The Pentagon awarded AI agreements to eight companies covering analysis, logistics, and large-scale classified data processing. Anthropic was excluded after it refused to waive safety guardrails that would restrict use of its models in autonomous weapons and mass surveillance contexts; the department responded by labeling Anthropic a supply-chain risk. Anthropic is contesting that designation in court, making this the first known instance of a safety-policy disagreement translating into a formal procurement exclusion. The contracts carry no independent enforcement mechanism for the non-use commitments stated by the participating vendors, leaving the governance of AI in military contexts largely self-policed. For technology leaders, the episode signals that AI procurement decisions are now bound up with safety and ethics positioning: companies that adopt harder safety stances may face commercial exclusion from defense markets, while those that participate accept reputational exposure without binding safeguards. Engineering and product leaders advising on AI strategy will need to treat military-use policy as a live commercial variable, not a hypothetical.