The Pentagon has begun asking major defense contractors—including Lockheed Martin and Boeing—to assess how dependent they are on Anthropic’s AI services, a step that signals the Defense Department is seriously weighing whether to treat the company as a potential “supply-chain risk.” The outreach was first reported by Axios, which said the inquiries come as the government presses Anthropic to change the rules governing how its AI can be used in military contexts.
At the core of the dispute is Anthropic’s insistence on keeping strict usage safeguards for its model Claude. The Department of Defense wants Anthropic to relax restrictions that the company says are meant to prevent certain uses—such as autonomous weapon targeting and forms of domestic surveillance—that the company deems unacceptable. Anthropic, according to the reporting, has resisted those demands and has indicated it does not intend to substantially soften its policies.
The tension escalated after a meeting between Defense Secretary Pete Hegseth and Anthropic CEO Dario Amodei. Following that meeting, the Pentagon set a deadline—Friday at 5 p.m. ET—for Anthropic to respond to new government-imposed terms. Pentagon officials have discussed a range of possible actions if Anthropic refuses, including formally designating the company a supply-chain risk or exploring legal mechanisms that could compel changes in how Anthropic governs its AI use.
Why is the Pentagon focusing on this now? Claude is currently the only AI model integrated into certain classified U.S. military systems, which makes the relationship unusually sensitive: if the Pentagon concludes it cannot rely on Claude under its current restrictions, it may need to shift to other providers—potentially disrupting projects, procurement plans, or contractor workflows. That helps explain why the Defense Department is checking whether key contractors would be affected if Anthropic were restricted or blacklisted.
Apparently, Boeing said it has no active contracts with Anthropic, though the broader inquiry suggests the Pentagon is not only looking for direct contracts but also indirect dependence through subcontractors, platforms, or embedded tools.
As the Pentagon pushes to deploy AI more broadly—including on classified networks—it is increasingly colliding with commercial AI firms’ safety policies and reputational concerns. Some rival providers with more permissive terms are trying to gain ground in defense contracting, increasing the pressure on Anthropic from both the government and competitors.
