Anthropic's response to the Secretary of War and advice to customers.

The U.S. Department of War is moving to designate AI developer Anthropic a supply chain risk, according to a statement from the company. The action follows an impasse in negotiations over Anthropic’s refusal to allow its flagship model, Claude, to be used for mass domestic surveillance or fully autonomous weapons systems. This public conflict between a leading American AI firm and a major government agency marks a significant escalation in the debate over the ethical application of artificial intelligence in national security.

Anthropic stated it held to its two exceptions because it believes current AI models are not reliable enough for use in autonomous weapons, and that mass domestic surveillance violates fundamental rights. The company, which has deployed its models on classified government networks since June 2024, asserts that these exceptions have not affected any government missions to date. In its response, Anthropic also challenged the legal scope of the potential designation, clarifying that under federal statute it could only restrict the use of Claude on Department of War contracts, not prohibit contractors from using the AI for other commercial purposes.

This standoff forces a critical question for the broader technology industry about where the ethical red lines for AI deployment should be drawn, and by whom. Anthropic's decision to publicly resist and legally challenge the Pentagon could compel other AI developers to solidify their own use policies, potentially creating a divide between firms willing to accept unrestricted government contracts and those imposing ethical limitations. The use of a supply chain risk designation, a tool historically applied to foreign adversaries, against a domestic company establishes a new, and contentious, dynamic in government-tech relations.

Anthropic's public conflict with the Department of War moves the discussion of AI ethics from theoretical policy to direct confrontation, establishing a key test case for whether a technology creator can successfully enforce its own safety restrictions on a sovereign government user.