Anthropic, a San Francisco-based artificial intelligence company, has taken its fight with the Trump administration to federal court. The company was recently designated a “Supply-Chain Risk to National Security” by the Department of Defense (DOD) after refusing to loosen restrictions on how its AI models can be used.
Anthropic says the government is retaliating against it for maintaining two safety guardrails. One prohibits fully autonomous lethal warfare; the other bars mass surveillance of Americans.
The lawsuit (pdf), filed Monday in the U.S. District Court for the Northern District of California, targets the DOD and Defense Secretary Pete Hegseth along with numerous federal agencies and officials. Anthropic is asking the court to block the blacklist and declare the government’s actions unlawful.
The Clash Over AI Use
The complaint portrays the dispute as a breakdown in negotiations between the Pentagon and one of its most important AI partners.
Anthropic says it had worked closely with the military and that its flagship model, Claude, had already become deeply embedded in defense systems. According to the lawsuit, Claude is “reportedly the Department’s most widely deployed and used frontier AI model, and the only frontier AI model on the Department’s classified systems.”
The relationship unraveled during negotiations over how the technology could be used.
The Pentagon demanded what it called “all lawful use” of Anthropic’s AI models across defense operations. Anthropic says it agreed broadly but refused to remove two restrictions it considers essential:
Anthropic’s Usage Policy has always conveyed its view that Claude should not be used for two specific applications: (1) lethal autonomous warfare and (2) surveillance of Americans en masse.
The company argues that those limits reflect the real capabilities and risks of current AI systems:
Anthropic currently does not have confidence, for example, that Claude would function reliably or safely if used to support lethal autonomous warfare.
The lawsuit adds that the restrictions are rooted in the company’s technical understanding of the system:
These usage restrictions are therefore rooted in Anthropic’s unique understanding of Claude’s risks and limitations — including Claude’s capacity to make mistakes and its unprecedented ability to accelerate and automate analysis of massive amounts of data, including data about American citizens.
The Pentagon, however, insisted that military use of the technology should not be constrained by a private company’s policies.
Anthropic Accused of “Wokeness”
According to the lawsuit, the Pentagon did not treat the dispute as an ordinary contract disagreement. Instead, the Department publicly castigated Anthropic and its leadership. Among other things, the DOD said CEO Dario Amodei was “ok putting our nation’s safety at risk.”
The confrontation escalated into an ultimatum. Officials allegedly warned Anthropic to accept the government’s demands by 5:01 p.m. on February 27, 2026, or “pay a price.”
That price, the complaint says, would take one of two forms. The company could be expelled from the defense supply chain under 10 U.S.C. § 3252, or the government could force Anthropic to provide unrestricted access to its AI models under the Defense Production Act.
Amodei publicly responded that the company could not “in good conscience accede to” the Pentagon’s demands.
Retaliation followed almost immediately.
The next day, President Donald Trump publicly attacked the company. The lawsuit recounts that the president posted a message on social media, described in the filing as a Presidential Directive, “directing EVERY Federal Agency in the United States Government to IMMEDIATELY CEASE all use of Anthropic’s technology.” He also “derided Anthropic as ‘out-of-control’ and a ‘RADICAL LEFT, WOKE COMPANY’ of ‘Leftwing nut jobs’.” The lawsuit also cites,
“Anthropic better get their act together,” the President threatened, or he would “use the Full Power of the Presidency to make them comply, with major civil and criminal consequences to follow.”
The directive set the stage for a broader government crackdown.
Government Agencies Move to Cut Ties
That same afternoon, Hegseth issued what the lawsuit calls a Secretarial Order.
Acting on the president’s directive, Hegseth announced on social media that he was designating Anthropic a “Supply-Chain Risk to National Security.” He also ordered that:
Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic.
Hegseth criticized what he described as Anthropic’s “Silicon Valley ideology,” “defective altruism,” and “corporate virtue-signaling.” He also suggested the company lacked patriotism. At the same time, the secretary required Anthropic to continue providing services to the Department for up to six months.
Other federal agencies soon followed.
The General Services Administration (GSA) terminated Anthropic’s “OneGov” contract, cutting off access to the company’s services across much of the federal government. The Treasury Department and the Federal Housing Finance Agency (FHFA) announced they were ending their use of Anthropic technology.
The State Department and the Department of Health and Human Services (HHS), according to the complaint, circulated internal memoranda directing employees to stop using the company’s AI systems.
The pressure intensified in early March.
On March 4, Hegseth sent Anthropic a formal letter confirming the designation. The letter, dated March 3 and referred to in the lawsuit as the Secretarial Letter, informed the company that:
the Department of War (DoW) has determined … that the use of [Anthropic’s] products in [the Department’s] covered systems presents a supply chain risk.
The letter said invoking 10 U.S.C. § 3252 was “necessary to protect national security.” It also stated that the determination applies to all Anthropic “products” and “services,” including those that “become available for procurement.”
The Department further asserted that “less intrusive measures are not reasonably available” to address the supposed risk.
Five Fronts of Legal Attack
Anthropic’s lawsuit challenges the government’s actions on several legal grounds.
First, the company argues that the Pentagon’s designation violates the Administrative Procedure Act (APA). The complaint says the statute cited by the government, 10 U.S.C. § 3252, is meant to address sabotage or subversion by foreign adversaries, not policy disagreements with domestic contractors.
“The Secretary has not determined, and cannot reasonably determine, that Anthropic’s services present a risk of sabotage or subversion by an adversary to the United States,” the lawsuit states.
Second, Anthropic claims the government violated the First Amendment by retaliating against the company’s speech about the limits of its own technology.
The complaint argues that while the government may disagree with those views, it cannot punish them:
But the government may not employ “the power of the State to punish or suppress [Anthropic’s] disfavored expression.”
Third, the lawsuit argues that Trump’s directive was ultra vires, meaning beyond the president’s legal authority. It cites Youngstown Sheet & Tube Co. v. Sawyer, the landmark Supreme Court case that blocked President Harry S. Truman’s seizure of steel mills.
Fourth, Anthropic alleges violations of the Fifth Amendment’s due process protections. The company says the government inflicted severe reputational and economic damage without notice, evidence, or an opportunity to respond.
Fifth, the lawsuit claims multiple federal agencies imposed sanctions without statutory authority.
Taken together, the company argues, the actions amount to an “unprecedented” use of government power against a private firm.
A Broader Warning
The lawsuit summarizes its core claim bluntly:
The Constitution does not allow the government to wield its enormous power to punish a company for its protected speech. No federal statute authorizes the actions taken here.
Nicolas M. Chaillan, a former Air Force and Space Force officer and the U.S. Air Force’s first chief software officer, said the dispute raises serious questions about the use of government power.
“The moment we turn the law or tools on our own citizens, we’ve become the thing we fight against,” Chaillan told The New American, referring to China.
