The Trump administration filed its appeal of Judge Rita Lin's preliminary injunction on Thursday, the seventh and final day of the stay Lin had granted for exactly this purpose. [1] The case now moves to the Ninth Circuit Court of Appeals, where it will test a question that extends well beyond one AI company: can the federal government designate a private technology firm as a "supply chain risk" to punish it for public criticism of military policy?

The timeline is compressed and worth recounting. On February 27, Defense Secretary Pete Hegseth posted on X that Anthropic was "sanctimonious" and delivered "a master class in arrogance" -- a response to the company's public position that its AI model, Claude, should not be used for domestic mass surveillance or fully autonomous weapons systems. [2] The same day, the Pentagon formally designated Anthropic as a supply chain risk. The GSA removed Anthropic from USAi.gov and the government's Multiple Award Schedule. [3] Trump issued a directive ordering all federal agencies to "immediately cease all use of Anthropic's technology." [2]

Anthropic sued. Judge Lin, of the Northern District of California, held a hearing on March 24 at which she said, on the record, that the government's actions looked "like an attempt to cripple Anthropic." [4] She called the administration's reasoning "Orwellian." [2] On March 26, she issued a forty-three-page preliminary injunction blocking both the supply chain risk designation and the federal usage ban. [2] Her ruling found that the measures were "likely unlawful" and constituted "classic illegal First Amendment retaliation" -- the government punishing a company for its public speech. [2]

Lin gave the government seven days to appeal before the injunction would take full effect. On day seven, the appeal landed.

The legal questions for the Ninth Circuit are substantial. The supply chain risk designation originates in the Federal Acquisition Supply Chain Security Act of 2018, which was designed to allow the government to exclude companies whose products pose genuine national security threats -- the paradigmatic case being Huawei's telecommunications equipment. [5] The statute grants broad discretion to the executive branch in identifying risks. The question is whether that discretion encompasses designating a company not because its products are compromised but because its executives publicly disagreed with the Defense Secretary.

Lin's opinion drew a sharp line. She found that Anthropic's products pose no supply chain risk in any conventional sense -- they are software models, not hardware with embedded vulnerabilities. [2] The Pentagon's own intelligence agencies used Claude during the early days of the Iran strikes, hours after Trump ordered all agencies to stop using it, a contradiction that Lin noted in her ruling. [6] The government's argument, reduced to its essentials, was that Anthropic's refusal to allow unrestricted military use of Claude created an operational dependency risk. Lin found this reasoning pretextual, noting that the government's restrictions on military AI already exceeded Anthropic's usage policies through existing federal law and DoD internal directives. [2]

The Ninth Circuit appeal raises the stakes. If the appellate court upholds Lin's injunction, it establishes precedent that the executive branch cannot use supply chain security authorities as a retaliatory weapon against companies that exercise First Amendment rights. If the court reverses, it creates a framework in which any technology company that publicly criticizes military policy can be effectively blacklisted from government contracts -- not through a transparent procurement decision, but through a national security designation that carries the force of law and the stigma of being labeled a threat.

Roger Parloff, the legal journalist whose coverage of the case has been the most granular, noted the timing significance on X: the government waited until the final day of the stay to file, suggesting internal deliberation about whether to pursue the appeal at all. [1] Zvi Mowshowitz, the AI policy analyst, flagged that Anthropic had never seen the classified memorandum justifying its supply chain risk designation until it was filed in court proceedings -- meaning the company was blacklisted on the basis of a document it was not allowed to read. [7]

For Anthropic, the appeal means continued uncertainty. The injunction remains in effect during the appellate process, so the immediate business impact is contained. But the political message is clear: the administration intends to fight for the principle that it can punish companies that set ethical limits on their products. For the AI industry more broadly, the case functions as a warning. OpenAI, Google, and Meta all maintain usage policies that restrict military applications to varying degrees. If the government's theory prevails, any of those restrictions could become the basis for a supply chain risk designation.

Hannah Arendt observed that the most dangerous moment for a republic is when the instruments of security become instruments of political control. The supply chain risk statute was written to protect the government from compromised technology. It is being used to compromise a company that said no.

-- ANNA WEBER, Berlin