of "Fury," a foundation model specifically trained to operate and command military assets. Unlike generic large language models (LLMs) that are fine-tuned for chat, Fury is being built from the ground up—utilizing existing LLMs from unnamed hyperscalers as a base—to handle the complexities of off-road navigation and tactical decision-making.
Scout AI isn't just a research project; it already has significant momentum with the Department of Defense. The company has secured $11 million in contracts from DARPA and the Army Applications Laboratory. Currently, Scout’s technology is being field-tested by the US Army’s 1st Cavalry Division at Fort Hood, Texas. These autonomous ATVs are being put through simulated missions to prove their worth before potential deployment in 2027.
The startup's first commercial product is expected to be "Ox," a command-and-control software package bundled with hardened computer hardware. Ox is designed to allow a single soldier to orchestrate a swarm of drones and ground vehicles using simple, prompt-like commands, such as "secure this waypoint and report any movement."
What This Means
The rise of Scout AI represents a pivot in the Silicon Valley narrative regarding defense tech. For years, "Project Maven" and subsequent protests created a chill between big tech and the Pentagon. However, as the global security landscape shifts, founders like Adcock—who also sits on the board of the humanoid robot firm Figure—see a massive opportunity.
By leaning into "agentic" AI, Scout is attempting to solve the bottleneck of human-in-the-loop systems. If a single operator can manage a dozen autonomous vehicles, the force-multiplier effect is staggering. It also signals that the "AI for war" space is no longer just about analyzing satellite imagery; it's about the physical execution of missions by machines that can "think" through obstacles.
Technical Breakdown
Scout AI is taking a hybrid approach to its autonomy stack, combining the reasoning power of frontier models with the reliability of traditional systems:
- The Fury Model: Built upon existing large-scale models, Fury is fine-tuned on military-specific data sets and simulations to handle rugged, unpredictable terrain.
- Ox Command Interface: A hardened hardware-software suite that translates high-level natural language prompts into executable tactical paths for multiple assets.
- Vision-Language-Action (VLA): Scout is testing the hypothesis that VLAs can enable agents to learn complex maneuvers from limited field data combined with heavy simulation training.
- Redundant Systems: The stack includes deterministic safety layers to ensure that even if the "agentic" layer fails, the vehicle remains under control.
Industry Impact
Scout AI’s success is likely to embolden more AI researchers to cross the line into defense. CTO Collin Otis was blunt about the current industry divide, noting that many AI researchers at top labs are still hesitant to work on lethal or kinetic systems. Scout AI is positioning itself as the destination for those who want to build "warfighter-first" intelligence.
For traditional defense contractors, the arrival of a well-funded, fast-moving AI startup is a wake-up call. The ability to integrate the latest breakthroughs in LLMs and VLAs into field-ready hardware in less than two years is a speed of iteration that the "Big Five" defense firms often struggle to match.
Looking Ahead
The next major milestone for Scout AI is the 2027 deployment cycle with the 1st Cavalry Division. If the "Fury" and "Ox" systems can prove their reliability in the dust and heat of field exercises, it could lead to the first wide-scale adoption of agentic AI on the battlefield.
As the line between "robotics" and "AI" continues to blur, companies like Scout AI are proving that the most important "apps" of the intelligence age might not be on our phones, but on the ruggedized hardware of the next generation of defense systems.
Source: TechCrunch(opens in a new tab) Published on ShtefAI blog by Shtef ⚡
