The U.S. military used the most sophisticated artificial intelligence it has ever employed in combat to hit a scorching 1,000 targets in the first 24 hours of its attack on Iran. The Pentagon may find it challenging to abandon this weapon even as it cuts ties with the company that developed it.
According to three people familiar with the system, the military’s Maven Smart System, developed by the data mining firm Palantir, is producing insights from an astounding amount of classified data from satellites, surveillance, and other intelligence, helping to provide real-time targeting and target prioritization to military operations in Iran.
The system incorporates Anthropic’s artificial intelligence tool, Claude, which was outlawed by the Pentagon last week following contentious discussions on its application in combat.
According to two of the sources, military strategists have witnessed Claude, along with Maven, develop into a tool that is used on a daily basis throughout the majority of the military during the past year.
Maven—powered by Claude—suggested hundreds of targets, provided exact position coordinates, and ranked those targets based on significance while preparations for a possible strike in Iran were in progress. According to one of the persons, Maven and Claude’s combination has produced a weapon that is accelerating the campaign, decreasing Iran’s capacity to counterattack, and converting weeks-long war planning into real-time operations. According to the person, the AI tools also assess a strike once it has been initiated.
Additionally, Claude was utilized in the raid that captured Nicolás Maduro, the president of Venezuela, and in the fight against terrorist plans. However, two of the persons claim that this is the first significant war operation in which it has been deployed.
Despite the fact that the tool is being used to support the US military effort in Iran, Anthropic CEO Dario Amodei’s relationship with the Trump administration has deteriorated. Hours before the Iran bombing began, President Donald Trump stated that government agencies would no longer be able to use Anthropic’s technologies, giving the department six months to phase them out. The move occurred after a heated debate between the firm and the military over who would have authority over the tools’ use in mass domestic surveillance and fully autonomous weapons. The military will continue to use existing technology while it waits for a replacement to be implemented, as per sources.
According to one of the persons, military leaders have grown so reliant on the AI system that the Trump administration would utilize governmental authority to keep the technology until it could be replaced if Amodei ordered the military to stop. “Whether his morals are right or wrong or whatever, we’re not going to let Amodei’s decision making cost a single American life”.
A request for comment was not immediately answered by the Pentagon. Eduardo Maia Silva, a representative for Anthropic, declined to comment. Lisa Gordon, a representative for Palantir, declined to comment. The Iran campaign’s deployment of cutting-edge generative AI coincides with a heated discussion on the ethics and speed of employing such technologies in combat.
Paul Scharre, executive vice president at the Center for a New American Security and author of articles on AI in warfare, stated, “It is noteworthy that we’re already at the point where AI has gone from hypothetical to supporting real-world operations being conducted today.” “The U.S. military can now develop targeting packages at machine speed instead of human speed due to AI, which is a major paradigm shift.”
The drawbacks, according to him, are that “AI gets it wrong. . . We need humans to check the result of generative AI when the stakes are life and death.”
Public disclosures indicate that the Pentagon started incorporating Anthropic’s Claude chatbot into Maven in late 2024. The system has been used to deliver summaries of intelligence received from the field, track logistics, and produce suggested targets. With over 20,000 military personnel use Maven as of last May, the Trump administration has significantly increased its use in numerous other branches of the military.
The commanders presently in charge of the Iran campaign are well-versed in the use of Maven, having utilized earlier versions of the system during the US pullout from Afghanistan in 2021 and to defend Israel following the Oct. 7, 2023, assaults, according to a talk by Navy Rear Adm. Liam Hullin in 2024. Hulin, who is now Central Command’s deputy director of operations, stated at the time that the system gathered data from 179 sources.
“Centcom is heavily using MSS,” Hulin stated, referring to Maven with an acronym.
Although it was unclear at the time whether Maven’s target lists were given to the Israelis before the operation, the two sides worked together for months to decide what to strike. Soon after operations started, the Israel Defense Forces released a statement stating that they “worked for thousands of hours to build as valuable and extensive a target bank as possible in close cooperation with the U.S. Army.”
Anthropic was the first significant AI firm to work with sensitive data, and the Pentagon is attempting to use the technology to modernize its military operations. Claude was “extensively deployed” throughout the Defense Department and other security agencies, according to Amodei’s statement from last week. He was used for intelligence analysis and operation planning.
As talks with the Pentagon came to a standstill, Amodei said on her blog, “I believe deeply in the existential importance of using AI to defend the United States and other democracies, and to defeat our autocratic adversaries.”
It has been swiftly embraced. In a recent video, NATO, which inked a contract with Palantir last year, described its version of Maven as providing commanders with battle management capabilities like to those seen in video games. According to a Georgetown University evaluation of the system’s application by the Army’s 18th Airborne Corps, one artillery unit in the US military could do the tasks of 2,000 personnel with a crew of just 20.
However, Anthropic is about to be replaced at the Pentagon by other AI firms. Both Anthropic’s main competitor, OpenAI, and Elon Musk’s xAI signed a deal last week to work on classified government systems.
“The baseline use case is chat and advanced search functions — essentially summarizing information,” according to Ben Van Roo, CEO and cofounder of Legion Intelligence, a defense software startup, who has spent the last 2.5 years integrating generative AI into Department of War software systems.
According to him, it is not heavily incorporated into weapons or mission-critical systems. Although he was unaware of its application in Iran, he expressed curiosity about how it improved upon already-existing software that can prioritize targets.
The quickness of deployment in Iran, according to Scharre, impressed him. “To see this in the middle of an operation is quite remarkable,” he continued.
