Pentagon quits Claude, which sources say ran until last week

As recently as last week, Claude was still helping the US military sort intelligence, including for operations against Iran, according to several people who spoke to the BBC. On Monday a Defense Department official said the Pentagon "has ceased the use of Anthropic products." That came months after Defense Secretary Pete Hegseth set a late August deadline for dropping them.
At a glance
- Anthropic's trouble began when it refused Pentagon demands to remove Claude's safety guardrails, citing mass surveillance and autonomous weapons, and Hegseth answered by designating the company a national security supply-chain risk.
- Claude ran inside Maven Smart System, the Palantir-operated platform where analysts organize satellite imagery and drone footage, then pass it to language models for tasks such as identifying potential military targets.
- It is not clear why removal ran past the late August deadline, while Anthropic's lawsuit calling the designation "unprecedented and unlawful" continues to play out in the courts.
If you have not followed the dispute, here is the background. The US government and military have used Anthropic's tools since 2024, and Anthropic was the first advanced AI company to have its tools deployed in government agencies doing classified work. Earlier this year the Pentagon pressed it to strip Claude's safety guardrails and give the military unfettered access to its AI products. Anthropic refused, citing concerns about mass surveillance and autonomous weapons.
Hegseth's February designation came from a law usually aimed at companies in countries that threaten the US
After the refusal, Hegseth designated Anthropic a "national security supply-chain risk." That designation comes from a law typically reserved for companies based in countries that pose a threat to the US. On 27 February he set a six-month deadline to phase Anthropic out of the Pentagon, which put the cutoff in late August. The order came one day before the US and Israel attacked Iran.
Anthropic called the move "unprecedented and unlawful" and sued the Trump administration to overturn the designation. While that fight has played out in court, the Pentagon has signed contracts with Google, xAI and OpenAI. Two people familiar with AI use in the defense department told the BBC that OpenAI tools have been more widely adopted in some military departments in recent months.
Claude sat inside Maven Smart System, the Pentagon's primary platform for organizing intelligence
Palantir operates Maven Smart System, the Pentagon's primary platform for organizing intelligence and other data. Some US allies also use it to share intelligence with the Trump administration. Multiple sources told the BBC that Claude was embedded in it.
People familiar with how Maven works describe a pipeline. Analysts and other defense officials use Maven to access and organize visual data, including satellite imagery and drone footage. That data was then fed into Claude and other large language models for tasks such as identifying potential military targets.
A source who was given a demonstration of Maven by Palantir earlier this year said officials often used the technology to compile one-page briefs that were passed up the chain of command. Think of an appliance built into a fitted kitchen. The cooks use the kitchen rather than the appliance on its own, so taking the appliance out means opening up the cabinets around it.
Sources say Claude, including the Mythos models, was in use until last week
The Pentagon's Monday statement was its answer to a BBC question, asked after several people said in interviews that Claude was still widely used. Among them were former US defense officials and contractors who have worked closely with the Pentagon on AI. They called Claude a critical tool for sorting through vast quantities of information and said it was used in research, analysis, intelligence gathering and military operations against Iran.
One said Pentagon employees kept using Claude, including Anthropic's powerful Mythos models, "throughout the whole period of the controversy." A former senior defense official recalled that when Hegseth first ordered the phase-out, "there was concern with moving beyond Anthropic to OpenAI" and other tools.
Lauren Kahn is a senior research analyst at Georgetown's Center for Security and Emerging Technology and a former US defense official. She said Claude's deep integration into Maven would have made it hard to remove quickly. "These things are not just plug and play," she said. "Once they become integrated it can be painful to remove them."
Trump met Dario Amodei twice at the White House last week
The cutoff arrives as relations between Anthropic and the White House appear to be warming, even though Hegseth has publicly criticized Anthropic's chief, Dario Amodei. Trump met Amodei twice at the White House last week. After one meeting with Trump and other technology leaders, Amodei said he still had concerns about the risks posed by AI.
"How we address those risks is still under discussion," Amodei told reporters as Trump stood by his side, adding that "if we work with the president, we can win safely." Trump then praised Amodei, calling him "great" and "fantastic." The Pentagon's statement on Monday came days after those meetings.
The BBC reporting leaves two things open: why the removal ran past the late August deadline, and what, if anything, took Claude's place inside Maven. In our view, the telling detail is the design. Claude reached Pentagon analysts as a component wired into a Palantir platform, so an order against one vendor turned into a slow extraction job rather than a switch someone could flip.
What Anthropic's lawsuit could still change
Anthropic's challenge to the supply-chain-risk designation is still in the courts, and the BBC report gives no date for a ruling. If the designation is overturned, we don't know whether Claude would go back into Maven or how long reintegration would take. It is also unclear whether last week's White House meetings will affect the designation at all.
Related stories
- D.C. Circuit calls Claude's refusals a supply chain risk
- Pentagon to finish cutting Anthropic loose by month's end
- Claude flagged a diary entry and a human sent it to police
- FTC plans to make Anthropic and OpenAI executives testify
- Australian senators want Amodei and Altman in Canberra
- Trump invites Amodei to a private White House dinner
Comments
No comments yet. Be the first.
Join the conversation
Sign in with Google to leave a comment. Your name and avatar come from your Google profile, and the comment appears after moderation.
