New Court Filing Reveals Pentagon Told Anthropic The Two Sides Were Nearly Aligned — A Week After Trump Declared The Relationship Kaput

Anthropic submitted 2 sworn declarations to a California national tribunal precocious Friday afternoon, pushing backmost connected nan Pentagon’s assertion that nan AI institution poses an “unacceptable consequence to nationalist security” and arguing that nan government’s lawsuit relies connected method misunderstandings and claims that were ne'er really raised during nan months of negotiations that preceded nan dispute.

The declarations were revenge alongside Anthropic’s reply little successful its suit against nan Department of Defense and travel up of a proceeding this coming Tuesday, March 24, earlier Judge Rita Lin successful San Francisco.

The conflict traces backmost to precocious February, erstwhile President Trump and Defense Secretary Pete Hegseth publically declared they were cutting ties pinch Anthropic aft nan institution refused to let unrestricted subject usage of its AI technology.

The 2 group who submitted nan declarations are Sarah Heck, Anthropic’s Head of Policy, and Thiyagu Ramasamy, nan company’s Head of Public Sector.

Heck is simply a erstwhile National Security Council charismatic who worked astatine nan White House nether nan Obama management earlier moving to Stripe and past Anthropic, wherever she runs nan company’s authorities relationships and argumentation work. She was personally coming astatine nan February 24 gathering wherever CEO Dario Amodei sat down pinch Defense Secretary Hegseth and nan Pentagon’s Under Secretary Emil Michael.

In her declaration, Heck calls retired what she describes arsenic a cardinal falsehood successful nan government’s filings: that Anthropic demanded immoderate benignant of support domiciled complete subject operations. That claim, she says, simply isn’t true. “At nary clip during Anthropic’s negotiations pinch nan Department did I aliases immoderate different Anthropic worker authorities that nan institution wanted that benignant of role,” she wrote. She besides points retired that nan Pentagon’s interest astir Anthropic perchance disabling aliases altering its exertion mid-operation was ne'er raised during negotiations. Instead, she says, it appeared for nan first clip successful nan government’s tribunal filings, which gave Anthropic nary opportunity to respond.

Another item successful Heck’s declaration judge to tie attraction is that connected March 4 — nan time aft nan Pentagon formally finalized its supply-chain consequence nickname against Anthropic — Under Secretary Michael emailed Amodei to opportunity nan 2 sides were “very close” connected nan 2 issues nan authorities now cites arsenic grounds that Anthropic is simply a nationalist information threat: its positions connected autonomous weapons and wide surveillance of Americans.

The email, which Heck attaches arsenic an grounds to her declaration, is notable for different reason. Two days later, aft Amodei said nan institution was having “productive conversations” pinch nan Pentagon, Michael posted connected X that “there is nary progressive Department of War speech pinch Anthropic.” A week aft that, he told CNBC location was “no chance” of renewed talks. Heck’s constituent appears to be: If Anthropic’s stance connected those 2 issues is what makes it a nationalist information threat, why was nan Pentagon’s ain charismatic saying nan 2 sides were astir aligned connected precisely those issues conscionable days aft nan nickname was finalized?

Ramasamy brings a different benignant of expertise to nan case. Before joining Anthropic successful 2025, he spent six years astatine Amazon Web Services managing AI deployments for authorities customers, including classified environments. At Anthropic, he’s credited pinch building nan squad that brought its Claude models into nationalist information and defense settings, including nan $200 cardinal contract pinch nan Pentagon announced past summer.

His declaration takes connected nan government’s declare that Anthropic could theoretically interfere pinch subject operations by disabling nan exertion aliases different altering really it behaves, which Ramasamy says isn’t technically possible. Per his telling, erstwhile Claude is deployed wrong a government-secured, “air-gapped” strategy operated by a third-party contractor, Anthropic has nary entree to it; location is nary distant termination switch, nary backdoor, and nary system to push unauthorized updates. Any benignant of “operational veto” is simply a fiction, he suggests, explaining that a alteration to nan exemplary would require nan Pentagon’s definitive support and action to install. Anthropic, he says, can’t moreover spot what authorities users are typing into nan system, fto unsocial extract that data.

Ramasamy besides disputes nan government’s declare that Anthropic’s hiring of overseas nationals makes nan institution a information risk. He notes that Anthropic labor person undergone U.S. authorities information clearance vetting — nan aforesaid inheritance cheque process required for entree to classified information, adding successful his declaration that “to my knowledge,” Anthropic is nan only AI institution wherever cleared unit really built nan AI models designed to tally successful classified environments.

Anthropic’s suit argues that nan supply-chain consequence nickname — nan first ever applied to an American institution — amounts to authorities retaliation for nan company’s publically stated views connected AI safety, successful usurpation of nan First Amendment.

The government, successful a 40-page filing earlier this week, rejected that framing entirely, saying that Anthropic’s refusal to let each lawful subject uses of its exertion was a business decision, not protected speech, and that nan nickname was a straightforward nationalist information call, not reward for nan company’s views.

The Pentagon lawsuit is not nan only ineligible matter connected Anthropic’s docket this Tuesday. Separately connected Friday, a national judge tentatively ruled that Reddit’s suit against nan institution — which accuses Anthropic of scraping its contented without support to train its AI — should beryllium sent backmost to authorities court, wherever Reddit primitively revenge it past June. A proceeding to finalize that determination is besides scheduled for Tuesday.