The US Department of Defense no longer uses AI tools from Anthropic, a department official confirmed in a statement. The Pentagon, according to the statement, has "stopped using Anthropic's products", with the statement following a response to a question from the BBC about the use of these tools.
The US Secretary of Defense, Pete Hegseth, designated Anthropic a supply chain risk in February, and announced that the Pentagon would stop using the company's tools by the end of August. A six-month phase-out period was set for 27 February, the day before the US and Israel attacked Iran. It is not clear why the department has now stopped using the tools, and details of the complete withdrawal have not previously been published.
Several people familiar with the matter told the BBC that the company's Claude model was still being used by the department in intelligence gathering, analysis and military operations against Iran last week. Claude was considered a key tool for sorting through vast amounts of information. One source said that Pentagon staff continued to use Claude, including its powerful Mythos models, "throughout the period of controversy" with the company.
Claude was embedded in a larger data platform used by the Pentagon, known as the Maven Smart System, several sources said. Maven, managed by Palantir, is the Pentagon's primary platform for organising intelligence and other data. Sources familiar with the way the technology is used said that analysts and other defence officials rely on Maven to access and organise visual data, including satellite and drone images, with the data then fed to large language models for tasks such as identifying potential military targets. Officials have often used this technology when compiling one-sided summaries to be sent up the chain of command, according to a source who had previously seen a demonstration of Maven by Palantier.
Lauren Kahn, a senior research analyst at Georgetown University's Center for Security and Emerging Technology, said that due to the deep integration of Claude, it would be difficult to remove it quickly from the system. She added that the fact that it took the Pentagon this long to remove all use of Claude from the system showed that such things were not "plug and play" and that "once integrated, removing them can be painful".
The US government had been using Anthropic since 2024, and the company was the first advanced AI company whose tools were deployed in government agencies working on classified work. Earlier this year, the Pentagon pressured Anthropic to remove safety guardrails from its models, allowing the military unrestricted access to its AI tools. Anthropic rejected this, citing concerns about mass surveillance and autonomous weapons, following which Hegseth described the company as a "risk to national security supply chains". Anthropic described the move as "unprecedented and illegal" and sued the Trump administration in an attempt to overturn the designation.
The Pentagon has signed contracts with other technology companies, including Google, xAI and OpenAI. OpenAI tools have in recent months been more widely accepted in some military departments, according to two people familiar with the use of AI in the Department of Defense. When Hegseth initially ordered the phase-out of Anthropic from the government, there was concern about the shift from Anthropic to OpenAI and other tools, a former US defence official said.
The Anthropic spokesperson declined to comment on the Pentagon's statement. The BBC spoke to defence contractors specialising in AI who have worked closely with the Pentagon, as well as former US defence officials and others familiar with the use of artificial intelligence in the Pentagon. Several spoke on condition of anonymity as they were not authorised to speak to the media.
Trump met with Anthropic's head, Dario Amodei, twice last week at the White House. Following one of the meetings with Trump and other technology leaders, Amodei said he had concerns about the risks posed by AI. He told reporters: "How we deal with the risks posed by AI is still a matter of debate", while Trump stood beside him, but "if we work with the president, we can win safely". Trump then praised Amodei, saying he was "great" and "fantastic".
