Anthropic Challenges Pentagon’s National Security Claims in High-Stakes AI Dispute
In a legal battle that could reshape the relationship between Silicon Valley and the U.S. military, artificial intelligence company Anthropic has vehemently pushed back against the Pentagon’s assertion that its technology poses an “unacceptable risk to national security.” Late Friday afternoon, Anthropic submitted two sworn declarations to a California federal court, asserting that the government’s case relies on technical misunderstandings and unfounded claims. The filings come ahead of a crucial hearing scheduled for Tuesday, March 24, before Judge Rita Lin in San Francisco, marking a pivotal moment in the escalating conflict between the AI firm and the Department of Defense (DoD).
The dispute traces back to late February, when the Pentagon abruptly severed ties with Anthropic after the company refused to allow unrestricted military use of its AI technology. This decision, publicly announced by Defense Secretary Pete Hegseth, was followed by the DoD’s unprecedented move to designate Anthropic as a supply-chain risk—the first such designation ever applied to an American company. Anthropic’s lawsuit alleges that this action constitutes retaliation for its publicly stated principles on AI safety, a stance the company says is protected under the First Amendment.
The declarations filed by Anthropic were authored by two key figures within the company: Sarah Heck, Head of Policy, and Thiyagu Ramasamy, Head of Public Sector. Both documents aim to dismantle the government’s narrative, arguing that the Pentagon’s concerns were never raised during months of negotiations and that the claims are rooted in technical inaccuracies.
A Clash Over AI Ethics and Military Applications
Sarah Heck, a former National Security Council official during the Obama administration, played a central role in Anthropic’s negotiations with the Pentagon. In her declaration, she directly refutes the government’s claim that Anthropic sought an “approval role” over military operations, calling it a “central falsehood.” “At no time during Anthropic’s negotiations with the Department did I or any other Anthropic employee state that the company wanted that kind of role,” Heck wrote.
Heck also highlights the Pentagon’s failure to raise concerns about Anthropic potentially disabling or altering its technology mid-operation during negotiations. Instead, she notes, these claims appeared for the first time in the government’s court filings, leaving Anthropic with no opportunity to address them.
Perhaps most striking is Heck’s revelation that on March 4—one day after the Pentagon finalized its supply-chain risk designation—Under Secretary Emil Michael emailed Anthropic CEO Dario Amodei, stating that the two sides were “very close” on resolving the very issues now cited as evidence of a national security threat. This email, attached as an exhibit to Heck’s declaration, raises questions about the Pentagon’s motivations. If the government believed Anthropic’s stance on autonomous weapons and mass surveillance posed a threat, why was an official expressing optimism about alignment on these issues immediately after the designation?
Technical Expertise Counteracts Government Claims
Thiyagu Ramasamy, Anthropic’s Head of Public Sector and a veteran of Amazon Web Services’ government AI deployments, brings a technical perspective to the case. His declaration challenges the government’s assertion that Anthropic could interfere with military operations by disabling or altering its technology. Ramasamy explains that once Anthropic’s Claude models are deployed within a government-secured, “air-gapped” system operated by a third-party contractor, the company has no access to it.
“There is no remote kill switch, no backdoor, and no mechanism to push unauthorized updates,” Ramasamy wrote. He emphasizes that any changes to the model would require explicit Pentagon approval and action to install, rendering the notion of an “operational veto” a fiction. Additionally, Ramasamy notes that Anthropic cannot access or extract data inputted by government users, further countering concerns about misuse or surveillance.
Ramasamy also addresses the government’s claim that Anthropic’s hiring of foreign nationals constitutes a security risk. He points out that Anthropic employees undergo U.S. government security clearance vetting, the same rigorous process required for access to classified information. “To my knowledge,” Ramasamy writes, “Anthropic is the only AI company where cleared personnel actually built the AI models designed to run in classified environments.”
Legal Implications and Broader Context
At the heart of the dispute lies a fundamental clash between ethical AI development and military imperatives. Anthropic has long positioned itself as a leader in responsible AI, advocating for stringent safeguards against misuse. The company’s refusal to allow unrestricted military applications of its technology reflects this commitment but has also placed it at odds with the Pentagon’s vision for AI-enhanced defense capabilities.
The Pentagon, in a 40-page filing earlier this week, rejected Anthropic’s First Amendment argument, framing the company’s stance as a business decision rather than protected speech. The government insists that the supply-chain risk designation was a straightforward national security measure, not retaliation for Anthropic’s ethical positions.
This case, however, transcends the immediate legal and technical disputes. It highlights the broader tensions between Silicon Valley’s cautious approach to AI and the military’s desire to integrate advanced technologies into defense operations. The outcome could set a precedent for how the U.S. government engages with AI companies moving forward, particularly those with firm ethical boundaries.
A Pivotal Hearing Ahead
As the hearing approaches, all eyes will be on Judge Rita Lin’s courtroom in San Francisco. The declarations submitted by Anthropic provide a robust counterargument to the government’s claims, but whether they will sway the court remains to be seen. The case underscores the complexities of balancing innovation, ethics, and national security in an era where AI’s potential—and risks—are increasingly apparent.
For Anthropic, the stakes are high. A ruling in its favor could reaffirm the importance of ethical AI development and protect companies’ rights to set boundaries on technology use. A ruling against it, however, could embolden the government to impose stricter controls on AI firms, potentially stifling innovation and ethical considerations.
As the legal drama unfolds, one thing is clear: the intersection of AI and national security is fraught with challenges, and the decisions made today will shape the future of technology, ethics, and governance. Whether this clash results in reconciliation or deeper divisions remains uncertain, but its implications will undoubtedly resonate far beyond the courtroom.
