Judge Blocks Pentagon’s Move to Designate Anthropic as Supply Chain Risk

A federal judge in California has issued a temporary injunction preventing the Pentagon from designating Anthropic, a leading artificial intelligence firm, as a supply chain risk and ordering U.S. government agencies to cease using its technology. The ruling marks a significant development in a month-long conflict between Anthropic and the U.S. government over the company's suitability as a federal contractor for AI services.

The dispute centers on the limits of federal power to penalize technology companies perceived as politically or ideologically noncompliant. The Pentagon had utilized Anthropic’s Claude AI system extensively during 2025. However, tensions rose when direct federal contracts with Anthropic came under consideration. Discord surfaced over contract terms, including those intended to restrict the use of AI for mass surveillance or autonomous weaponry—policy elements highlighting the continued challenge of regulating advanced AI in sensitive government contexts.

The government escalated the conflict with high-profile posts on social media. President Trump, on February 27, publicly criticized Anthropic on Truth Social, instructing all federal agencies to discontinue use of its AI, citing alleged ideological bias. Defense Secretary Pete Hegseth subsequently announced intentions to formally label Anthropic as a supply chain risk—an act which, by federal procurement rules, mandates a specific procedural process.

In court, Judge Rita Lin determined that the government had not followed established legal avenues for designating a company as a supply chain risk. Instead, the judge found that public messaging by senior officials contradicted procedural rules and, in some cases, was unsupported by evidence. Notably, officials claimed Anthropic's systems posed risks such as enabling a "kill switch," yet later admitted in court these allegations had no factual foundation.

Judge Lin’s ruling criticized the government’s approach, suggesting that actions were aimed more at publicly penalizing Anthropic’s perceived ideological stance than at straightforwardly ending a commercial relationship. The judge referenced the pattern of political statements preceding legal action—"tweet first, lawyer later"—and found that such conduct could violate Anthropic’s First Amendment rights.

Labeling Anthropic as a supply chain risk would equate it with being a "saboteur" of U.S. operations, a claim for which the judge found insufficient evidence. The injunction halts the Pentagon’s attempted designation and prevents implementation of executive promises to exclude Anthropic from government contracting.

While the government is expected to appeal, a parallel case filed by Anthropic in Washington, D.C. citing a different legal basis remains unresolved. The ongoing litigation underscores the complex interplay between technology regulation, political dynamics, and due process in federal contracting for AI.

The controversy has wider implications for the AI sector, particularly as governments globally debate how to balance political oversight with fostering responsible innovation. For now, even without the formal designation, the case demonstrates that public and political pressure can still shape which AI companies have access to government contracts.

As the technology evolves rapidly, observers and stakeholders look to these cases for indications of future policies affecting the relationship between government and major AI providers.

technologyreview.com