On March 26, 2026, US District Judge Rita Lin granted Anthropic a preliminary injunction, temporarily blocking the Department of Defense from designating the company a “supply chain risk.” Her language was unsparing: the designation is “likely both contrary to law and arbitrary and capricious,” the broad measures “appear designed to punish Anthropic” rather than serve national security — in her words, “classic illegal First Amendment retaliation.”
For the first time since the government and a frontier lab went to war, a court has hit the brakes. At a hearing two days before the ruling, Lin had already signaled her read of the case, saying the government appeared to be illegally trying to “cripple” and “punish” the company. The order takes effect in one week, giving the administration time to appeal. But the ruling draws a line at the same time: the Pentagon may still lawfully phase out Claude and pick another vendor — it just cannot cite the supply-chain-risk label as the reason.
Back to February 27: What the Injunction Actually Does
The injunction “restores the status quo” to February 27, before the punitive directives were issued. Quick recap of how we got here: the Pentagon (renamed the Department of War under Trump) had used Claude for roughly two years, including for sensitive documents and classified-data analysis, before deciding the company “could not be trusted” because it insisted on usage restrictions. On February 27, Trump directed federal agencies to stop using Anthropic technology, and Defense Secretary Pete Hegseth designated the company a supply chain risk the same day — a label normally reserved for foreign adversaries. WIRED reports the label froze Claude usage across government and damaged Anthropic’s sales and reputation.
The Judge’s Reasoning: Retaliation, Not National Security
Lin’s chain of reasoning is direct. First, the official rationale fails: “the record strongly suggests” the government’s stated reasons were pretextual. Second, if command integrity were genuinely the concern, the Pentagon could simply stop using Claude — it did not need a blacklisting label. Third, she rejected the suggestion that Anthropic might ideologically sabotage the military: there is “no legitimate basis to infer from Anthropic’s forthright insistence on usage restrictions” that it would sabotage anything. The record, she found, supports the opposite inference — that Anthropic is being punished for criticizing the government’s contracting position in the press.
Where It Began: The Usage-Limits Red Line
Anthropic’s red line never moved: Claude is not to be used for fully autonomous weapons — AI making final battlefield targeting decisions without a human in the loop — or for mass domestic surveillance. On March 9, the company filed two federal lawsuits naming officials including Hegseth and the Secretary of State, arguing the sanctions are unconstitutional. After the ruling, Anthropic said it was pleased the court found it likely to succeed on the merits, and that its focus remains working productively with the government.
What the Ruling Does Not Settle
This is a preliminary injunction, not a final judgment, and Lin set no schedule for a merits ruling. POLITICO reports she stayed the order precisely to leave room for appeal; subsequent coverage indicates the Pentagon has appealed, and its CTO said the ban still stands after the decision. Anthropic’s second lawsuit — over a separate law barring it from supplying software to the military — is still pending before a federal appeals court in Washington, DC. And agencies remain free to cancel contracts or ask contractors to drop Claude, as long as they do not invoke the designation.
What It Means for the AI Industry
Three takeaways. First, procurement weapons like the supply-chain-risk label are no longer a free move: courts will examine their motives and legality, which gives vendors a real defense and gives agencies a reason to think twice before reaching for the blacklist. Second, safety red lines got judicial backing — insisting on usage restrictions cannot itself be spun into evidence of sabotage, an important precedent for every lab walking the tightrope between the government market and its own policies. Third, the case is far from over: the appeal, the second lawsuit, and the Pentagon’s transition period are all still running. The rules of government AI procurement will be written in courtrooms, not press conferences, over the coming months.
Sources
- Anthropic Supply-Chain-Risk Designation Halted by Judge — WIRED
- Judge temporarily blocks Pentagon’s supply chain risk designation — ABC News
- Judge pauses Pentagon’s punishment for Anthropic — POLITICO
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.