The US Department of Defence has discontinued its use of Anthropic's AI tools, a department official confirmed to the BBC, several months after labelling the company a national security risk and ordering a phase-out.
Defence Secretary Pete Hegseth designated Anthropic a supply chain risk in February and set a deadline for the Pentagon to stop using the company's products by late August. The department official stated in a response that the Pentagon "has ceased the use of Anthropic products."
However, multiple sources with direct knowledge of Pentagon operations told the BBC that Anthropic's Claude model remained in active use across the department as recently as the previous week, supporting research, analysis, intelligence gathering and military operations against Iran. The tool was regarded as essential for processing enormous volumes of information, according to former defence officials and contractors with expertise in Pentagon AI systems.
Anthropic declined to comment when contacted about the Pentagon's statement regarding the discontinuation of its products.
Why did the Pentagon blacklist Anthropic?
The Pentagon's decision to cease using Anthropic stemmed from the company's refusal to remove safety protections from Claude. Earlier in the year, military officials pressed Anthropic to strip away its guardrails and grant the military unrestricted access to its AI capabilities. Anthropic declined, citing concerns about mass surveillance and the development of autonomous weapons systems. In response, Hegseth invoked a national security designation typically reserved for firms based in adversarial nations, labelling Anthropic a "national security supply-chain risk."
Anthropic characterised the move as "unprecedented and unlawful" and initiated legal action against the Trump administration to challenge the designation. The company has continued to contest the decision in court as the Pentagon signed new contracts with competing technology firms, including Google, xAI and OpenAI.
How was Claude integrated into Pentagon systems?
Claude had been embedded within Maven Smart System, the Pentagon's primary platform for organising intelligence and other data, according to multiple sources. Maven is operated by Palantir, a defence technology company founded by billionaire Peter Thiel, a prominent Trump ally. The deep integration of Claude into Maven created significant technical challenges for removal.
Lauren Kahn, a senior research analyst at Georgetown's Center for Security and Emerging Technology and former defence official, explained the difficulty:
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. Once they become integrated it can be painful to remove them.
Defence analysts and officials relied on Maven to access and organise visual data, including satellite imagery and drone footage. This information was then processed through Claude and other large language models to identify potential military targets and support other analytical tasks. Officials frequently used the technology when preparing briefings for senior command, according to a source who received a demonstration of Maven from Palantir earlier this year.
Cameron Stanley, who leads AI work within the defence department, stated during a Maven demonstration that the system is deployed "across the entire department." Recent private demonstrations of Maven for government officials have also included Claude, according to another person familiar with the platform.
What was the timeline of the phase-out?
Hegseth issued the six-month phase-out order on 27 February, one day before the US and Israel launched attacks on Iran. Despite this directive, Pentagon employees continued using Claude tools, including its advanced Mythos models, throughout the entire period of the dispute with the company, according to one source.
According to reporting from defence industry sources, the Pentagon stated that approximately 90% of classified systems had transitioned away from Anthropic models, with completion expected by the end of September.
The Pentagon's statement to the BBC on Monday came after multiple sources reported in interviews that Claude was still being widely used in recent weeks. The delay in full removal underscores the technical complexity of extracting deeply integrated AI systems from military infrastructure.
What legal developments have occurred?
The legal battle between Anthropic and the Trump administration has continued to evolve. A federal appeals court upheld the Pentagon's designation of Anthropic as a supply-chain risk on 25 September in a 2–1 decision, affirming a separate designation from one being challenged in California. The appeals court delayed the ruling from taking immediate effect to allow Anthropic time to seek a rehearing or petition for review by the full D.C. Circuit.
In a separate development, the Pentagon filed an appeal on 4 October against a California ruling that had barred federal agencies from designating Anthropic a supply-chain risk and removing its technology from federal systems. Anthropic may petition for a rehearing following the appeals court's decision, with the delayed implementation providing time for such requests.
What contracts replaced Anthropic?
As the legal dispute unfolded, the Pentagon signed contracts with other major technology companies. In July 2025, the Pentagon awarded contracts to four AI companies—Anthropic, Google, OpenAI and xAI—each valued at up to $200 million to accelerate adoption of advanced AI for national-security applications. OpenAI tools have been more widely adopted in recent months in some military departments, according to two people familiar with AI use in the defence department.
A former senior defence official noted that when Hegseth initially ordered the phase-out of Anthropic, "there was concern with moving beyond Anthropic to OpenAI" and other tools. Maven is also used by some US allies to share intelligence with the Trump administration, and a current defence department official highlighted Maven's role in intelligence sharing during the conflict with Iran in a private briefing with industry leaders in Washington last week, according to a former US official who attended the meeting.
How have relations between Anthropic and the White House evolved?
Despite the Pentagon's designation and the ongoing legal battle, relations between Anthropic and the Trump administration appear to be improving. Trump met with Anthropic Chief Executive Dario Amodei twice at the White House last week. Following one of these meetings with Trump and other technology leaders, Amodei expressed that he retained concerns about risks posed by AI.
How we address those risks is still under discussion,Amodei told reporters at the White House, with Trump standing nearby, adding that
if we work with the president, we can win safely.Trump subsequently praised Amodei, describing him as "great" and "fantastic."
What happens next?
Anthropic has the opportunity to petition for a rehearing or seek review by the full D.C. Circuit following the appeals court's 25 September decision. The delayed implementation of that ruling provides time for such requests. The Pentagon continues its transition away from Anthropic systems, with the company's future role in government AI applications remaining uncertain as the legal proceedings continue.
