Army AI War Test Hits Snags

In the blistering California desert, the U.S. Army just let artificial intelligence help run a mock war — and the system struggled in ways that raise hard questions about who really controls tomorrow’s battlefield.

Story Snapshot

Nearly 10,000 U.S. and allied troops tested artificial intelligence-driven command systems at Fort Irwin during Project Convergence Capstone 6.
The Army’s new Next Generation Command and Control platform, meant to link sensors, drones, weapons, and soldiers in real time, faced heat-related shutdowns and network glitches.
Leaders still hailed the event as a success and a “campaign of learning,” using failures to decide which technologies will be kept, fixed, or cut.
The experiment highlights both the promise of faster decision-making and the risk of handing more control to software built by powerful defense tech companies.

Army Turns the Mojave Desert into an AI War Lab

For ten days in late July, the sand and rocks around Fort Irwin, California, became a test bed for the Army’s vision of future war. Project Convergence Capstone 6 brought together almost 9,700 participants, including U.S. soldiers, allied troops from “Five Eyes” nations, government officials, and industry partners. They ran large mock battles meant to copy a fight with a powerful enemy, using about 100 new technologies, from smart drones to digital maps, under brutal heat in the Mojave Desert.

Project Convergence began in 2020 as what Army Futures Command calls a “campaign of learning,” not a normal training exercise. The idea is to plug together sensors, radios, command posts, satellites, and weapons and then see what breaks and what works before any of it goes to real units. Officials say the goal is simple but huge: help commanders and soldiers see threats sooner, share that information faster, and strike targets seconds after they are spotted, not minutes or hours.

Next Generation Command and Control: Promise and Problems

At the heart of Capstone 6 was the new Next Generation Command and Control system, known as NGC2. This software and network platform is supposed to pull in data from sensors, drones, radios, and troops and give commanders a live digital picture of the battle on the move. The Army has built NGC2 with major defense tech firms such as Anduril and Palantir, and this desert test was its largest tryout so far. Leaders told reporters the system “turned the dials up” on networking and would soon move into bigger units if results hold.

On paper, NGC2 aims to knit the Army into a wider plan called Combined Joint All-Domain Command and Control, which seeks to link all U.S. services and key allies in one web of sensors and shooters. In earlier Project Convergence events, artificial intelligence tools helped cut the “sensor-to-shooter” time from about 20 minutes down to 20 seconds in some trials. One demonstration saw an artificial intelligence system identify and generate firing solutions for 15 targets in a single hour, showing the raw speed these tools can reach. Supporters say this kind of speed is vital as rivals like China and Russia build their own high-tech war networks.

What Did Not Go According to Plan

Despite the upbeat public tone, not everything worked in the Mojave heat. Reporting from the exercise says some soldiers struggled with spotty connectivity as radios and networks dropped out or clogged under heavy use. In other cases, computer gear and sensors overheated and shut down as temperatures soared, forcing troops to fall back on older tools and manual work-arounds. The Army framed these issues as expected stress-test results, but for many watching, they showed how fragile complex digital systems can be when the environment fights back.

Officials and outside analysts note that this mix of success and failure is built into Project Convergence by design. The event is meant to expose weak points in artificial intelligence systems, networks, and hardware long before they reach front-line units. After Capstone 6, Army leaders and engineers will sort through data to decide which systems are ready to grow across the force, which need redesign, and which should be cut from future plans entirely. Those choices will guide modernization spending and contract awards starting in 2027, shaping how billions of dollars are used.

Shared Public Concerns: Speed, Control, and the “Deep State”

Many Americans, both conservative and liberal, look at this kind of experiment and see their deeper worries about power and accountability. Conservatives frustrated with global wars, huge Pentagon budgets, and high-tech spending may wonder if artificial intelligence war labs in the desert bring real safety or just more profit for defense contractors. Liberals worried about civilian harm, inequality, and what they view as discrimination against minorities may fear that faster kill chains could lead to less human judgment and more deadly mistakes, especially far from public view.

Across the political spectrum, there is a growing belief that government elites and the “deep state” protect their own systems first. Project Convergence shows that the state is racing to automate life-and-death decisions, while basic problems at home — wages, health costs, border security, crime — feel stuck. At the same time, the Army’s choice to invite reporters and allies into this experiment suggests a push for some transparency and shared control, even as software from powerful private firms becomes central to war.

Why This Desert Test Matters Going Forward

Fort Irwin’s mock battles are not just science fiction games; they are early steps toward the real rules of future war. The more the military leans on artificial intelligence and complex networks, the more regular citizens must ask who writes the code, who can see the data, and who is allowed to say “no” when a system recommends a strike. The Army describes Project Convergence as a way to keep humans “in the loop,” using artificial intelligence to support, not replace, human judgment. Whether that promise holds under pressure is a question that this year’s desert glitches keep alive.

For now, Capstone 6 shows both sides of the artificial intelligence future: breathtaking speed and reach, and basic breakdowns when reality intrudes. It also reminds us that behind every shiny network map stand human soldiers sweating in the heat, trying to make sense of what the machines tell them. As the government moves forward with these systems under President Trump and a Republican Congress, citizens who feel shut out of power have strong reasons to pay attention — and to demand that the race for battlefield advantage does not outrun the values the country was built on.