Pentagon Cuts Ties with Anthropic AI Following National Security Risk Label
After months of scrutiny, the US Department of Defense has officially discontinued the use of Anthropic’s AI technologies. This action follows the company’s prior recognition as a key supplier for advanced military data analysis, intelligence, and operational support—roles that Anthropic will no longer occupy since being designated as a supply chain risk due to national security concerns.
Pentagon Ends Use After Extended Phase-Out
On Monday, a Pentagon spokesperson stated the Department of Defense “has ceased the use of Anthropic products.” The move follows Defense Secretary Pete Hegseth’s formal identification of Anthropic as a supply chain risk in February, with a deadline imposed for complete disengagement by late August. Despite this, insiders report that the Claude AI model was actively serving military intelligence and operational goals—including missions targeting Iran—as recently as last week.
Anthropic initially held a unique status as the first advanced AI supplier to US government entities with classified mandates. Its influential Claude and Mythos models allowed analysts to process huge intelligence databases, facilitating tasks like visual organizing of information and target spotting. The Pentagon’s Maven platform, operated by Palantir, regularly depended on these Anthropic tools.
Safety Guardrails Spark Controversy
Tensions escalated between the Pentagon and Anthropic earlier this year when the military pushed for Claude’s safety restrictions to be lifted. Anthropic leadership refused this request, citing serious concerns about enabling mass surveillance and AI-powered autonomous weapons. As a counter, Secretary Hegseth classified Anthropic as a national security supply chain risk—a rare move for a US-based firm, as such labels are typically reserved for foreign adversaries.
Challenging this decision, Anthropic responded with a lawsuit against the Trump administration, describing the order as “unprecedented and unlawful.” The company’s legal contestation forms part of larger ongoing disputes with the Pentagon. Further escalating the matter, a recent federal decision ruled that the administration had illegally retaliated against Anthropic.
Pentagon Shifts to Alternative AI Providers
In the interim, the Pentagon signed agreements with leading tech companies such as Google, xAI, and OpenAI. With Anthropic’s exit underway, according to sources with Defense Department insight, OpenAI’s products have recently been adopted more broadly across certain Pentagon units.
The complex nature of removing Anthropic systems from military architecture proved challenging. Lauren Kahn, a senior research analyst at Georgetown’s Center for Security and Emerging Technology, described the Claude model as deeply woven into core defense platforms like Maven, complicating the extraction process. Kahn emphasized the “pain” inherent in such technical and operational untangling as a reason for the transition’s extended timeline.
According to additional sources, Claude had been a backbone in the Maven Smart System, a primary Pentagon hub for integrating satellite and drone data along with generating rapid-response intelligence summaries. Various US allied partners also utilized Maven in coordination with the Trump administration for intelligence sharing.
Anthropic and White House Engagement
Though legal conflicts persist and the Pentagon has officially separated from Anthropic, the company’s dialogue with the White House appears to be on a more positive trajectory. Last week, President Trump conducted two meetings with Anthropic CEO Dario Amodei at the White House. Speaking to the press afterward, Amodei stressed continued vigilance over AI risks but acknowledged a willingness to collaborate, stating, “if we work with the president, we can win safely.” President Trump, during the event, commended Amodei’s leadership.
Anthropic said the risk designation is “unprecedented and unlawful” and is seeking to challenge the decision in court. To explore further details about the legal disputes and technical issues associated with defense AI usage, see the related articles linked below.