Key Findings
Anthropic entered a classified Pentagon contract in June 2024 for Claude deployment in intelligence and planning systems. The relationship functioned for approximately 20 months before the government demanded removal of contractual restrictions on surveillance and autonomous weapons. See Statutory Analysis.
The core demand was not merely contract renegotiation: the Pentagon sought access to Claude for mass domestic surveillance capabilities — a demand that, if met, would require fundamental changes to the model's safety architecture. Anthropic refused.[1]
OpenAI's Pentagon deal, signed on the same day as Anthropic's designation, contained "the same two limitations" Anthropic demanded — restrictions on mass domestic surveillance and fully autonomous weapons.[2] This simultaneous opposite treatment of identical contractual terms is the factual foundation of the equal protection claim. See Constitutional Analysis § III.
The supply chain risk designation was announced via social media (Hegseth's X post and Trump's Truth Social post) with no published risk assessment, written determination, or congressional notification — bypassing every procedural safeguard established by prior designations and required by federal statute. See Statutory Analysis.
Bipartisan congressional alarm preceded the designation: the Wicker/Reed/McConnell/Coons letter urged resolution, and the designation proceeded anyway — potentially establishing "implied will of Congress" against the action under Youngstown Category 3 analysis. See Constitutional Analysis § V.
The government imposed a six-month phase-out for all military contractors to sever ties with Anthropic — demonstrating that no emergency required immediate action, while creating a compounding-harm dynamic as each day brings additional relationship terminations. See PI Framework § III.
Trump called Anthropic "Leftwing nut jobs" and threatened "the Full Power of the Presidency" with "major civil and criminal consequences" — direct evidence of retaliatory animus for First Amendment analysis.[3] See Constitutional Analysis § I.
Chronological Record
Phase I: Contract & Cooperation (June 2024 – Late 2025)
In June 2024, Anthropic entered a contract with the Pentagon to provide Claude for use in classified intelligence analysis and military planning systems. Claude became the only AI model deployed on U.S. military classified networks.[4] The contract included explicit restrictions prohibiting use of Claude for mass domestic surveillance of Americans and for fully autonomous weapons systems.
For approximately 20 months the relationship functioned. Claude supported classified operations — reportedly including intelligence work contributing to the capture of Venezuelan President Nicolás Maduro.[4] The contractual restrictions were not in dispute during this period.
Phase II: Escalation (Early–Mid February 2026)
In early 2026, reporting indicated the Pentagon sought to expand Claude's deployment into surveillance capabilities that the contractual restrictions prohibited. CNBC reported that the Department of War wanted Claude deployed for capabilities including monitoring of domestic communications — uses that Anthropic's contract explicitly barred.[1]
By mid-February, the dispute had become public. Axios reported that the Pentagon presented Anthropic with a 72-hour ultimatum: remove the restrictions on mass surveillance and autonomous weapons, or face contract termination.[5]
Phase III: Congressional Intervention (Feb. 24–26, 2026)
The dispute drew immediate bipartisan congressional attention. On February 26, Senators Wicker (R-MS), Reed (D-RI), McConnell (R-KY), and Coons (D-DE) — the chair and ranking members of the Senate Armed Services Committee and the Senate AI Caucus — sent a letter to the Pentagon urging negotiated resolution.[6]
Additional congressional voices weighed in: Senator Tillis (R-NC) called the situation "sophomoric." Senator Peters (D-MI) questioned the tight deadline. Senator Slotkin (D-MI) observed that "Congress has not put clear limits around AI's use in the military," implying these were legislative questions.[7]
Phase IV: The Designation (Feb. 27, 2026)
On February 27, 2026, Secretary of War Pete Hegseth posted on X announcing that Anthropic had been designated a "supply chain risk" under national security authorities. The post directed that no military contractor, supplier, or partner "may conduct any commercial activity with Anthropic" and imposed a six-month transition period for divestiture.[8]
That same day, President Trump posted on Truth Social: "Anthropic, a group of Leftwing nut jobs, made a DISASTROUS MISTAKE trying to STRONG-ARM the Department of War into accepting limits on America's AI defense capabilities. They will now feel the Full Power of the Presidency. Major civil and criminal consequences to follow."[3]
Hours later, OpenAI announced it had signed a Pentagon contract for Claude-equivalent deployment in classified systems. Fortune subsequently reported the OpenAI deal contained "the same two limitations" — restrictions on mass domestic surveillance and fully autonomous weapons — that Anthropic had demanded.[2]
Phase V: Immediate Aftermath (Feb. 27–28, 2026)
Anthropic issued a statement that evening noting it "ha[d] not yet received direct communication from the Department of War or the White House" and vowing to pursue legal remedies. The company stated: "No amount of intimidation or punishment from the Department of War will change our position on mass domestic surveillance or fully autonomous weapons."[9]
Expert and congressional reaction was swift. Dean Ball (Foundation for American Innovation, former Trump AI policy advisor) called the broad designation "almost surely illegal" and "a psychotic power grab." Peter Harrell (Georgetown/former NSC) stated the Department of War "can't legally tell contractors don't use Anthropic in private contracts."[2]
Senator Markey called the designation "reckless and unprecedented."[10] Senator Gillibrand issued a statement opposing the rejection of AI guardrails.[11] A coalition of over 50 organizations signed a letter calling for congressional oversight.[12]
The business impact analysis was severe. One analyst noted: "Every general counsel at every Fortune 500 company with any Pentagon exposure is going to ask one question: is using Claude worth the risk?" Defense officials privately acknowledged that disentangling from Claude would be a "huge pain in the ass" given its deep integration into classified systems.[4]
Comparative Context: Prior Supply Chain Risk Actions
The Anthropic designation departed radically from every prior supply chain risk action:
Kaspersky Lab (2024)
The Bureau of Industry and Security conducted a "lengthy and thorough investigation" before publishing the Final Determination (Case No. ICTS-2021-002) in the Federal Register at 89 FR 52434. The process included formal investigation, initial determination, mitigation opportunity, and final determination with published findings. The entity in question was a Russian-origin company with documented ties to Russian intelligence services.[13]
Huawei/ZTE (2020)
The FCC's formal designation followed the Secure and Trusted Communications Networks Act, published at 85 FR 230. The process involved formal rulemaking, public comment periods, and published findings. The entities were Chinese-origin companies with documented ties to the Chinese military.[14]
The Contrast
Formal multi-year investigations for foreign adversaries with documented state intelligence ties. Social media posts — with no investigation, no risk assessment, no Federal Register publication, no congressional notification — for a domestic AI company in a contractual dispute about ethical guardrails.
Sources
- CNBC (Novet), "Anthropic faces lose-lose scenario in Pentagon conflict," Feb. 27, 2026. ↩
- Jeremy Kahn, "OpenAI sweeps in to ink deal with Pentagon as Anthropic is designated a 'supply chain risk,'" Fortune, Feb. 28, 2026. ↩
- NPR (Bond & Brumfiel), "OpenAI announces Pentagon deal after Trump bans Anthropic," Feb. 27, 2026. ↩
- Ina Fried & Andrew Lawler, "Pentagon blacklists Anthropic, labels AI company 'supply chain risk,'" Axios, Feb. 27, 2026. ↩
- Andrew Lawler & Ashley Curi, "Trump blacklists Anthropic: Here's what being a 'supply chain risk' means," Axios, Feb. 27, 2026. ↩
- Axios, "Scoop: Top Senate defense leaders intervene in Pentagon-Anthropic AI dispute," Feb. 27, 2026. ↩
- Axios, "Congress rips Pentagon over 'sophomoric' Anthropic fight," Feb. 26, 2026. ↩
- CBS News (Walsh), "Hegseth declares Anthropic a supply chain risk," Feb. 28, 2026; ABC News (Wang), "Trump orders US government to cut ties with Anthropic," Feb. 27, 2026. ↩
- Anthropic, "Statement on the comments from Secretary of War Pete Hegseth," Feb. 27, 2026. ↩
- Sen. Markey, "Markey Demands Immediate Congressional Action," Feb. 27, 2026. ↩
- Sen. Gillibrand, "Statement On Secretary Hegseth's Rejection Of AI Guardrails," Feb. 2026. ↩
- Protect Democracy, "Call for Congressional oversight of the Department of Defense's demand for unrestricted AI use," Feb. 2026. ↩
- BIS, "Commerce Department Prohibits Russian Kaspersky Software," June 2024; Federal Register, 89 FR 52434 (June 24, 2024). ↩
- DWT, "Huawei and ZTE Designated as Threats to National Security," Aug. 2020; Federal Register, 85 FR 230 (Jan. 3, 2020). ↩
This document is an independent legal research compilation. It does not constitute legal advice. All factual claims sourced to publicly available primary documents, official statements, court filings, and reporting by major news organizations.