The Pentagon began testing artificial intelligence models from OpenAI, Google, and other vendors in March following Defense Secretary Pete Hegseth's designation of Anthropic as a supply chain risk. The move came after the Defense Department and Anthropic failed to revise a July 2025 contract, with a January 2026 AI strategy memo pushing Defense Department deals toward standard "any lawful use" terms that Anthropic declined to accept. The Defense Department currently relies on Anthropic's Claude in its Maven Smart System for classified operations, and Anthropic is challenging the risk designation in federal court. US undersecretary Emil Michael said talks with Anthropic remain frozen.
The Contract Dispute and Supply Chain Designation
The supply chain risk label stems from a fundamental disagreement over contract terms. Anthropic maintained red lines that bar the use of Claude for mass domestic surveillance and fully autonomous weapons systems, refusing to adopt the Pentagon's broader "any lawful use" language. This contract dispute triggered Hegseth's designation, marking what appears to be an unprecedented application of supply chain risk laws. Traditionally, such designations have targeted sabotage or subversion in government suppliers, often tied to foreign threats rather than domestic contract disagreements.
Anthropicis challenging the designation in federal court.
Pentagon Testing and Early Results
A senior defense official stated that early tests show models from the alternative vendors answer the same prompts as Anthropic's Claude. The testing began in March as the Pentagon evaluated replacements across multiple AI providers.
Scope of the Designation
The precise scope of the supply chain risk designation remains contested. According to legal analysis, the cited procurement authority would not lawfully block non-defense commercial business with Anthropic, and the designation does not reach commercial contracts outside the defense sector. However, the clash could split the AI market, with some vendors accepting Pentagon terms while others maintain ethical limits and lose access to certain sensitive defense work. Rivals such as OpenAI may pursue defense projects Anthropic leaves behind, though available sources do not confirm any signed replacement deals.
Broader Implications
Human rights groups have warned about the risks of using AI in war. The dispute could make tech companies more wary of national security work if ethical safeguards start to appear as a trigger for procurement penalties.
