Pentagon keeps its Anthropic blacklist; the $2T IPO inherits the fight

The blacklist stands, on a 2-1 vote

A federal appeals panel in Washington ruled on 25 September that the Pentagon lawfully branded Anthropic a supply-chain risk↗, a ban that keeps Claude out of the US military and away from every contractor working for it ¹. Judges Gregory Katsas and Neomi Rao, both Trump appointees, sided with the department; Judge Karen LeCraft Henderson dissented ¹. The panel held its ruling back only so Anthropic can try for a rehearing, an en banc hearing, or the Supreme Court ¹.

The defense account was never the money

Anthropic's Pentagon work sat under a July 2025 contract with a ceiling of up to $200 million↗, the same ceiling OpenAI, Google and xAI signed that month ². What is actually at stake is the listing: a November debut that could value the company near $2 trillion and raise as much as $100 billion↗ ³. Anthropic has said the designations cost it revenue as customers hesitated to buy from a company the government was pushing away, and it has never put a figure on the loss ⁴.

Gatekeeper on Thursday, security risk on Friday

The ruling landed the day after word surfaced that the White House had asked Anthropic and OpenAI to hold new models back from Britain's AI Security Institute until the US government tested them first↗ ⁵. Anthropic complied: Claude Mythos 5.1 shipped "only available to a set of U.S. organizations" ⁵. Judge Rita Lin in San Francisco ruled last month that the parallel designation covering federal contracts was unlawful↗; Friday's decision leaves the defense label standing indefinitely ¹⁴⁶.

What the majority accepted, and the dissent rejected

Katsas wrote that Anthropic's built-in limits mean "these restrictions have stopped Claude from performing tasks requested by government users," and that "it is the President and the Secretary of War who must determine how best to balance the competing risks" ¹⁶. Defense Secretary Pete Hegseth had accused the lab of trying to "seize veto power over the operational decisions of the United States military" ¹. Henderson answered that Congress wrote the 2018 supply-chain law against sabotage by hostile foreign powers, not safety rules from an American company ⁶.

The loss may not survive a fuller court

Neither remaining appeal is guaranteed. En banc review by the full D.C. Circuit and review by the Supreme Court are both discretionary, and legal analysts expect Anthropic to win before a full bench that mixes four presidents' appointees, and probably at the high court, if either takes the case↗ ⁶. The account moves on regardless: the Pentagon has been replacing Claude with SpaceX's Grok, Google's Gemini and OpenAI's GPT models, and OpenAI's classified-network deployment deal was announced on 28 February ²⁴.

Days to pick a door

The petition window is open now and the roadshow is close: marketing could begin in mid-October, with the listing set for days before the November midterms ⁷. Shareholders are meanwhile being asked to hand the seven co-founders, who own roughly 2% each, a 50.1% voting bloc↗ before the bell ⁸⁹. The company that wants a majority of the vote is still waiting on two courts for the rest.