What it is
“Supply-chain risk” is a formal designation the U.S. government can attach to a contractor to block it, and often its business partners, from federal procurement. The concept originates in laws meant to keep adversary-linked hardware and software out of sensitive government systems: the Federal Acquisition Supply Chain Security Act of 2018 created a government-wide council (the FASC) that can issue removal or exclusion orders against suppliers, a power first used against the backup-software firm Acronis in 2025. The Department of Defense has a parallel, narrower authority under 10 U.S.C. § 3252, implemented through DFARS clause 252.239-7018, that lets it exclude a source from procurements for “covered systems” — intelligence, cryptologic, command-and-control, weapons, or classified systems — based on supply-chain risk information, without disclosing the basis and without the decision being reviewable in a bid protest or in federal court.
That authority was built for cases like Kaspersky or Huawei — firms whose ownership or jurisdiction created a plausible espionage or sabotage risk. In February 2026 it was used differently. Defense Secretary Pete Hegseth designated Anthropic a “supply-chain risk to national security,” the first time the label had been applied to a U.S. company rather than one tied to an adversarial state. President Trump simultaneously directed federal agencies to stop using Anthropic’s products. The practical effect, per Hegseth’s order, is that “no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic” — a restriction that reaches well beyond Anthropic’s direct dealings with the Pentagon.
The designation followed the collapse of a July 2025 contract under which Claude became the first frontier model cleared for classified Pentagon networks, with the department agreeing to abide by Anthropic’s acceptable-use policy. That policy bars use of Claude in fully autonomous weapons systems and mass domestic surveillance without human oversight. The Pentagon later sought to renegotiate those terms, asking Anthropic to permit “all lawful purposes” without those carve-outs; Anthropic refused, and the designation followed.
Why it matters for AI governance and narratives
The episode turns a national-security procurement tool into leverage in a fight over what constraints an AI company may impose on its own government customer. Critics — including a Just Security legal analysis and a group of Senate Democrats — have called the designation pretextual: a mechanism designed for genuine espionage risk redeployed to punish a safety policy the government found inconvenient, rather than a finding about Anthropic’s actual supply chain. Anthropic’s own public statement framed it as retaliation for declining to waive its weapons and surveillance restrictions.
This matters for how the AI safety debate is narrated because it demonstrates that safety commitments carry a concrete, government-imposed cost — lost defense revenue and disrupted commercial relationships — that has not, to date, been matched by any comparable cost in Anthropic’s access to private capital. That asymmetry is precisely what makes the designation a data point about where the real selection pressure on AI companies’ safety postures currently operates: procurement, not valuation.
Key facts and dates
- December 2018: Federal Acquisition Supply Chain Security Act establishes the FASC and general exclusion-order authority (separate from the DoD-specific mechanism used against Anthropic).
- July 2025: Anthropic and the Pentagon sign a contract making Claude the first frontier model approved for classified networks, conditioned on the Pentagon honoring Anthropic’s acceptable-use policy.
- February 27, 2026: Trump directs federal agencies to halt use of Anthropic’s AI; Hegseth designates Anthropic a supply-chain risk under DoD’s 10 U.S.C. § 3252 authority after Anthropic refused to waive its autonomous-weapons and domestic-surveillance restrictions.
- September 2026: A divided (2-1) D.C. Circuit panel rejects Anthropic’s legal challenge, allowing the designation and resulting restrictions to stand.
Accounts of the underlying contract dispute (the Pentagon’s demand versus Anthropic’s refusal) are consistent across legal-industry analyses (Mayer Brown, Goodwin) and news coverage (NPR, CBS, Fortune), and Anthropic’s own public statement does not dispute the basic sequence of events, though it disputes the designation’s legality and motive.
Where to learn more
- 10 U.S.C. § 3252 — Requirements for information relating to supply chain risk — the statute underlying the designation
- DFARS 252.239-7018 — Supply Chain Risk clause — how the authority is implemented in contracts
- Anthropic, “Statement on the comments from Secretary of War Pete Hegseth” — Anthropic’s own account
- Just Security, “What Hegseth’s ‘Supply Chain Risk’ Designation of Anthropic Does and Doesn’t Mean” — independent legal analysis of the designation’s scope and limits
- NPR, “Pentagon labels AI company Anthropic a supply chain risk” — news account of the announcement and its stated rationale