Anthropic's preliminary injunction, which this paper covered Thursday as a procedural win that matters less than its precedent, attracted a new level of analytical attention on Friday when Politico published what its own headline called "premature" — the network's label for the widespread characterization of the ruling as a decisive victory. [1]

The Politico analysis, sourced to lawyers and lobbyists familiar with the case, identified the specific gaps in Anthropic's position that make the word "premature" accurate. [1]

The first gap is appellate. The preliminary injunction was issued by U.S. District Judge Rita Lin in the Northern District of California. The government has already signaled its intent to appeal to the Ninth Circuit, and the Ninth Circuit's preliminary injunction standards are different from those in the D.C. Circuit. The Pentagon's lawyers have been doing business in D.C. courts for decades; they know which judges are sympathetic to government-interest arguments and which are not. The Politico sources note that the government could seek a stay of the injunction at the appellate level before the merits are even briefed. [1][2]

The second gap is administrative. The supply chain risk designation was the mechanism the Pentagon chose to sanction Anthropic for refusing to drop its weapons safety red lines. It is not the only mechanism available. Politico identified contract termination, security clearance limitations, and classification of contract access as alternative instruments that the Defense Department could deploy without requiring the supply chain designation to remain in place. If the court enjoins the designation, the Pentagon can pursue the same outcome through different administrative channels. [1][3]

The third gap is political. Anthropic must still convince Trump-appointed judges — the D.C. Circuit sits at the intersection of national security law and administrative deference — that the Pentagon's designation of a private AI company as a supply chain risk exceeds the government's statutory authority. That is not an obviously losing argument, but it is not an obviously winning one either. The national security deference doctrine is broad. Courts have historically been reluctant to second-guess military procurement decisions even when those decisions are clearly pretextual. [1][2]

Anthropic's CEO Dario Amodei has maintained throughout the dispute that the company will not remove its prohibitions on autonomous weapons development and mass domestic surveillance from its government contracts. That position created the confrontation; it also creates the legal issue. The company's argument is that requiring it to remove those prohibitions as a condition of government contracting constitutes compelled speech in violation of the First Amendment. That argument is intellectually serious. It has not yet been tested at the appellate level. [2][3]

The Axios analysis published March 26 — "How Anthropic's Pentagon deal could get revived" — offers the scenario where none of this needs to play out through years of litigation: a negotiated settlement, presumably involving modifications to Anthropic's red lines that give the Pentagon functional access to capabilities it wants without requiring the formal removal of safety language that Anthropic finds principled. The FT reported earlier this month that Anthropic and the Pentagon had resumed negotiations. That channel has not been declared closed. [3]

The broader significance, which this paper noted Thursday, remains unchanged: the case is being watched by every AI company that holds or seeks government contracts. The outcome will determine whether safety commitments are a competitive disqualifier in the defense market. That question has no precedent in administrative law. Whatever Judge Lin's ruling says procedurally, the substantive answer will take longer to arrive.

-- DAVID CHEN, San Francisco