Pentagon Says It Stopped Using Anthropic's Claude, Six Weeks Past Its Own Deadline

The Pentagon says it finally did it. A Defense Department official told the BBC on Monday that the military has “ceased the use of Anthropic products.” That ends, on paper at least, the department’s long and awkward entanglement with Claude.

The statement arrived more than six weeks after Defense Secretary Pete Hegseth’s own deadline. Hegseth set a six-month phase-out clock on Feb. 27, putting the cutoff in late August. Instead, sources told the BBC that Claude was still being used by Pentagon staff as recently as last week, including in military operations against Iran.

It was, by multiple accounts, a quiet defiance of the secretary’s order. The department had blacklisted one of Silicon Valley’s most prominent AI labs as a national security risk and kept using its software anyway.

How the Pentagon Anthropic Claude split started

The break traces back to a demand Anthropic wouldn’t meet. Earlier this year, the Pentagon pressed the company to remove 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 then designated Anthropic a “national security supply-chain risk,” a legal category normally reserved for companies based in countries that threaten the US.

Anthropic called the move “unprecedented and unlawful” and sued the Trump administration to overturn the designation. That lawsuit is still working through the courts; in September, an appeals court upheld the Pentagon’s blacklist designation.

Anthropic was, notably, the first advanced AI company to have its tools deployed in US government agencies doing classified work, dating back to 2024. The February designation made it the first to be thrown out over a fight about how those tools should be allowed to behave.

The divorce took longer than the paperwork.

Claude was still in the building last week

Multiple people familiar with the matter told the BBC that as recently as last week, Pentagon staff were still using Claude for research, analysis, and intelligence gathering, including during recent military engagements against Iran. It was, one source said, a critical tool for sorting through vast quantities of information. Another said Pentagon employees used Claude’s models “throughout the whole period of the controversy.”

The date on the order matters. Hegseth set the phase-out deadline on Feb. 27, one day before the US and Israel attacked Iran, according to the BBC’s reporting. Claude then stayed embedded through months of live operations.

An Anthropic spokesman declined to comment when asked about the Pentagon’s claim that it had stopped using the company’s products.

The reason it took so long: Maven

Claude wasn’t just sitting on someone’s laptop. It was wired into Maven Smart System, the Pentagon’s primary platform for organizing intelligence and other data, operated by Palantir. Analysts relied on Maven to access and organize visual data — satellite imagery, drone footage and fed it into Claude and other large language models for tasks like identifying potential military targets and compiling the one-page briefs that move up the chain of command.

Maven is deployed “across the entire department,” defense AI lead Cameron Stanley said during a demonstration earlier this year. Several US allies use it to share intelligence with the US as well.

Ripping a frontier model out of that system turned out to be an engineering problem, not a memo.

“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,” Lauren Kahn, a senior research analyst at Georgetown’s Center for Security and Emerging Technology and a former defense official, told the BBC. “Once they become integrated it can be painful to remove them.”

A former senior US defense official told the BBC there had also been unease about what came next: “there was concern with moving beyond Anthropic to OpenAI” and other tools.

The replacements are already here

While the legal fight has played out, the Pentagon has been signing contracts with other AI labs. Last July, the Defense Department awarded contracts worth up to $200 million each to Anthropic, Google, OpenAI, and xAI to push advanced AI into military use. With Anthropic on the way out, Google, xAI, and OpenAI have filled the vacuum, and OpenAI’s tools, two people familiar with AI use in the department said, have been more widely adopted across some military departments in recent months.

The optics at the top are strange. Anthropic CEO Dario Amodei met twice with President Trump at the White House last week. Trump praised him afterward as “great” and “fantastic.” Amodei, standing beside Trump, said “how we address those risks is still under discussion,” but “if we work with the president, we can win safely.”

So the president is praising Anthropic’s CEO in the White House while the Pentagon tells the BBC it has cut the company’s last connection. Welcome to 2026.

Why this matters

This is the first time a US AI lab has been formally blacklisted by the Pentagon, and the first hard test of what happens when a frontier model becomes infrastructure. Claude wasn’t hard to remove because the military dragged its feet. It was hard to remove because it worked too well and sat too deep in the system, inside the primary intelligence platform, during active operations. Every other AI lab courting defense contracts just got a preview of how sticky this business is: once your model is the one reading the satellite footage, you’re not a vendor anymore. You’re load-bearing.

There’s also a quieter lesson for AI safety. Anthropic lost its biggest government customer because it wouldn’t hand over the keys, and the customer kept using the product anyway for eight months. The guardrails were apparently valuable enough to keep, right up until the system could be rebuilt without them. Watch whether that holds: the Pentagon’s statement to the BBC says the cessation is complete, but the department declined to say when exactly the last use ended. The lawsuit over the designation is still live. And the White House, at least for now, still takes Amodei’s meetings.

FAQ

Did the Pentagon stop using Anthropic’s Claude?

Yes — at least officially. A Pentagon official told the BBC on Oct. 5 that the department “has ceased the use of Anthropic products.” That came more than six weeks after Defense Secretary Pete Hegseth’s late-August phase-out deadline.

Why did the Pentagon cut ties with Anthropic?

The Pentagon demanded Anthropic strip Claude’s safety guardrails and grant the military unfettered access. Anthropic refused over concerns about mass surveillance and autonomous weapons, and Hegseth labeled the company a national security supply-chain risk in February. Anthropic called the designation “unprecedented and unlawful” and sued.

Why did removal take longer than the deadline?

Claude was deeply embedded in Maven Smart System, the Palantir-operated platform that organizes Pentagon intelligence data, and was still in use as recently as last week, including in operations against Iran. Georgetown analyst Lauren Kahn called it “indicative that these things are not just plug and play”: once a model is integrated, pulling it out is painful engineering, not paperwork.

Which AI companies is the Pentagon using now?

The Pentagon has signed contracts with Google, xAI, and OpenAI, originally part of a July 2025 round of awards worth up to $200 million each, and OpenAI’s tools have been more widely adopted across some military departments in recent months as Anthropic was phased out.

Sources: BBC News, Techstrong.ai, Android Headlines.