Pentagon and Anthropic: Tensions Over National Security

A Legal Conflict Between Anthropic and the Pentagon

Anthropic, a company specializing in artificial intelligence, recently submitted two sworn statements to a federal court in California. These statements aim to contest the Pentagon's claim that the company poses an "unacceptable risk to national security." According to Anthropic, the accusations are based on technical misunderstandings and allegations that were never raised during previous negotiations.

These statements were filed alongside Anthropic's response memorandum in its lawsuit against the Department of Defense. A hearing is scheduled for this Tuesday, March 24, before Judge Rita Lin in San Francisco.

Background of the Conflict

The dispute began in late February when President Trump and Secretary of Defense Pete Hegseth publicly announced the end of relations with Anthropic. This decision came after the company refused to allow unlimited military use of its AI technology.

Sarah Heck, head of policy at Anthropic, and Thiyagu Ramasamy, head of public sector, submitted the statements. Heck, a former official at the National Security Council, worked at the White House under the Obama administration before joining Stripe and then Anthropic. She was present at the February 24 meeting where CEO Dario Amodei met with Secretary of Defense Hegseth and Pentagon Undersecretary Emil Michael.

In her statement, Heck highlights what she describes as a central falsehood in the government's documents: that Anthropic demanded an approval role over military operations. This claim, she says, is simply not true. "At no point during Anthropic's negotiations with the Department did I or any other employee of Anthropic state that the company wanted this type of role," she writes.

Heck also asserts that the Pentagon's concern about the possibility of Anthropic disabling or modifying its technology during operations was never raised during negotiations. On the contrary, she says, this concern first appeared in the government's court documents, preventing Anthropic from responding.

Internal Revelations

Another detail in Heck's statement that is likely to attract attention is that on March 4 — the day after the official designation of supply chain risk against Anthropic — Undersecretary Michael emailed Amodei to say that the two parties were "very close" on the two issues that the government now cites as evidence that Anthropic poses a national security threat: its positions on autonomous weapons and mass surveillance of Americans.

The email, which Heck attaches as an exhibit to her statement, deserves to be read alongside what Michael publicly stated in the following days. On March 5, Amodei issued a statement saying that the company had had "productive conversations" with the Pentagon. The next day, Michael posted on X that "there are no active negotiations between the Department of Defense and Anthropic." A week later, he told CNBC that there was "no chance" of renewing discussions.

Heck's point seems to be: if Anthropic's position on these two issues is what makes it a threat to national security, why did the Pentagon official say that the two parties were nearly aligned on these same issues just after the final designation? (She stops short of saying that the government used the designation as leverage, but the timeline she presents leaves the question open.)

Technical Expertise of Ramasamy

Thiyagu Ramasamy brings a different type of expertise to the case. Before joining Anthropic in 2025, he spent six years at Amazon Web Services managing AI deployments for government clients, including in classified environments. At Anthropic, he is recognized for having built the team that introduced its Claude models in national security and defense contexts, including the $200 million contract with the Pentagon announced last summer.

His statement responds to the government's claim that Anthropic could theoretically interfere with military operations by disabling the technology or altering its functioning, which Ramasamy says is not technically possible. According to him, once Claude is deployed in a government-secured system, "isolated" and operated by a third-party contractor, Anthropic does not have access; there is no "kill switch" remotely, no backdoor, and no mechanism to push unauthorized updates. Any form of "operational veto" is a fiction, he explains, noting that changing the model would require explicit approval from the Pentagon and action to install it.

Anthropic, he says, cannot even see what government users input into the system, let alone extract that data.

Ramasamy also contests the government's claim that hiring foreign nationals by Anthropic poses a security risk. He notes that Anthropic employees have undergone security clearance by the U.S. government — the same background check process required for access to classified information — adding in his statement that "to my knowledge," Anthropic is the only AI company where individuals with clearance have actually built AI models designed to operate in classified environments.

Retaliation Allegations

Anthropic's lawsuit argues that the supply chain risk designation — the first ever applied to a U.S. company — constitutes government retaliation for the company's publicly expressed views on AI safety, in violation of the First Amendment.

The government, in a 40-page document filed earlier this week, rejected this interpretation outright, asserting that Anthropic's refusal to allow all legal military uses of its technology was a business decision, not a protected expression, and that the designation was a clear call for national security and not a punishment for the company's views.