Personal Claude Agent Ran on a Pentagon Network for Two Months
A personal Claude agent belonging to an Accenture employee reportedly ran on a U.S. Department of Defense network for two months, according to WIRED. The issue was identified and the vulnerability was patched, but the episode raises questions about oversight of personal AI tools on sensitive government infrastructure.
Accenture is a defense contractor, and the report describes the agent as having access to a secure Pentagon system. While the agent may have been intended to assist with routine work, allowing an employee’s personal AI environment to operate on a Department of Defense network expands the potential consequences of errors, data exposure, or unauthorized access.
The incident underscores the need for clear controls on which AI agents can connect to restricted networks, how their activity is monitored, and how quickly access can be revoked when configuration or security issues are found.