Table of Contents
ToggleThe fact that the US military used AI to assist in artillery strikes against Iran is not surprising in itself.
But what's surprising about this war is its speed.
On February 27, 2026, the United States and Israel jointly launched an operation codenamed "Epic Fury" against Iran. Although the name sounds a bit childish, in the first 24 hours, more than 1,000 targets in Iran were hit.
The operation involved more than 2,000 targets, and General Brad Cooper, commander of U.S. Central Command, described it as almost "twice" the size of the 2003 invasion of Iraq.
The difference of twice is not in quantity, but in time.
This is not human speed.
During the Iraq War in 2003, it often took dozens of hours from intelligence confirmation to approval of an attack on a target. Satellite imagery was reviewed frame by frame by humans, and intelligence documents were passed, translated, reported, and reconfirmed between different departments. This is "human speed," which is limited by human attention, working hours, and cognitive capacity (memory).
Things are different now.
The Maven Smart System integrates with over 150 data sources, including satellite imagery, drone footage, signals intelligence (SIGINT), human intelligence (HUMINT), and historical strike records. This data is input into the system every millisecond. Claude is one of the core language models Maven uses to understand this data.
The specific division of labor is as follows: AI is responsible for initial screening and semantic understanding. After completing the initial analysis, it assigns priority levels to the targets and pushes them to human analysts. Analysts no longer need to start from scratch and be overwhelmed by the flood of data. They only need to process the information that the AI has already predicted and sorted, and then make the final decision.
U.S. Central Command spokesman Captain Timothy Hawkins said that AI technology "can perform initial screening of incoming data, allowing analysts to focus on higher-level analysis and verification."
AI makes strategic choices, human officers make the decisions.
Maven Smart System Battlefield Brain
The Maven Smart System is not new. Developed by Palantir Technologies, the system originated from the U.S. Department of Defense's "Project Maven" in 2017, which sparked massive employee protests within Google and ultimately led to Google's withdrawal from a military AI contract.
But Palantir continued.
To date, the Maven Smart System has become one of the most deeply integrated AI platforms in the U.S. military's classified networks. Claude's ability to run on the Pentagon's classified network is a result of a 2024 contract between Anthropic and the Department of Defense, with a maximum contract value of approximately $200 million and a duration of two years. Claude thus became the first commercial AI model to be allowed into the U.S. military's classified network.
In the operations in Iran, Maven's specific functions include:
- Real-time aggregation of intelligence data from multiple sources (satellites, drones, signal interception).
- Automatically sort targets based on their importance and threat level.
- Generate precise coordinates and a list of strike recommendations
- Let human analysts focus on verification and final authorization, rather than data processing.
If a traditional intelligence system is a library, then Maven is an assistant that has already read all the books, compiled summaries, and ranked them by importance. You only need to tell it what you want, and it already knows where to look and what to find.
The decision-making process has been compressed to the extreme.
The traditional military decision-making chain is roughly as follows:
Sensors collect data → Analysts read and organize the data → Report submission → Commander assessment → Legal counsel review → Strike approval
At each stage, someone is waiting. Waiting for translation, waiting for review, waiting for reports, waiting for confirmation. This process retains a significant "decision-making buffer time," but it also means that the pace will remain at the slowest stage.
What Maven + Claude does is compress the process of "sensors collecting data → analysts reading and organizing it" from human time to machine time. AI does not replace the final strike decision, but it significantly reduces the time required to reach the decision point.
In the first 24 hours, there were 1,000 targets, approximately 41 targets per hour, averaging less than one per minute. This may not sound fast at first glance, but behind it lies the intelligence gathering, location confirmation, and threat assessment required for each target. With the assistance of AI, these tasks have been compressed to an efficiency that humans could not achieve on their own.
The speed of warfare has been ruined by AI.
It was banned and then used to bomb Iran.
Here are some of the most unusual tech news stories of 2026, which deserve a separate mention.
On February 27, 2026, the Trump administration announced a ban on Anthropic, citing "supply chain risks" and "national security threats," and ordered all federal agencies to immediately cease using Claude. The reason given was that Anthropic refused to remove Claude's security barriers, particularly those restrictions that prevented Claude from being used for autonomous weapons control and mass surveillance.
Further Reading: Anthropic CEO Angrily Denounces OpenAI's Pentagon Contract as a Complete Lie; Altman Disguised as a Peace Ambassador
At almost the same time, the U.S. military was using Claude to conduct target analysis for Operation Epic Fury.
It was banned. Then it was used to bomb people again, on the same day.
The Pentagon later explained that there was a "six-month transition period." But anyone familiar with combat systems knows that you can't "replace" a core AI system in the middle of an active military operation. Military commanders are already heavily reliant on the Maven + Claude workflow; replacement isn't a technical issue, but rather an operational risk.
This is the reality that once AI is deeply embedded in military systems: while policies can ban a company, it cannot be removed quickly in practice.
From "could be faster" to "had to be faster"
One thing is certain: the speed boost from the militarization of AI will not disappear just because a company loses a military contract.
If Claude is removed, Maven will integrate with OpenAI's models. In fact, after Anthropic was banned, OpenAI immediately announced it would take over the Pentagon's AI system. The logic of speed won't change; what will change is who provides that speed. Come on, Sam Altman isn't going to fight over a chair.
This is the core meaning that everyone should understand when commenting on this Iranian action.
Human warfare has never been static. Gunpowder changed the meaning of city walls, aircraft changed the concept of battle lines, and nuclear weapons changed the bottom line of war. Every technological leap brings both increased efficiency and magnified errors; the faster the leap, the less time there is to recover.
AI is changing the length of the decision-making chain. When a system can assist in the intelligence gathering and prioritization of 1,000 targets within 24 hours, war will no longer wait for the slowest link—human intervention. Moreover, the US military possesses fully automated weapon systems that have never been used (or at least not acknowledged), meaning weapon systems that can independently identify, make decisions, and execute strikes.
This Iranian operation may be the first time that AI-assisted warfare has been publicly revealed. The "machine speed war" is already underway, and there is no turning back.
