Pentagon’s Anthropic Supply Chain Risk Designation Upheld: What It Means for AI in Defense

A Landmark Decision in AI Governance

On September 25, 2026, a federal appeals court in Washington, D.C., delivered a significant blow to Anthropic, upholding the Pentagon’s controversial designation of the AI company as a “supply chain risk.” The 2-1 decision by the U.S. Court of Appeals for the District of Columbia Circuit marks a pivotal moment in the intersection of artificial intelligence, national security, and corporate autonomy. This ruling has profound implications not only for Anthropic but for the entire AI industry and how government agencies will approach AI procurement and security.

The Background: A Relationship Gone Wrong

The conflict between Anthropic and the Department of Defense (DoD) represents a fundamental clash of values and operational requirements. In July 2025, Anthropic signed a $200 million contract with the Pentagon, positioning itself as an early partner for integrating AI into military operations. However, when negotiations began in September 2025 regarding Claude’s deployment on the DoD’s GenAI.mil platform, talks collapsed spectacularly.

The core disagreement centered on usage restrictions. Anthropic wanted assurances that its Claude AI models would not be used for fully autonomous weapons systems or domestic mass surveillance. The Pentagon, conversely, demanded unfettered access to Claude across all lawful military purposes. When the two sides couldn’t reach an agreement, the relationship deteriorated rapidly.

In March 2026, the DoD took an unprecedented step: it designated Anthropic as a “supply chain risk” under the Federal Acquisition Supply Chain Security Act (FASCA). This designation prevents the U.S. military from using Anthropic’s models directly and, more significantly, blocks defense contractors from using them in any work performed for the DoD.

The Legal Battle

Anthropic immediately challenged the designation in two separate courts, reflecting the complexity of the legal framework. A San Francisco federal judge ruled in August 2026 that one of the two designations was illegal. However, the D.C. Circuit Appeals Court upheld the second designation, creating a split decision that leaves Anthropic’s situation in limbo.

The majority opinion, written by Judge Gregory Katsas (appointed by President Trump) and joined by Judge Neomi Rao (also Trump-appointed), found that “the Department had ample support for its conclusion that the continued integration of Claude into the Department’s information systems, by the Department or its contractors, presented a statutorily covered national-security risk.”

Judge Karen LeCraft Henderson, appointed by former President George H.W. Bush, dissented, suggesting that the court’s interpretation of the statute was overly broad.

The Court’s Reasoning

The majority opinion focused on a critical concern: the potential for Anthropic to manipulate Claude’s design to prevent it from performing functions the DoD deemed necessary. Judge Katsas highlighted what he called “deeply sobering” concerns raised by Defense Secretary Pete Hegseth, including the possibility that “overly constrained” AI models could shut down unexpectedly or be “subject to manipulation.”

Crucially, the court interpreted the supply chain risk statute broadly, moving beyond the traditional definition that focuses on adversaries and sabotage. The opinion states: “Whatever paradigmatic examples individual members of Congress may have had in mind, the statutory definition is not limited to ‘adversar[ies],’ and instead covers ‘any person.'” This interpretation significantly expands the government’s authority to designate companies as supply chain risks based on concerns about their operational constraints.

The court essentially sided with the Pentagon’s argument that a company’s refusal to remove safety constraints from its AI models constitutes a potential supply chain risk, even if that company is not an adversary and has no intention of sabotaging military operations.

The Implications: A Chilling Effect on AI Ethics

The decision raises profound questions about the future of AI development and corporate responsibility in the defense sector. Several critical implications emerge:

1. Corporate Autonomy vs. Government Authority

The ruling establishes a precedent that companies cannot unilaterally impose usage restrictions on their products when selling to the government. While the DoD certainly has the right to choose not to purchase from Anthropic, the supply chain risk designation goes further — it prevents any defense contractor from using Anthropic’s products for any purpose, even non-military applications.

This creates a troubling dynamic where companies that prioritize ethical constraints on their technology face potential government retaliation through supply chain designations.

2. The Definition of “Supply Chain Risk”

The court’s broad interpretation of supply chain risk is particularly concerning. Traditionally, such designations were reserved for companies with ties to adversarial nations or those suspected of sabotage. By expanding the definition to include companies that refuse to remove safety features, the government has created a new category of risk that could be applied to any company with ethical guardrails.

This interpretation could have far-reaching consequences for the entire tech industry, not just AI companies. Any company that refuses to implement certain features or capabilities could potentially face similar designations.

3. The Future of AI Safety in Defense

Anthropic’s original position — that Claude should not be used for autonomous weapons or mass surveillance — reflects a growing consensus in the AI safety community about responsible AI development. The court’s decision effectively punishes companies for taking such positions, potentially discouraging other AI developers from implementing similar safeguards.

This creates a perverse incentive structure where companies that prioritize safety and ethics face government penalties, while those willing to implement any capability without question face no such consequences.

4. The Political Dimension

The decision cannot be divorced from its political context. President Trump has been openly hostile to Anthropic and its CEO Dario Amodei, repeatedly criticizing them on social media. Trump’s appointees on the appeals court voted to uphold the designation, while the judge appointed by a Republican president (George H.W. Bush) dissented.

This political dimension raises questions about whether the designation was driven by legitimate national security concerns or by political animus toward Anthropic’s leadership and values.

The Broader AI Industry Impact

While the immediate impact is on Anthropic, the decision sends a clear message to the entire AI industry:

For AI Companies

AI developers must now consider whether implementing safety constraints could expose them to government supply chain risk designations. This creates pressure to develop AI systems without ethical guardrails, at least for government applications.

For Defense Contractors

Defense contractors are now prohibited from using Anthropic’s Claude models for any DoD work. This includes using Claude for software development, code review, documentation, or any other purpose. The breadth of this prohibition suggests that the government views even non-military uses of Anthropic’s technology as problematic.

For the AI Safety Community

The decision represents a setback for the AI safety movement. It suggests that government agencies may view safety constraints as obstacles rather than features, and that companies prioritizing safety may face regulatory penalties.

What Happens Next?

Anthropic has indicated it is “considering all options, including further review.” The company could:

1. Petition for Rehearing: Request that the same panel reconsider its decision
2. En Banc Review: Ask all judges on the D.C. Circuit to review the case
3. Supreme Court Appeal: Petition the Supreme Court to take the case

The appeals court has delayed the decision from taking immediate effect to give Anthropic time to pursue these options. However, the company faces an uphill battle, as the court’s interpretation of the statute appears to give the Pentagon broad discretion in making supply chain risk designations.

The Bigger Picture: AI Governance in Crisis

This case highlights a fundamental tension in AI governance: how should governments balance national security concerns with corporate autonomy and ethical AI development? The court’s decision suggests that national security concerns will generally prevail, even when those concerns are based on a company’s refusal to remove safety features.

This approach differs significantly from how other democracies are approaching AI governance. The European Union’s AI Act, for example, emphasizes responsible AI development and includes provisions for high-risk AI systems. The U.S. approach, as exemplified by this decision, appears to prioritize government access to AI capabilities over ethical constraints.

Conclusion

The appeals court’s decision to uphold the Pentagon’s supply chain risk designation of Anthropic represents a significant moment in the evolution of AI governance. By interpreting supply chain risk broadly to include companies that refuse to remove safety constraints, the court has effectively given the government a powerful tool to pressure AI companies into compliance with its demands.

For Anthropic, the decision is a major setback. For the AI industry more broadly, it signals that companies prioritizing safety and ethics may face government penalties. For the AI safety community, it represents a troubling precedent that could discourage the development of responsible AI systems.

As AI becomes increasingly central to military operations and national security, these questions about governance, corporate autonomy, and ethical development will only become more pressing. The Anthropic case may be just the beginning of a broader reckoning about how governments and companies should work together to develop AI systems that are both powerful and responsible.

The court’s decision stands for now, but the broader debate about AI governance, national security, and corporate ethics is far from over. How this case ultimately resolves — and how other companies respond to this precedent — will shape the future of AI development for years to come.