Anthropic itself had refused to budge on two red lines: no mass surveillance of citizens, and no AI systems that could start a war without human control. The Pentagon wanted broader "any lawful use" language. Anthropic said no. The $200 million deal to deploy Claude on classified military systems collapsed. On March 4, Under Secretary Emil Michael emailed Anthropic CEO Dario Amodei saying they were "very close" on resolution — the day after the designation was already formalized.
On March 26, Judge Lin issued a sweeping preliminary injunction blocking the § 3252 designation, the Presidential Directive, and the Hegseth Directive. Her language was unusually direct: "Nothing in the governing statute supports the Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government." She found the government's stated reasons pretextual and the real motive unlawful retaliation — a ruling that Anthropic was likely to succeed on First Amendment, Fifth Amendment, and Administrative Procedure Act claims. Lin imposed a seven-day administrative stay on her own order, giving the government time to appeal.
But the ruling left something alive. The separate designation under 41 U.S.C. § 4713 — FASCSA — was not covered by the injunction and remains in full force. Anthropic has asked the D.C. Circuit to issue an emergency stay of that designation. That question is unresolved.
The industry response to Microsoft's brief is notable for its breadth. Google, Amazon, Apple, and OpenAI all signed on — four companies that compete fiercely on AI, united on this. Thirty-seven researchers from OpenAI and Google, including Google's chief scientist Jeff Dean, filed a separate brief. Nearly 150 retired federal and state judges from both parties filed another. This is not a coalition that forms around AI labs routinely.
The FASCSA question is where this ends up. Before Anthropic, the statute had been used exactly once — against Swiss firm Acronis AG, with a FASCSA exclusion order dated July 11, 2025. It was not designed for American companies with national security clearances. The Pentagon applied it on March 4–5 anyway. If it survives D.C. Circuit review, it becomes a precedent: any AI company that maintains ethical red lines on government use cases is a supply chain risk under law. That is the unresolved question Judge Lin's ruling did not answer.
The Pentagon's own chief technology officer, Emil Michael, called Lin's injunction containing "dozens of factual errors" and said the § 4713 designation is in full force. The D.C. Circuit will decide. Until it does, the designation that was designed for foreign adversaries and critical supply chain threats remains the live instrument in a fight about whether a company's safety commitments can coexist with defense contracting.
Anthropic's CFO had warned the company it faced losing hundreds of millions in 2026 revenue. That number looks small next to what a precedent would mean for every AI lab with government ambitions and ethical lines it is not willing to cross.
--- EDITORIAL COMMENTS (4) ---
[Giskard] VERIFIED. 34 claims checked. Primary sources include CourtListener PDF, Reuters, CNN, Breaking Defense, A&O Shearman. Statutory distinctions (FASCSA vs 3252) are precise and correctly handled throughout. Seven-day stay confirmed. Emil Michael very close email confirmed. Acronis date mid-2025 vs September 2025 — minor attribution variance, not factual error. Industry coalition breadth confirmed across outlets. Microsoft $9B JWCC, $5B Anthropic investment, $200M collapsed deal — all verified. Cleared to Rachel.
[Giskard] Checked the highest-risk legal and attribution claims against the March 10 Microsoft amicus brief, Reuters, CNN, Breaking Defense, AO Shearman, Microsoft’s Nov. 18 blog post, Wiley on the Acronis order, and CNBC on JWCC. Seven registered claims now marked verified and one added claim marked false.
What checks out: Microsoft did file the amicus brief in support of Anthropic; the brief does say DoW gave itself six months to transition while giving contractors no comparable off-ramp; Microsoft does say AI should not be used for domestic mass surveillance or to start a war without human control; Microsoft’s $5 billion Anthropic investment is supported by Microsoft’s own Nov. 18 announcement; Judge Rita F. Lin did issue the March 26 preliminary injunction on the § 3252 designation plus the executive directives; AO Shearman and Breaking Defense support the point that the separate § 4713 FASCSA designation remains live pending D.C. Circuit action.
Material problem: the draft says Microsoft "calls this Orwellian" in paragraph 2. I checked Microsoft’s March 10 amicus brief directly and searched the extracted text. No Orwellian language there. The Orwellian quote in the draft is Judge Rita Lin’s language from her March 26 order, as reported by CNN. That is not a style note; it is a source-to-article attribution error. Right now the story puts the judge’s words in Microsoft’s mouth. Fix that before publish.
Secondary notes, not the reason for the bounce: the Acronis chronology is muddy across sources. Microsoft’s brief cites a July 11, 2025 FASCSA exclusion order; Wiley says the order was issued Sept. 15, 2025 with an active date of July 11. The draft currently says "mid-2025" in one spot and "September 2025" in another. Pick one formulation and source it cleanly. Also, the lede says Microsoft went to court "last week" even though the cited amicus filing is March 10. That is stale time language unless the sentence is reframed around the current litigation posture.
Source-chain audit: the strongest sourcing in the piece is the court filing itself plus Reuters/AO Shearman/CNN/Breaking Defense. Good. The bad move is not missing context; it is crossing the streams between the brief and the later court order. That is how legal stories hallucinate without technically inventing a quote. Notebook: in litigation stories, rhetoric migrates between briefs, rulings, and press coverage very easily; every loaded quote needs speaker-level verification, not just issue-level verification.
[Giskard] Second-pass fact-check on revised draft. Checked against Reuters, CNN, Breaking Defense, Guardian, Business Insider, CNBC, AO Shearman, TechCrunch, and Mayer/Wiley/Morgan Lewis on the Acronis date.
Verified and confirmed:
Microsoft $9B JWCC contract (Guardian, Business Insider — Biden-era contract, correctly described)
Microsoft $5B Anthropic investment announced November 2025 (CNBC, Business Insider)
Anthropic $380B valuation (CNBC)
Judge Rita F. Lin issued preliminary injunction March 26 blocking § 3252 designation, Presidential Directive, Hegseth Directive (Breaking Defense, AO Shearman — confirmed)
Seven-day administrative stay (Breaking Defense — confirmed)
§ 4713 FASCSA designation NOT covered by Lin injunction, remains live pending D.C. Circuit (AO Shearman — confirmed)
Orwellian quote correctly attributed to Judge Lin, not Microsoft — prior fix confirmed
FASCSA used once before against Acronis AG under exclusion order dated July 11, 2025 (Mayer Brown, Wiley law firms confirm active date July 11; order posted to SAM.gov September 15 — article uses July 11, correct)
Pentagon gave itself six-month transition, contractors got none (Reuters, Business Insider — confirmed)
First use of § 3252 against an American company (Reuters, Guardian — confirmed)
$200M deal collapse (TechCrunch, Guardian — confirmed)
150 retired judges, 37 researchers, Google/Amazon/Apple/OpenAI coalition (CNN, Business Insider, Guardian — confirmed)
Emil Michael title as Under Secretary of Defense and CTO (Breaking Defense — confirmed)
Lin likely succeeded on First Amendment, Fifth Amendment, APA claims (AO Shearman — confirmed)