Pentagon drops Anthropic seven months after guardrails refusal
A US defence official says the department has stopped using Anthropic's tools, having branded the company a national security supply chain risk in February. The designation is the one usually aimed at firms based in hostile states.
The US Department of Defence has stopped using Anthropic's AI products, a department official said in a statement on Monday. The official told the BBC the Pentagon "has ceased the use of Anthropic products".
The removal follows a designation made in February. Defence Secretary Pete Hegseth labelled Anthropic a "national security supply-chain risk", a measure normally reserved for companies based in countries considered a threat to the United States.
The trigger was a refusal. The Pentagon had pressed Anthropic to strip the safety guardrails from Claude and give the military unrestricted access to its products, and the company declined, citing mass surveillance and autonomous weapons.
Hegseth set the phase-out order on 27 February with a six-month deadline, which pointed at late August. The cut-off slipped by about five weeks, and no reason has been given.
People familiar with the matter told the BBC that Claude was still in use as recently as last week, in research, analysis and intelligence gathering, and in military operations against Iran. Those people include former defence officials and contractors who have worked with the Pentagon on AI.
Part of the delay appears to be plumbing. Claude sat inside Maven Smart System, the Pentagon's main platform for organising intelligence data, which is operated by Palantir.
Analysts use Maven to pull together satellite imagery and drone footage, according to sources familiar with the system. That data was then fed into Claude and other large language models for tasks including identifying potential military targets.
Anthropic had been selling to the US government and military since 2024, and was the first advanced AI company with tools deployed in agencies doing classified work. That position has now been taken apart contract by contract.
The company calls the designation "unprecedented and unlawful" and has sued the Trump administration to overturn it. A judge ruled on 28 August that the administration had illegally retaliated against Anthropic.
While the case has run, the Pentagon has signed contracts with Google, xAI and OpenAI. Two people familiar with defence department AI use said OpenAI's tools have been more widely adopted across some military departments in recent months.
Relations at the top look less frosty than the procurement record suggests. Donald Trump met Anthropic's chief executive Dario Amodei twice at the White House last week and called him "great" and "fantastic".
Anthropic declined to comment on the Pentagon's statement.
For any founder selling AI into US federal procurement, the lesson is in the mechanism rather than the politics. A product term the buyer dislikes can be answered not with a cancelled contract but with a supply chain risk label, which travels across every agency at once and takes a lawsuit to shift.
The second lesson is about integration. Being embedded deep inside a customer's platform bought Anthropic months of continued use after the order to remove it, and bought the Pentagon a slower exit than it announced.
More: Anthropic