Judge blocks DOD's ban on Anthropic, calls it First Amendment retaliation
Judge Rita Lin in the U.S. District Court of Northern California is also blocking the government's enforcement of a presidential directive that all government agencies stop using the company’s artificial intelligence products.
Lin cited several grounds in her Thursday ruling, which says the Defense Department’s declaration was retaliation for Anthropic exercising its First Amendment rights.
DOD also violated Anthropic’s due process rights by not giving the company advance notice or an opportunity to respond before the ban went into effect. DOD also did not follow the procedures laid out in the federal law they relied on to ban the company from federal work, Lin said.
The judge also called DOD’s action “arbitrary and capricious” and contrary to the law.
Anthropic has been working with the Defense Department since late 2024 through a partnership with Palantir Technologies. Since March 2025, Anthropic has also gone to market with a standalone product Claude Gov.
The dispute emerged in the fall of 2025 when DOD pushed for unrestricted access to Claude for “all lawful uses.” Anthropic refused to remove two long standing restrictions – no mass surveillance of U.S. citizens and no lethal autonomous warfare.
DOD wanted those restrictions lifted, but Anthropic refused. Negotiations were cordial and Anthropic offered to help DOD transition to another vendor, according to the judge’s ruling.
But things went south in January, when Anthropic and DOD went public with the dispute. CEO Dario Amodei posted an essay that month talking about AI safety, and the company issued a statement on Feb. 26 on its position about how DOD should use AI.
Within 24 hours, President Trump issued his government-wide ban on Truth Social and Defense Secretary Pete Hegseth designated Anthropic a supply chain risk.
“Neither the President nor Secretary Hegseth cited any statutory authority for the Directives,” Lin wrote in her ruling.
Anthropic worked for DOD for years and had gone through a lengthy national security vetting process. The company received nothing but praise, the judge wrote.
"The day after the designation was finalized—and before it was communicated to Anthropic—Under Secretary (Emil) Michael and Amodei cordially exchanged drafts of Anthropic's usage terms, with Under Secretary Michael writing to Amodei: 'After reviewing with our attorneys and seeing your last draft (thanks for being fast), I think we are very close here,'" the judge wrote.
On the First Amendment question, Lin found that Anthropic's public statements about AI safety — including Amodei's essay and the company's public statement on its dispute with DOD — were protected speech on matters of public concern.
Courts have long held that matters of public concern are at the core of First Amendment protections.
The judge said the government's own words undermined its national security argument. Trump called Anthropic a "radical left, woke company" and Hegseth attacked its "sanctimonious rhetoric" and "Silicon Valley ideology."
Most tellingly, an internal DOD memo stated that Anthropic's risk level escalated because it was engaging in an "increasingly hostile manner through the press."
"Punishing Anthropic for bringing public scrutiny to the government's contracting position is classic illegal First Amendment retaliation," Lin wrote.
The injunction takes effect in seven days. That timeline gives the government until around April 2 to seek an emergency stay from the Ninth Circuit Court of Appeals, which it has indicated it will do.
Meanwhile, a separate but related case challenging the supply chain designation under a different federal statute is already pending in the D.C. Circuit Court of Appeals. That means the legal battle over Anthropic's status as a government contractor is likely to play out on multiple fronts simultaneously.
Anthropic’s battle with DOD and the Trump administration has drawn a variety of supporters, who have filed amicus briefs with the court.
Among them are employees of its competitors Google and OpenAI. Microsoft also filed a brief as did several industry associations.
]]>