Summary: The Pentagon has confirmed it has stopped using Anthropic's AI tools after Secretary Pete Hegseth declared the company a national security supply‑chain risk in February and ordered a phase‑out. Sources told the BBC Claude remained embedded in Palantir's Maven intelligence platform until recently and was used for analysis, intelligence gathering and operations. Anthropic refused Pentagon demands to remove safety guardrails, citing risks of mass surveillance and autonomous weapons, and has sued to overturn the designation. The department is shifting contracts toward other firms, including Google, xAI and OpenAI.
Pentagon Confirms It Has Stopped Using Anthropic’s AI After National Security Blacklist
Defence secretary Pete Hegseth has publicly criticised Anthropic's boss Dario Amodei [Reuters]
The US Department of Defense has confirmed it has stopped using Anthropic's AI tools, a Pentagon official told the BBC, months after Secretary of Defense Pete Hegseth designated the company a national security supply‑chain risk.
Official Statement and Context
"The Pentagon has ceased the use of Anthropic products," the department said in a statement.
Hegseth labeled Anthropic a supply‑chain risk in February and ordered the department to phase out the company's tools by late August. Despite that directive, multiple people familiar with Pentagon operations told the BBC that Anthropic's Claude model was still in use as recently as last week for research, analysis, intelligence gathering and even operations related to Iran.
How Claude Was Used
Sources said Claude — including its Mythos family of models — had become a critical tool for sorting and analyzing large volumes of data. The model was reportedly embedded in the Maven Smart System, a data‑management platform operated by Palantir that the Pentagon uses to organise intelligence such as satellite imagery and drone footage. Analysts fed organised visual and textual data into Claude and other large language models for tasks including target identification and drafting concise briefs for senior decision‑makers.
Why the Break With Anthropic
Earlier this year the Pentagon requested that Anthropic remove certain safety guardrails from Claude and provide broader, less restricted access for military use. Anthropic refused, citing concerns that such changes could enable mass surveillance and the development of autonomous weapons. In response, Hegseth designated the company a "national security supply‑chain risk," a label typically reserved for firms tied to countries judged to threaten US security.
Trump recently met with Anthropic boss Dario Amodei at the White House [EPA]
Anthropic has called the designation "unprecedented and unlawful" and has filed suit seeking to overturn it. The legal battle between the company and the US government is ongoing.
What Comes Next
As the dispute played out in court, the Pentagon moved to diversify its AI suppliers and has signed contracts with other major tech firms, including Google, xAI and OpenAI. Two people familiar with Defence Department AI use told the BBC that OpenAI tools have been adopted more widely in some branches in recent months.
Experts say removing deeply integrated tools is technically and operationally difficult. Lauren Kahn, a senior research analyst at Georgetown’s Center for Security and Emerging Technology and a former US defence official, noted that once a model is embedded in systems like Maven "it can be painful to remove them." The Pentagon has not publicly explained the precise timing or operational steps behind the complete cutoff of Anthropic tools.
Political Backdrop
The personnel and political context remains notable: Anthropic CEO Dario Amodei met President Trump twice at the White House recently, and reports suggest warming ties even as the company fights the Pentagon's designation in court.
Reporting note: The BBC reported the original details and spoke with current and former defense officials, contractors and analysts, some on condition of anonymity because they were not authorised to discuss internal operations.
Help us improve.
