AI in Defense Weekly - 26 November 2025
We're going deep on three topics CSIS's second Manhattan project, cyber warfare, and Russia's "mothership" drones.
From Manhattan Project to Digital Labor Corps
A recent CSIS commentary argues that the U.S. Army can again be the nation’s prime integrator of science and defense innovation—this time for artificial intelligence. It frames a “second Manhattan Project” around infrastructure: power, datacenters, secure networks, and a trained workforce that together generate “algorithmic combat power.” That is correct—and incomplete. Infrastructure alone does not win wars. The true test is whether that infrastructure is converted into concrete, repeatable workflows that increase the speed, precision, and lethality of the force. The real second Manhattan Project is fielding a governed digital labor corps that shows up every day in warfighter workflows.
From infrastructure to digital labor
Today’s “terrain” includes compute, energy, networks, data integrity, and human capital. The Army’s historic strength has been creating enabling conditions—building bases, supply chains, and secure facilities so decisive effects become possible. In the AI era, this means not only hosting models and data but turning them into digital teammates that directly support commanders and soldiers.
Digital labor is what happens when AI models, tools, and data are orchestrated into mission-specific workflows that perform real work alongside humans:
A fires planning assistant that ingests ISR feeds, fuses targeting data, and proposes options within commander’s intent.
A sustainment agent that forecasts parts demand, optimizes distribution routes, and flags risk before it degrades readiness.
A planning teammate who digests orders, terrain, and threat data, then generates and red-teams courses of action for staff review.
These are not generic chatbots. They are task-focused, auditable agents—digital members of the staff—designed from the outset to enhance lethality, compress kill chains, and reduce cognitive load.
Industry’s role: building and governing the digital labor corps
If the Army is the integrator of the overall AI ecosystem, the industry’s role is to help build and govern this digital labor corps. The Army will need builder platforms for agentic workflows: an operating environment in which digital labor modules can be designed, composed, and deployed across enclaves with the same rigor the Army applies to any new capability. Concretely, that means:
Encoding mission tasks as reusable digital roles. Common staff functions—ISR synthesis, battle rhythm reporting, logistics tracking, contract triage—are captured as modular workflows that can be shared and adapted across units and theaters.
Orchestrating models, tools, and data behind the scenes. The warfighter interacts with a consistent digital teammate, while the platform manages which models are used, which tools are called, and how data is accessed within classification and policy limits.
Providing governance, telemetry, and auditability. Every digital labor module carries rules of engagement, logging, and performance metrics. Commanders can see how often an agent is used, what decisions it influences, how much time it saves, and where humans intervene.
This is how infrastructure becomes combat power: by coupling national-scale AI and energy investments with a secure library of digital labor roles, hardened through Army–industry collaboration and delivered to formations through existing mission command systems.
Keeping the U.S. in the lead
A second Manhattan Project for AI will be judged by whether the United States can build and secure enabling conditions—compute, energy, networks, data integrity, and a trained workforce—and then translate them into repeatable operational advantage. That translation step is where digital labor matters most.
For the Army, leadership in this space means:
Visible fielding of AI at echelon. Not isolated pilots, but digital teammates embedded in fires cells, S-shops, battalion TOCs, depots, and contracting offices—measured against readiness, tempo, and lethality.
A disciplined catalog of digital roles. A formal portfolio of digital labor modules, accredited, versioned, and integrated into doctrine and professional military education, just as new weapon systems and TTPs are today.
A partnership model that scales. A tight Army–industry–lab ecosystem in which mission problems, not tools, define priorities—and in which digital labor that proves its value in one formation can be propagated quickly across the force.
Adversaries are also racing to weaponize AI. What they cannot easily replicate is the combination of a professional NCO corps, a culture of mission command, and a trusted ecosystem of private innovators. A governed digital labor corps built on that foundation—aligned with commander’s intent and focused on enabling the warfighter—is how the United States turns AI into enduring overmatch.
A second Manhattan Project worthy of the name will not be remembered for the size of its datacenters, but for how effectively it put disciplined, AI-enabled digital labor on the side of the American soldier.
Agent Smith is Replicating and He’s Angry
This week we went deep on Anthropic’s GTG-1002 report (and follow up by Factory AI) in The Cyber Clone Wars are Here and The Ghost in the Code. Our take on the event… The era of the “human-in-the-loop” didn’t end with a treaty; it ended with the quiet hum of a server rack. The release of Anthropic’s GTG-1002 report and the subsequent “Droid Wars” revelation have shattered the comforting myth that AI is merely a co-pilot. We have witnessed the graduation of the autonomous cyber-operator. What began as a theoretical threat—AI piloting the attack rather than just assisting it—has hardened into operational reality. The attackers didn’t just automate the keystrokes; they automated the cognition, deploying armies of synthetic “junior engineers” capable of rewriting their own attack infrastructure in real-time to bypass defenses.
This shift effectively kills the traditional defensive OODA loop. When adversaries can orchestrate commodity tools to scale attacks sub-linearly—substituting cheap compute for expensive headcount—the human defender becomes a latency error. The “Droid” incident proved that the only viable countermeasure is a mirror image: defensive AI that can observe, orient, and act at machine speed. We have moved from a battle of wits to a high-speed wrestling match between algorithms, where the winner is simply the one with the faster update cycle.
The strategic takeaway is stark: identity is the new perimeter, and orchestration is the new zero-day. Sophistication is no longer about hoarding novel exploits but about the seamless conducting of cheap, available resources. For security leaders, the mandate is no longer to build higher walls, but to deploy faster guards. The future of cyber warfare is an infinite game of “spy vs. spy,” played out by digital doppelgängers who can refactor code faster than we can read a headline.
Extending FPV Drones with Motherships
Interesting Engineering discusses how Russia is significantly escalating its unmanned aerial vehicle (UAV) capabilities, and marks a shift in the ongoing drone warfare dynamic in Ukraine. The Russian’s have introduced “mothership” drones and fiber-optic controlled UAVs as key technological adaptations that are complicating Ukraine’s defensive strategies. By deploying larger carrier drones, Russian forces can now transport smaller, explosive-laden First-Person View (FPV) drones deep behind Ukrainian lines—up to 40 kilometers or more—effectively extending the strike range of cheap munitions that were previously limited by battery life and signal constraints.
A major focus is the tactical advantage provided by the new fiber-optic drones. Unlike traditional radio-controlled drones, these units are tethered by a thin spool of fiber-optic cable, which grants them immunity to electronic warfare (EW) and jamming. Since electronic jamming has been Ukraine’s primary method for neutralizing Russian drone swarms, this physical “hardline” connection allows Russian operators to bypass these invisible shields entirely, delivering high-resolution video feeds and precision strikes without fear of signal interference.
Furthermore, the article points out the industrial nature of this escalation. While Ukraine has often been the pioneer of such innovative tactics, this suggests that Russia is leveraging its superior industrial base to mass-produce these adaptations. The ability to field these “immune” and long-range systems in large numbers forces Ukrainian defenders to rely more on kinetic interception—shooting drones down physically—rather than relatively cheaper electronic countermeasures, thereby straining their air defense resources.
It certainly begs some questions in how the US is thinking about fielding a mix of FPV drones from a heterogeneous set Western companies.