Anthropic Pentagon Blacklist Ruling: Findings and Injunction

What the court held

United States District Judge Rita Lin issued a 59-page written order on Thursday evening, August 27, blocking the Department of War’s designation of Anthropic as a supply chain risk. Reporting on the order describes three holdings: that the designation was retaliation for protected speech under the First Amendment, that it was arbitrary and capricious, and that Anthropic was denied the due process the Fifth Amendment requires. [1][3]

Lin wrote that the government’s "words and deeds confirm that the challenged actions were based on a desire to make a public example out of Anthropic for its ‘arrogance’ in criticizing the government, not based on any articulable basis to believe that Anthropic would actually sabotage its model." She added that "the empty invocation of national security is not a blank check to punish and retaliate against government critics," while noting that the department "is undisputedly free to select the AI vendor of its choice." [1][2]

Atlas interpretation: Those two sentences bound the ruling on both sides. The government lost on how it acted, not on whether it has to buy anything from Anthropic. A procurement decision made for ordinary reasons would have survived; a punishment dressed as one did not. [1][3]

How a usage policy became a national security label

The dispute began over two carve-outs. In February, Anthropic refused terms that would have allowed any lawful government use of Claude, holding back mass domestic surveillance and fully autonomous weapons. The department’s position was that a vendor should not constrain the military’s use of what it has bought. [3][4]

The escalation was fast and public. The administration directed agencies to treat Anthropic as a supply chain risk, the department made the designation official on March 5, and Anthropic sued in two courts four days later. An appeals court declined to pause the designation in April. It was the first time a U.S. company had been publicly labelled a supply chain risk under a procurement statute written to keep foreign sabotage out of military systems. [2][3]

OpenAI signed with the department hours after Anthropic was designated, saying it had secured prohibitions on domestic mass surveillance and human responsibility for the use of force. Those are close to the conditions Anthropic would not drop, which is part of why the department’s stated rationale did not hold together in court. [3]

The facts that decided it

Lin pointed at the government’s own conduct rather than at its rhetoric. Secretary Hegseth had proposed applying the Defense Production Act to Anthropic, which treats a company as essential to national security rather than as a threat to it. The department kept pursuing a contract with the company. The government was collaborating with Anthropic’s Mythos model on cybersecurity. And Anthropic, she found, "undisputedly lacks" any backdoor access to models once they are handed over. [2]

The government’s argument for why the ordinary evidentiary standard should not apply was that AI models are "so staggeringly enormous and opaque" that the department cannot evaluate one the way it evaluates a physical component. At a July 30 hearing Lin called the government’s position "really troubling" and said the record had "gotten worse for the government" over time. [1][4]

Atlas interpretation: The opacity argument is the interesting one, because it is true and it still lost. Nobody can inspect a set of weights the way they can inspect a circuit board. What the court refused was the inference drawn from that: that unverifiable means dangerous, and dangerous enough to justify a penalty without a hearing. Applied consistently, the government’s reasoning would have covered every frontier vendor it was simultaneously signing with. [2]

What this does not settle

The government was expected to challenge the ruling, and a separate, narrower Anthropic case over a different rule the department is using toward the same end remains pending in the federal appeals court in Washington. The order addresses the designation; it does not by itself restore the contracts, the certifications every Pentagon contractor had to sign, or the revenue Anthropic put in the billions when it sued. [1][3]

Atlas interpretation: The durable question is not this designation but the leverage it demonstrated. A frontier lab’s safety policy is now a commercial exposure the government can price, and the only reason it cost the government anything this time is that the retaliation was documented well enough to be found. A quieter version of the same decision, made through routine procurement discretion, is the version the opinion leaves available. [1][2]