Skip to content

Anthropic

Pentagon quits Claude, which sources say ran until last week

5 min read

Written by an AI model from the linked sources and checked against them by a second model. How we work

Image: bbc.co.uk

The US Defense Department has stopped using Anthropic’s products, a department official told the BBC on October 5. That is more than a month after the late-August cutoff Defense Secretary Pete Hegseth set, and multiple people familiar with the matter said Claude was still in use as recently as last week for research, analysis and intelligence gathering, and in military operations against Iran.

The official’s statement that the Pentagon “has ceased the use of Anthropic products” answered a question from the BBC, after several people had said in interviews that Claude was still widely used. It is not clear why the phase-out ran late.

Inside Maven

Claude was embedded in Maven Smart System, the Palantir-run platform the Pentagon uses as its primary tool for organizing intelligence and other data, multiple sources told the BBC. Analysts use Maven to pull together visual data such as satellite imagery and drone footage. That data was then fed into Claude and other large language models for tasks including identifying potential military targets. Officials often used the setup to compile one-page briefs passed up the chain of command, according to a source who was given a Maven demonstration by Palantir earlier this year. Pentagon employees kept using Claude, including Anthropic’s Mythos models, “throughout the whole period of the controversy,” one person said.

Lauren Kahn, a senior research analyst at Georgetown’s Center for Security and Emerging Technology and a former US defense official, said that deep integration would have made Claude hard to pull out quickly. “The fact that it took the Pentagon until now to remove all use of Claude from across the system is indicative that these things are not just plug and play,” she said. “Once they become integrated it can be painful to remove them.”

The Pentagon’s own account in September was different. Under Secretary of Defense Emil Michael told reporters on September 10 that “about 90%” had transitioned, that Palantir’s Maven Smart Systems work had moved off Anthropic months earlier, and that the rest of the military’s classified systems would follow by the end of the month. The BBC’s sources describe Claude as part of Maven and in use across the department until last week.

Replacements were already lined up. The Pentagon has signed contracts with Google, xAI and OpenAI, and two people familiar with military AI use told the BBC that OpenAI tools have been more widely adopted in some military departments in recent months. When Hegseth first ordered the phase-out, “there was concern with moving beyond Anthropic to OpenAI” and other tools, a former senior US defense official said.

The six-month window was itself contested. Just Security noted that Claude was “extensively deployed” across US military and intelligence systems, and argued that allowing up to six months of continued use undermined the claim of a true supply chain risk. By the BBC’s sources’ account, use ran more than a month beyond that window.

How the dispute got here

Anthropic’s tools had been used by the US government and military since 2024, and Anthropic was the first advanced AI company to have its tools deployed in agencies doing classified work. The break came when the Pentagon pressed it to drop Claude’s safety guardrails and give the military unfettered access; Anthropic refused, citing mass surveillance and autonomous weapons. The supply chain risk label it got in return is typically reserved for companies based in countries that pose a threat to the US.

  • February 24: Hegseth gives Dario Amodei until 5:01 p.m. on February 27 to allow unrestricted use of Claude “for all legal purposes,” per CBS News.
  • February 27: Trump directs federal agencies to stop using Anthropic’s products, and Hegseth designates the company a supply chain risk, with up to six months for the Defense Department to transition. A day later, the US and Israel attack Iran.
  • March 24: At a hearing, Judge Rita Lin says, “If the worry is about the integrity of the operational chain of command, [the Department of Defense] could just stop using Claude,” per Tech Policy Press.
  • March 26: Lin grants a preliminary injunction blocking the ban and the designation, finding the government’s retaliatory actions likely violated the law, Tech Policy Press reports.
  • April 8: A D.C. Circuit panel denies Anthropic’s stay request, citing the need not to force the military to prolong dealings with an unwanted vendor of critical AI services during a significant ongoing conflict, according to Tech Policy Press.
  • June 4: The Pentagon tells the D.C. Circuit that Hegseth has denied Anthropic’s request to reconsider the designation, clearing the way for a three-judge panel, per POLITICO.
  • September: A D.C. Circuit majority of Judges Gregory Katsas and Neomi Rao rejects Anthropic’s challenge, reading a 2018 supply chain law broadly.

Warmer at the White House

Anthropic has called the designation “unprecedented and unlawful,” and its lawsuit against the Trump administration is still running. Relations with the White House, though, appear to be warming: Trump met Amodei twice at the White House last week. After one of those meetings, held with other technology leaders, Amodei said he still had concerns about the risks AI poses. “How we address those risks is still under discussion,” he told reporters with Trump standing beside him, but “if we work with the president, we can win safely.”

Trump then called Amodei “great” and “fantastic.”

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.