The Digester

A $200 million deal to put Anthropic’s Claude inside Pentagon systems has broken down after the company set limits on how its model can be used, prompting Defense Secretary Pete Hegseth to demand unfettered access and threaten penalties, and raising urgent questions about who should govern military AI.

Anthropic signed a reported $200 million Pentagon contract in July 2025 to deploy its Claude model inside classified military systems.
The dispute centers on Anthropic’s usage rules that prohibit facilitating violence, designing weapons, or enabling mass surveillance.
Defense Secretary Pete Hegseth issued a directive pushing AI firms to remove restrictions and reportedly threatened supply chain penalties or use of the Defense Production Act against Anthropic.
Tension escalated after an employee raised concerns about Claude’s role in a U.S. operation to capture Nicolás Maduro and Palantir alerted the Pentagon.
The dispute has exposed gaps in who sets rules for military AI and prompted calls for congressional oversight of frontier AI in national security

Pentagon and Anthropic Clash Over Limits on Military Use of Claude

Sources