Anthropic Won in Court. The Blacklist Still Worked.
A judge ruled the Pentagon's blacklisting of Anthropic was illegal retaliation — after the contracts had already moved and the guardrails were gone.
Anthropic won. On Wednesday a federal judge ruled that the Pentagon's blacklisting of the company was illegal retaliation for its refusal to strip safety guardrails from its AI — the "empty invocation of national security" the reader has already seen. And the company is marking the victory by signing a $45 billion compute deal and preparing to go public [1]. That is what winning looks like now: a company vindicated in court and exiled from the government work it was fighting to keep. The ruling bars the government from enforcing the blacklist going forward. It does not require the government to resume business with Anthropic [2]. The contracts have already moved. The six-month phase-out the Pentagon set in motion expired before the judge ruled [3]. The work went to xAI, which accepted the "all lawful purposes" standard — mass surveillance and autonomous weapons included — that Anthropic refused [4]. And it went to OpenAI, which added paper prohibitions only after a user revolt and a wave of resignations [5]. Amodei had a word for those prohibitions.
no amount of intimidation or punishment from the Department of War will change our position. — Anthropic
The guardrails Anthropic refused to strip are simply absent from the military AI stack now, and no court order puts them back. The timing is not an accident of the calendar. It is the mechanism. The blacklist was imposed in February; a judge granted a preliminary injunction in March, and the D.C. Circuit overturned the stay, letting the blacklisting continue through litigation [6]. The Supreme Court's emergency docket has stayed lower-court rulings blocking administration actions in 15 of 15 cases since April 2025 [7]. The courts' own procedural delays are the window the administration operates in: the weapon does its work in the months it takes to review it. The parallel is the tariff case. The Supreme Court ruled the IEEPA tariffs illegal, and the government is refunding $166 billion — real money, a real remedy [8]. But Trump's response was the most honest sentence in the whole pattern.
the Supreme Court’s decision had merely forced the administration to use a different method to impose them — Donald Trump
The ruling didn't undo the tariffs; it rerouted them. Same with Anthropic. A court can rule the weapon illegal. It cannot un-fire it. That gap — between ruling and remedy — is not a flaw in the strategy. It is the strategy.
- 1. Anthropic Signs $45 Billion Compute Deal With Nscale
- 2. Judge Blocks Trump Administration Ban on AI Firm Anthropic
- 3. Federal Judge Blocks Pentagon Risk Designation of Anthropic
- 4. Pentagon Deploys xAI Grok After Standoff With Anthropic
- 5. OpenAI Amends Pentagon Deal After User Exodus and Resignations
- 6. Appeals Court Upholds Pentagon Blacklisting of AI Firm Anthropic
- 7. Supreme Court Uses Shadow Docket to Enable Trump Policies
- 8. US Government Refunds $166 Billion in Illegal Trump Tariffs