Defense Secretary Pete Hegseth has reaffirmed the classification of AI startup Anthropic as a national security risk, setting the stage for a court ruling on the issue.
The Pentagon informed the D.C. Circuit Court of Appeals on Thursday that Hegseth has rejected Anthropic's plea to reconsider its risk designation, according to POLITICO. This decision enables a three-judge panel to address the unique issues presented in Anthropic's lawsuit against the Defense Department's designation and its authority over domestic companies.
The judges had previously indicated they might delay the case until the Pentagon responded to Anthropic's request for reconsideration. However, on Wednesday, Hegseth clarified that the original supply-chain risk concerns about Anthropic's Claude stemmed from a loss of trust and other pre-deployment risks, not from any belief that Anthropic could alter the AI model in real time after deployment.
Anthropic disputed the claim, saying it does not have such post-deployment control and that Hegseth's initial assessment was based on a misunderstanding.
NSA Partnership Emerges Amid Dispute
In May, judges questioned the Pentagon's rationale, with Judge Karen Henderson stating, "For the life of me, I do not see any evidence of maliciousness despite the best efforts of [Pentagon Under Secretary Emil Michael], who in his memo refers to you as having mal-intent, a bad motive, cannot be trusted."
Meanwhile, President Donald Trump has directed the Pentagon to remove Anthropic from its systems by August, giving it a six-month deadline. However, Trump recently suggested the company's relationship with his administration was improving, indicating the ban could be reconsidered. His remarks came after a White House meeting with CEO Dario Amodei and officials to discuss cooperation and safety measures for advanced AI technologies.
However, a report by the Financial Times on Friday indicated that despite the legal dispute with the Pentagon, Anthropic is working with the U.S. National Security Agency (NSA) to deploy its advanced Mythos AI model, placing several engineers inside the agency to help tailor and implement the technology for specific missions.
While it is unclear whether the engineers are directly involved in active cyber operations, sources said the AI could support offensive cyber activities, including efforts to penetrate the networks of countries such as China and Iran.
The Pentagon and Anthropic did not immediately respond to requests for comments.
