Artificial intelligence on the battlefield: the Pentagon decision reshaping the future of war
Artificial intelligence has reached modern battlefields in ways few imagined possible this quickly. And now, a Pentagon decision is shaking up the landscape for good.
The United States Department of Defense has ordered the removal of Anthropic AI from all its military systems within six months. The decision did not come out of nowhere. It stems from a direct conflict between the Trump administration and Anthropic CEO Dario Amodei, which escalated to the point where the company was officially classified as a supply chain risk by the U.S. 😬
But before passing any judgment, it is worth understanding the scale of what is at stake here. Anthropic’s Claude model was the only large-scale language model operating within the Department of Defense classified systems. According to retired Navy Admiral Mark Montgomery, senior director of the Center on Cyber and Technology Innovation at the Foundation for Defense of Democracies, the military was processing roughly a thousand potential targets per day, striking most of them, with turnaround time for the next strike potentially under four hours.
With Anthropic on its way out, a race has already begun. Google, OpenAI, and other big tech companies are moving fast to fill this space, each with their own set of rules. What really changes in the relationship between war and artificial intelligence from here on out? That is exactly what we are going to explore in this article 👇
What Claude was doing inside the Pentagon
Before grasping the full scale of the disruption, it is important to understand the role that Anthropic artificial intelligence played inside the classified operations of the American Department of Defense. The Claude model was not just being tested in a controlled or pilot environment — it was in live operation, integrated into systems dealing directly with critical national security areas, including nuclear weapons, ballistic missile defense, and cyber warfare, as revealed by an internal Pentagon memo.
According to a source directly familiar with Claude’s military capabilities, the model’s primary task was to sift through massive volumes of intelligence reports, synthesizing patterns, summarizing findings, and surfacing relevant information at a speed far beyond what any human analyst could manage. This kind of data processing at scale is not just an efficiency gain — it is a structural shift in how military operations are planned and executed.
The ability to process up to a thousand potential targets per day represents an unprecedented operational transformation in the context of modern warfare. Before Claude was adopted, that volume of analysis required entire teams of analysts working for days on end, cross-referencing intelligence data, satellite imagery, battlefield video, movement histories, and a long list of complex variables. With AI, that process was compressed to under four hours, which in practice means critical decisions started being made much faster.
Aaron McLean, national security analyst at CBS News, put this transformation into context well when he said we are living through a military revolution driven by the digital revolution. According to McLean, the current revolution is powered by the explosion of data: cameras everywhere, smartphones, connected cars. The battlefield is now flooded with information in ways that would have been unimaginable a generation ago. And there is far more data than any roomful of analysts could process within the timelines that matter. AI algorithms do the sorting to build target packages, assign strike assets, and assess damage — practically in real time.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
McLean also offered a pretty illustrative example: Israel’s missile defense system. When hundreds of drones and missiles are inbound within hours, no human team can decide in real time which ones to intercept, with which asset, and at what moment. That is exactly what AI is doing.
What makes this situation even more telling is the fact that Claude was classified as the only large-scale large language model operating inside the Pentagon high-classification systems up to that point. This shows that Anthropic did not hold a peripheral or experimental position in this ecosystem — it sat at the center of a critical national defense infrastructure. Removing a system this deeply integrated within six months is a challenging timeline even by military standards, where every protocol change requires extensive testing, security validation, and personnel training.
Beyond functions directly tied to combat operations, AI was also being used for administrative functions such as research, policy development, and acquisition processes, according to Josh Gruenbaum, commissioner of the Federal Acquisition Service, the government agency that helps decide which goods and services the government uses.
How AI works alongside physical weapons systems
An important point worth reinforcing is that artificial intelligence does not operate in isolation on the battlefield. There is a lot of human oversight and physical technology involved, from aircraft carriers to drones, supplied by traditional defense contractors like Northrop Grumman, Boeing, and Lockheed Martin. The large language models powering the AI are not flying planes or launching missiles. What they do is perform enormous volumes of analysis before those actions are carried out by humans.
According to Montgomery, this advancement compressed operational timelines from days to hours. He described AI as an important enabler for the military’s ability to plan and execute combat operations quickly, but emphasized that humans are still in the process. AI is used to help plan potential strikes, not to execute them autonomously.
And that is a detail that makes all the difference. The target selection process is still led by humans. Anthropic’s government use policy allows the Department of Defense to use Claude for analyzing foreign intelligence, but the terms of use require humans to make any decisions about military targets. This distinction between automated analysis and automated decision-making is essential to understanding the current limits of AI use in warfare contexts.
Sources familiar with the military use of artificial intelligence told CBS News that AI programs, including Anthropic’s, are likely being used as part of the American operation against Iran. The Pentagon has not specified exactly how AI tools are being deployed in that context, but experts with knowledge of military operations described likely scenarios.
Montgomery also pointed out that AI is a significant boost to operations, but that the war could still be fought without it. Traditional contractors still manufacture the vast majority of weapons. According to him, this war is being fought with armaments, 98% of them supplied by the major traditional contractors, and they are performing very well. It would be possible to fight a war without AI, but it would be less desirable. And that role is set to grow campaign after campaign.
The political conflict behind the decision
The rift between the American government and Anthropic did not originate from a technical issue. It has very clear political roots. Dario Amodei, the company’s CEO, went on a direct collision course with the Trump administration over who should have the final word in defining restrictions on how Claude would be used by the military. This clash of perspectives escalated to an extreme point when the company was formally classified as a supply chain risk by the United States.
That classification is not just symbolic — it carries immediate practical consequences. It means any active federal contract can be suspended, new partnerships are blocked, and the company is treated with the same level of scrutiny reserved for suppliers considered potentially hostile to American national interests. It is a serious measure, with a direct impact on Anthropic operations and reputation in the defense market.
It is worth remembering that the Pentagon had signed a 200 million dollar contract with Anthropic in July to integrate Claude into its systems. That contract was canceled in the wake of the conflict. Anthropic responded by suing the federal government, alleging retaliation. In the lawsuit, the company argued that the Constitution does not allow the government to use its enormous power to punish a company for its protected free speech, and that no federal statute authorizes the actions taken.
In a relevant development, Microsoft and professionals from OpenAI and Google filed amicus curiae briefs in support of Anthropic’s lawsuit. This signals that the industry as a whole is concerned about the precedent this case could set, regardless of commercial rivalries.
What this episode reveals, quite clearly, is that the artificial intelligence sector now operates in a field where technological decisions and political decisions are completely intertwined. A company can have the best model on the market, with proven performance in critical environments, and still be removed from a strategic position because of disagreements with the current administration. This creates a troubling precedent for the entire industry, especially for companies that bet on more cautious approaches to AI development.
The race to fill the space left by Anthropic
With Anthropic removal set in motion, the American defense market has become openly contested territory. Google and OpenAI have already signaled concrete moves to fill this space, and the interest is entirely understandable when you are talking about multimillion-dollar contracts with the Pentagon. Each of these companies comes with a different portfolio of capabilities and, more importantly, with distinct stances on the use of artificial intelligence in warfare and military operations.
Google announced in an official blog post the launch of AI agents for unclassified military uses, powered by its Gemini model. The initiative was announced in March 2026, shortly after the crisis between Anthropic and the Department of Defense began, which shows how quickly companies are positioning themselves to capture this opportunity.
Sam Altman, CEO of OpenAI and a direct rival to Anthropic, posted on X about the use of ChatGPT artificial intelligence models on the Pentagon classified network. OpenAI then released details about the deal struck with the Department of Defense, highlighting what it called three red lines for AI use:
- A ban on autonomous lethal weapons
- A ban on mass surveillance of Americans
- A ban on automated high-risk decisions without human oversight
These three guidelines are significant because they publicly establish the boundaries that OpenAI commits to respecting, creating a benchmark for any other company entering this market.
Beyond the major platforms, a number of startups specializing in AI for defense are also eyeing this opportunity. Companies like Palantir, Anduril, and Shield AI already have a strong presence in the American military ecosystem and could serve as either direct suppliers or integrators of third-party language models into Pentagon systems. The practical outcome of all this movement is that the vacuum left by Anthropic will be filled — the question is who will set the new ethical and operational rules guiding the use of AI in warfare scenarios from here on out. 🤖⚔️
And while the transition unfolds, the Pentagon is still using Anthropic products in operations in Iran, despite the supply chain risk designation. That alone illustrates how deep the technological dependency already runs and how complex it is to simply swap out an AI model in active military operations.
What changes in the relationship between AI and war
The episode involving Anthropic and the Pentagon is more than a corporate or political dispute. It marks a major turning point in how governments and armed forces around the world will think about and structure their relationship with artificial intelligence. The fact that a language model was capable of processing analyses that previously required human teams working for days straight shows that AI has already moved past the experimental phase in a military context. It is real operational infrastructure, and the decisions about who controls it carry direct consequences on the battlefield.
At the same time, the case raises questions that go well beyond operational efficiency. When an AI is integrated into systems for military target identification and analysis, the questions about accountability, human oversight, and ethical limits become urgent. Who is responsible for a decision made based on an analysis generated by a language model? How do you ensure these systems do not amplify errors or biases in contexts where the consequences are irreversible? These questions still do not have settled answers, and the acceleration of AI adoption in the defense sector is outpacing any existing regulatory or ethical effort.
Montgomery offered an interesting perspective when he stated that a human is still in the loop, but that AI is doing the work that used to take days of analysis — and doing it at a scale no previous campaign has matched. That statement captures the current moment well: AI is indispensable, but human oversight remains the last barrier before an irreversible action is taken.
What becomes clear after this episode is that artificial intelligence has moved beyond being just a productivity tool to become a strategic resource of national power. Countries and governments will increasingly treat their AI ecosystems as geopolitical assets, and the companies developing these models will need to navigate a landscape where technology, politics, and war are in constant friction.
Anthropic exit from the Pentagon is just the first chapter of a story still being written. And the outcome of that story will define not only who supplies AI to the American military, but how the entire global artificial intelligence industry positions itself in the face of the deepest dilemmas of our time. 🌐
