From AI to satellites: Anthropic, Palantir, and Starshield in the Iran raid.
On Saturday, February 28, Israel and the United States bombed several targets in Iran, striking the capital Tehran and other strategic cities. The two countries launched “major combat operations” in Iran overnight with the goal of destroying the country’s military capabilities and facilitating a regime change, announced US President Trump in a video statement. Afterwards, Trump announced the death of Ayatollah Ali Khamenei, Iran’s supreme leader, also confirmed by Iranian media.
According to US and Israeli sources cited by Axios, it was an operation of significantly greater scale than last June, when Washington struck three Iranian nuclear facilities before seeking a truce.
And to conduct the operations, the Pentagon reportedly used Claude, Anthropic’s AI, as reported by the Wall Street Journal. The US military allegedly used Claude in the attacks against Iran despite Trump’s ban the day before.
As Axios recalls, Anthropic’s AI model was used in the operation to capture Nicolás Maduro and could plausibly have been used in a potential military operation in Iran. Its artificial intelligence is integrated into the Maven Smart System, developed in collaboration with Palantir Technologies. The latter is the most controversial Californian data mining company, founded by Peter Thiel and led by Alex Karp, with deep ties to US intelligence and military agencies, on Israel’s side in the war against Hamas.
Not only that, Starshield also played a key role in the raid: the military variant of SpaceX’s Starlink ensured the continuity of operational data flow.
All the details.
THE ISSUE OF ARTIFICIAL INTELLIGENCE: CLAUDE USED DESPITE TRUMP’S BAN
“Overnight, on President Trump’s orders, the Department of War launched OPERATION EPIC FURY, the most lethal, complex, and precise air operation in history,” explained US Secretary of Defense Pete Hegseth on his X profile. The operation began at the expiration of the 10-day deadline imposed by Trump on Iran.
One of the most controversial aspects of “Epic Fury” concerns the use of artificial intelligence.
Just hours after declaring that the federal government would cease using AI tools produced by Anthropic, Trump launched a massive airstrike which – according to reports by the Wall Street Journal and Axios – made use of the Claude model.
Several US military commands, including Central Command (Centcom), indeed use Claude for analysis and decision support activities. Sources close to the dossier confirmed to the Wall Street Journal that the tool is integrated into intelligence operations. The episode highlights the difficulty of quickly “unplugging” AI systems already deeply embedded in the Pentagon’s operational infrastructure.
THE CLASH BETWEEN THE PENTAGON AND ANTHROPIC
On Friday, President Trump declared that the US government would blacklist Anthropic and that the Pentagon had declared the company a “supply chain risk,” in the most important and controversial political decision to date at the intersection of artificial intelligence and national security.
As Axios recalls, the AI company led by Dario Amodei rejected the Pentagon’s request to revoke all guarantees regarding the military’s use of its Claude model due to concerns about the use of AI for mass domestic surveillance and the development of weapons that fire without human intervention. The controversy follows tensions that had already emerged in January, when the US military reportedly used Claude in the raid to capture Venezuelan President Nicolás Maduro. Anthropic then reminded that its terms of service prohibit the use of its models for violent purposes, weapons development, or offensive surveillance.
Since then, relations between the White House, Pentagon, and the company have progressively deteriorated.
CLAUDE AT THE SERVICE OF THE PENTAGON
On February 26, Anthropic CEO Dario Amodei reiterated his refusal to allow Claude to be used for mass surveillance of US citizens or to pilot fully autonomous weapons, rejecting the Pentagon’s requests to use the model without restrictions.
Yet Claude is one of the two major generative artificial intelligence models the Pentagon has made available on classified networks, and it is the only one belonging to the group at the forefront of frontier models. The Department of Defense does not specify how it uses these models.
In 2024, Anthropic signed a partnership with Palantir Technologies and Amazon Web Services to make Claude available also in classified environments. In summer 2025, it also launched Claude Gov, a version designed for government and national security workflows.
Then comes July 2025, when Anthropic proudly announced the award of a $200 million contract from the Pentagon to accelerate the adoption of artificial intelligence by the armed forces. The company itself described it as “a new chapter in Anthropic’s commitment to supporting US national security.”
FROM DATA ANALYSIS TO TARGET SELECTION
According to press reports, in the attacks against Iran, the US armed forces employed Claude models for: intelligence assessment; target identification and prioritization; battle scenario simulation.
Centcom reportedly used the system to process large amounts of data – from intercepts to satellite images – generating operational summaries, threat assessments, and situational analyses. However, Claude reportedly did not directly control weapons systems nor make lethal decisions autonomously. Its role, according to sources, was limited to data processing and strategic support, without direct militarization.
THE RELATIONSHIP WITH PALANTIR
According to Axios, Palantir uses Anthropic’s Claude to support its most sensitive work with the US military.
THE USE OF STARSHIELD
Simultaneously, Starshield was also used for the operation. The latter is the military variant of Starlink, SpaceX’s low Earth orbit satellite constellation, the aerospace company founded by Elon Musk.
On February 28, according to a post on X by Andrea Stroppa, SpaceX advisor and Musk’s main contact in Italy, “the United States relied on Elon Musk’s Starshield for the operation in Iran. The system provided secure and fast connectivity, resisting jamming attempts. With terminals installed on drones, it ensured precise execution of the operation.”
When the Iranian Revolutionary Guard reportedly deployed jamming systems to disrupt frontline communications, Starshield allowed the operational data flow to be maintained.
In 2022, Musk’s aerospace company introduced Starshield as a new business line based on the Starlink satellite network. Starshield falls under SpaceX’s Special Projects group. “Starlink already offers unparalleled end-to-end encryption of user data,” the company states on its website. “Starshield uses additional high-security cryptographic capabilities to host classified payloads and securely process data, meeting the strictest government requirements.”
Finally, returning to the Anthropic case, the episode highlights a key point: artificial intelligence is no longer an experimental tool but a structural component of military operations.
