Statement from Dario Amodei on our discussions with the Department of War

Source: Anthropic Date: 2026-02-26 URL: https://www.anthropic.com/news/statement-department-of-war

Summary

Dario Amodei published a statement disclosing that the Department of War threatened to remove Anthropic from government systems — and potentially invoke the Defense Production Act — unless Anthropic removed two safeguards: a prohibition on mass domestic surveillance use cases, and a prohibition on fully autonomous lethal weapons systems. Anthropic declined on both counts. Amodei noted the department’s position is self-contradictory: simultaneously labeling Anthropic a security risk while pressuring it to become a compliant defense contractor. Anthropic stated willingness to facilitate transition to another provider rather than remove the safeguards.

Implications

• Marks the first publicly documented instance of a frontier lab refusing a government customer on safety-policy grounds and absorbing the threatened consequence.
• Distinguishes Anthropic’s positioning from competitors: Claude is deployed broadly across national security use cases (intelligence analysis, cyber operations), but the line is drawn at autonomous targeting and surveillance at scale — a specific and legible boundary rather than a blanket policy.
• Sets a reference point for how safety commitments interact with enterprise contract pressure; the question of whether other labs hold equivalent lines is now live.
• Relevant to any organization evaluating AI vendor stability under regulatory pressure: a lab willing to lose a significant government contract for a stated principle is a different risk profile than one that is not.