A federal appeals court upheld the U.S. Department of Defense's decision to exclude Anthropic from the Pentagon's supply chain. The ruling confirms the AI company's designation as a national security risk over restrictions on how its Claude models can be used in military operations.

The 2-1 decision by the U.S. Court of Appeals for the D.C. Circuit leaves intact the Department's March 3 order removing Anthropic products from Department systems and barring contractors from using them. The court found the government had sufficient grounds to conclude that Anthropic's continued integration into military systems created a risk, rejecting arguments that individual model testing could mitigate these concerns.

Conflict over AI usage restrictions

The dispute originated from a breakdown in negotiations earlier this year regarding the military's use of Claude. While Anthropic agreed to expand permitted government uses, it maintained two core restrictions: the models could not be used for lethal autonomous warfare or mass surveillance of Americans. The Department sought contractual language allowing "all lawful uses" of the technology.

When Anthropic refused to remove those restrictions, Defense Secretary Pete Hegseth designated the company as presenting a significant supply-chain risk under the Federal Acquisition Supply Chain Security Act of 2018. The court majority pointed to the way Anthropic embeds restrictions into Claude that can cause the model to refuse certain requests as evidence of the risk.

Legal reactions and ongoing challenges

Emil Michael, Under Secretary of Defense for Research and Engineering, welcomed the decision on social media, calling Anthropic a "supply chain risk" to the defense industrial base. He stated that no private company should insert their opinions into the chain of command.

An Anthropic spokesperson stated the company respectfully disagrees with the decision and is considering all options, including further review. The spokesperson noted that another federal court has already held the government's parallel designation unlawful, though that case involved a different statutory framework.

Constitutional claims rejected

Anthropic also argued that the government's action violated its constitutional rights and amounted to retaliation for its position on AI safety. The appeals court rejected those claims, stating the government acted because Anthropic refused to accept a contract term officials considered essential, rather than because of the company's advocacy around AI policy.

The court acknowledged the tension in the case, noting difficult questions about the government's use of increasingly powerful AI systems and who should decide how those systems are deployed in military operations.

What the numbers show

The judicial split highlights a critical divergence in legal interpretations of AI governance in defense contexts. While the D.C. Circuit upheld the exclusion based on national security risks, a separate federal court in California previously ruled against a related government designation involving the company. This indicates that while the Pentagon's specific supply-chain ban remains active, the broader legal landscape regarding Anthropic's status is contested across different jurisdictions and statutory frameworks.

Anthropic and medical knowledge platform OpenEvidence announced a collaboration on Tuesday to bring free AI-powered clinical decision support to physicians across nearly 100 low- and middle-income countries.

The initiative aims to close the information gap in healthcare by providing access to peer-reviewed research and treatment guidelines. While the platform is already free for clinicians in the U.S. and Europe, this expansion targets underserved regions including Uganda, Angola, Sudan, Haiti, and Mongolia. Anthropic will provide the back-end AI infrastructure, while OpenEvidence adapts the tool to local healthcare conditions and regional disease patterns.

Bridging the resource divide

OpenEvidence founder Daniel Nadler stated that access to medical knowledge should not depend on geography. The platform is described as "context adaptive," designed for physicians who may lack reliable electricity but possess smartphones. This design choice addresses specific infrastructural challenges in the target nations.

Daniela Amodei, President of Anthropic, noted that while the technology has advanced sufficiently to support such initiatives, market incentives alone would not drive this expansion without entrepreneurial and philanthropic efforts.

Usage metrics and partnerships

The collaboration follows significant domestic adoption of the platform. In August alone, U.S. clinicians logged 42 million OpenEvidence consultations. This high volume of usage in developed markets demonstrates the platform's scalability before its rollout in resource-constrained environments.

Strategic context

This move aligns with Anthropic's broader strategy in the healthcare sector. Recently, the company partnered with Novo Nordisk A/S (NYSE: NVO) to accelerate drug research using AI. These developments highlight a growing trend of leveraging large language models for specialized professional services beyond general consumer applications.