A federal appeals court in DC has upheld the Pentagon's designation of Anthropic as a supply-chain risk. This means Claude AI models remain banned from military and federal systems indefinitely. The ruling presents a challenge your legal team must prepare for: government customers can now exclude AI vendors based on contractual restrictions the vendor places on model use, even if those restrictions reflect the vendor's ethical stance.
This isn't just a compliance issue. It's a precedent that changes how you negotiate AI procurement contracts with government entities.
What Changed
The DC Circuit Court of Appeals ruled 2-1 that the Department of Defense acted within its authority when it labeled Anthropic a supply-chain risk. This designation arose from Anthropic's refusal to remove usage restrictions that prevent Claude from supporting autonomous weapons or domestic surveillance. Secretary of Defense Pete Hegseth deemed this position a significant national security risk.
The court's majority opinion stated: "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."
Importantly, the court viewed this as a standard contract negotiation, not a First Amendment violation. The Pentagon "excluded Anthropic from its supply chain based on the company's refusal to assent to a contract term that the Department deemed essential, not based on the company's support for greater governmental regulation of AI technology."
A separate federal judge in San Francisco dismissed one of the parallel supply-chain risk labels in March and confirmed that decision last month. However, Friday's DC ruling means the Pentagon's block continues under the second designation. Both rulings could face years of appeals.
Key Findings
Government customers can exclude vendors for refusing contract terms, even when refusal is based on ethical AI principles. The court rejected Anthropic's due process and free speech claims. Your vendor's commitment to responsible AI practices doesn't protect them from exclusion if those practices conflict with a government customer's operational needs.
Usage restrictions encoded in AI models are material contract terms. The court noted that "Anthropic admits, the company encodes restrictions into Claude that prevent the model from performing tasks that Anthropic wishes to prevent." If your vendor hardcodes limitations into the model, procurement teams will treat those as non-negotiable contract provisions.
Supply-chain risk designations can proceed on national security grounds without detailed public justification. The Pentagon doesn't need to publish comprehensive evidence of the security risk. The designation itself creates immediate procurement barriers across federal agencies and potentially with contractors who serve government customers.
Parallel designations under different statutes require separate legal challenges in different courts. Anthropic had to challenge two supply-chain risk labels in separate jurisdictions. One was overturned in San Francisco; one was upheld in DC. This fragmentation complicates legal strategy and creates prolonged uncertainty.
The market impact extends beyond direct government contracts. Anthropic reported lost revenue because customers were concerned about doing business with a government pariah. Your vendor's designation affects commercial relationships, not just federal procurement.
What This Means for Your Team
If you're procuring AI systems for government use or for commercial applications with potential government customers, you're now operating under a new risk calculus.
Vendor due diligence must include contractual flexibility assessment. Ask prospective vendors: What usage restrictions are hardcoded into your models? Can those restrictions be modified or removed for specific customer requirements? What's your policy on government contracts that require unrestricted model deployment?
Your contract language needs explicit model behavior clauses. Standard software licensing terms don't address whether the vendor can unilaterally restrict what tasks the model will perform. You need provisions that specify whether the vendor retains the right to encode usage limitations and whether those limitations can be overridden.
Government procurement teams should prepare for vendor exclusions based on AI ethics positions. The ruling establishes that agencies can reject vendors whose responsible AI commitments conflict with operational needs. Document the operational requirements that justify the exclusion, and ensure your procurement process treats encoded model restrictions as material contract terms.
Commercial customers with potential government use cases face supply-chain scrutiny. If you're a defense contractor or federal systems integrator, your AI vendor's supply-chain risk status affects your eligibility for government work. Build vendor risk assessment into your AI governance framework and maintain alternative vendor relationships.
Action Items by Priority
Immediate (next 30 days):
- Inventory your current AI vendors and identify any with government supply-chain risk designations or pending challenges.
- Review existing AI procurement contracts for clauses addressing model usage restrictions and vendor-imposed limitations.
- If you're a government customer, document operational requirements that could conflict with vendor AI ethics policies.
Near-term (next 90 days):
- Revise your AI vendor due diligence questionnaire to include questions about hardcoded model restrictions and government contract policies.
- Update procurement contract templates to include explicit provisions on model behavior control and restriction override rights.
- Assess whether your organization's AI systems could face similar challenges if you encode usage restrictions.
Strategic (next 6 months):
- Develop a vendor risk assessment framework that evaluates supply-chain designation risk alongside traditional security and compliance factors.
- For government customers: Establish a process for evaluating whether vendor AI ethics positions create material operational limitations.
- Build relationships with multiple AI vendors to avoid single-vendor dependency in case of supply-chain exclusions.
The Anthropic case won't be the last time government procurement collides with AI vendor ethics policies. Your legal and compliance teams need frameworks now for evaluating when a vendor's responsible AI commitments create contractual risk, and when those commitments align with your organization's own governance requirements. The line between ethical AI practice and supply-chain liability just got sharper.
