Amodei Rejects Pentagon Demands

On February 26, 2026, Anthropic CEO Dario Amodei published a company-wide memo — simultaneously released to media — stating unequivocally that Anthropic “cannot in good conscience accede to the Department of Defense’s demands as currently articulated.” The memo represented the first time a major AI company CEO had publicly refused a Pentagon directive during an active contract dispute.

Amodei drew two explicit red lines. First, Anthropic would not remove guardrails that prevent Claude from being used for mass surveillance of domestic or foreign civilian populations without judicial authorization. Second, Anthropic would not enable autonomous weapons systems — defined as systems that select and engage targets without meaningful human control over each engagement decision. Amodei emphasized that these were not negotiating positions but foundational commitments that predated the Pentagon contract and would survive its termination.

The memo acknowledged the gravity of the situation. Amodei wrote that Anthropic “deeply values its partnership with the Department of Defense” and recognized the “legitimate national security interests” at stake. But he argued that the specific demands being made would set a precedent that would ultimately undermine national security by establishing that AI companies must build whatever the government demands without regard to catastrophic risk. “If we remove these safeguards today because we are told to, we will have no credible basis to maintain any safeguards tomorrow,” he wrote.

The response from Washington was swift. Within hours, unnamed Pentagon officials told multiple outlets that Anthropic had “made its choice” and would face consequences. The Friday deadline passed without compliance, setting the stage for the supply chain designation that followed the next day.