The U.S. Department of Defense (DoD) has formally announced that it is discontinuing the use of Anthropic‘s artificial intelligence tools. This directive comes months after the Pentagon designated the company as a supply chain risk due to national security concerns.
Designation and Withdrawal Timeline
According to a department official speaking to the BBC on Monday, the Pentagon has completely stopped utilizing Anthropic products. Secretary of Defense Pete Hegseth had initially labeled Anthropic a supply chain risk in February and stated that the department would cease using the tools by late August. While the official announcement confirms the withdrawal, the timing has reportedly been delayed.
Multiple sources familiar with the matter told the BBC that, as recently as the previous week, the department was still using Anthropic’s Claude model. Its use spanned various activities, including research, analysis, intelligence gathering, and even military operations targeting Iran.
The Conflict Over Technology Use
Anthropic had been utilized by both the U.S. government and military since 2024, making it the first advanced AI firm whose tools were deployed within government agencies for classified work. However, tensions escalated earlier this year when the Pentagon required the company to eliminate its safety guardrails on the Claude model, giving the military unrestricted access to its AI features. When Anthropic declined, citing concerns regarding autonomous weapons and mass surveillance, Hegseth designated the firm a “national security supply-chain risk”—a classification typically reserved for companies based in nations that pose a threat to the United States.
Anthropic responded to the action, calling the designation “unprecedented and unlawful,” and subsequently initiated legal action against the Trump administration to challenge the classification.
Integration Challenges and Data Systems
The tools were crucial for processing vast amounts of information. Former U.S. defense officials and contractors specializing in AI stated that the technology was viewed as vital for sorting through extensive data sets. The Claude model, specifically, was integrated into the Pentagon’s primary intelligence and data organization platform, known as Maven Smart System, which is operated by Palantir.
Lauren Kahn, a senior research analyst at Georgetown’s Center for Security and Emerging Technology, noted that because Claude was so deeply embedded within Maven, removing it quickly would be technically difficult. Kahn, a former U.S. defense official, added that such integrated systems are not merely “plug and play,” and removal can be challenging.
Furthermore, sources indicated that analysts and defense officials rely on Maven to organize visual data, including satellite imagery and drone footage. This data is then fed into large language models, like Claude, to assist with tasks such as pinpointing potential military objectives. These systems were frequently used when compiling one-page summaries passed up the chain of command.
Aftermath and Future Contracts
In the wake of the withdrawal announcement, the Pentagon has reportedly entered into contracts with other major technology firms, including Google, OpenAI, and xAI. It has been reported by two people familiar with AI use in the defense department that OpenAI tools have seen wider adoption in some military departments recently.
The announcement followed a BBC query regarding the continued, widespread use of Claude by the department, despite earlier official statements. The Pentagon’s decision to phase out Anthropic came after Hegseth set a six-month deadline on February 27, the day before the U.S. and Israel launched attacks against Iran. Despite this directive, employees continued to use Anthropic’s Claude tools, including its powerful Mythos models, throughout the period of controversy.
The conflict between the Pentagon and Anthropic occurs as high-level relations appear to be improving. Last week, Donald Trump met with Dario Amodei, the head of Anthropic, twice at the White House. During one of these meetings, Amodei stated that while he still had concerns about the risks posed by AI, he was confident that working with the president would allow them to “win safely,” saying, “How we address those risks is still under discussion.”
