Large language models are no longer being discussed only as office assistants, coding tools, or search interfaces. In the defense industry, they are being tested as part of military AI systems that could help translate orders, coordinate autonomous platforms, and surface information for human operators.
A demonstration by Anduril at a secret US military base about 50 miles from the Mexican border showed the direction of travel. Four miniaturized jet aircraft, codenamed Mustang, flew in formation while a simulated fifth aircraft appeared on a nearby computer screen. After the command "Mustang intercept," an AI system parsed the order, communicated with the drones, and responded: "Mustang collapsing." Within about a minute, the drones converged on the target and destroyed it with virtual missiles.
What Anduril's drone demo showed
The Anduril test was not a case of a chatbot casually flying aircraft on its own. The source describes a model similar to the one that powers ChatGPT being used inside a larger command process. Its role was to understand a human instruction, relay that order to the drones, and produce a spoken response.
That distinction matters. Many military systems already use older forms of AI to operate autonomously. The new experiment is about bringing LLMs into the chain of command, where they may help make complex systems easier to direct and understand.
Anduril is also developing a larger autonomous fighter for the US Air Force through a project called Fury. The aircraft is designed to fly alongside crewed jets. In that context, LLMs are being explored less as independent commanders and more as interfaces that can help pilots issue orders and receive useful information.
Why defense AI is accelerating
The source frames Anduril's demo as part of a broader defense industry push. The promise is efficiency. Kill chains are complicated, and AI is being positioned as a way to streamline them. In military terms, that can also mean making them deadlier.
Strategists quoted in the source argue that whoever controls this technology will dominate the world. That belief helps explain why the United States is trying to limit China's access to cutting-edge AI and why the Pentagon plans to increase spending on it.
The war in Ukraine is also part of the backdrop. The source points to the widespread use of low-cost drones equipped with computer vision as evidence that autonomy has become valuable on the battlefield. Generative AI has intensified that interest by giving defense planners and contractors another class of tools to test.
The money is following. A 2024 Brookings report cited in the source says funding for AI-related federal contracts grew 1,200 percent from August 2022 through August 2023, with most of those contracts coming from the Department of Defense. The source also says the trillion-dollar 2026 defense, or "war," budget includes the first-ever dedicated allocation for AI and autonomy, at $13.4 billion.
Big AI companies are now in the mix
The shift is not limited to traditional defense contractors. Anthropic, Google, OpenAI, and xAI were each awarded AI-related military contracts worth up to $200 million. That marks a sharp change from 2018, when Google pulled out of Project Maven, an effort to use AI to analyze aerial imagery.
Project Maven has not disappeared. Emelia Probasco, who studies AI military use at Georgetown University, says it is now run by Palantir and, in the form of Maven Smart Systems, has become one of the military's most widely used AI tools.
The reason LLMs are attractive for the military is straightforward. According to Probasco, they are useful for intelligence gathering because they can parse large amounts of information. They are also suited to cyber offense because they can write and analyze code.
"The ambition that is a bit scary is that AI is so smart that it can prevent war or just fight and win it," Probasco says. "Like some sort of magical fairy dust."
That warning cuts through the hype. The same source notes that current models are still too unreliable, error-prone, and inscrutable to make battlefield decisions or receive direct control of hardware.
The next interface may be worn by soldiers
Anduril is also looking beyond drones and fighters. In September, Anduril and Meta jointly bid on a US Army contract worth up to $159 million to develop a rugged augmented reality helmet display for soldiers.
Anduril says that system would deliver mission-critical information to warfighters while sensing their surroundings. It would also use a newer generation of more capable AI models that can better interpret the physical world in real time.
This points to one of the central military AI use cases in the source: not replacing every human decision, but changing how people receive, interpret, and act on information. A soldier or pilot may not need to read through every feed or manually translate every system status if an AI layer can organize and present what matters.
But that benefit also creates risk. If the system is wrong, unclear, or overconfident, its output could shape decisions in high-stakes environments. The source's strongest caution is that the technology is advancing faster as a military interface than as a trustworthy battlefield decision-maker.
Automation is becoming the assumed direction
Michael Stewart, a former fighter pilot who led the US Navy's disruptive capabilities office and was involved in AI experimentation by the Fifth Fleet back in 2022, expects war to become heavily automated. He now runs a consulting firm and speaks with military planners around the world.
"In 10, 15, or 20 years, you're going to have robots that are pretty autonomous," he says. "That's where you're going."
If those future systems use LLMs as part of their control or explanation layer, they may do more than execute tasks. They may also be able to explain which actions they took and why, in their own words.
That is the larger implication of Anduril's test. Military AI is not just about faster drones or better image analysis. It is also about language becoming a control surface for machines, contracts, cyber tools, autonomous aircraft, and soldier displays. The technology is being pulled toward war because it can process information, translate commands, and connect complex systems. Whether it can be trusted with the consequences remains the unresolved question.