Defense contractor connects personal AI agent to Pentagon security systems
A Defense Contractor Gave Their Personal AI Agent Access to a Secure Pentagon System
According to WIRED, citing informed sources, an Accenture employee connected his personal Claude Code account to the production environment of the Department of Defense's unclassified network NIPRNet, letting an AI agent write and deploy code on his behalf; the agent ran on the network for two months before being discovered and patched. The sources said the agent needed external network access to carry out its tasks, and during that time sent login credentials, passwords and a NIPRNet topology map to a third-party website that NIST has designated as high-risk, while the systems it accessed held controlled unclassified information (CUI). The sources also said Accenture initially did not report the matter to the government; Accenture said the allegations were found not to be substantiated after a thorough independent investigation, and a Department of Defense spokesperson said it does not comment on its network conditions. Claude currently prohibits use by Department of Defense personnel.
The incident fully presents the path by which credentials leak after a personal AI coding agent is connected to a government internal network, and can serve as a reference for institutions assessing the deployment boundaries of similar tools.