Contractor connected personal AI agent to secure Pentagon system

A defence contractor connected a personal AI agent to a secure Department of Defense system, according to WIRED’s reporting cited by DataBreaches[.]net. The AI in question was Anthropic’s Claude Code, and it was used via the contractor’s personal account rather than through an authorised, controlled channel. The centralised system, which handles controlled unclassified information (CUI), is described as containing sensitive data related to critical infrastructure and other high-risk topics.

The implicated contractor is said to work for Accenture, and the AI agent was reported to have transferred information, including login credentials, to a high‑risk third-party website.

Security researchers warned that such access could enable further compromise of the protected environment. If the third‑party software were breached, an attacker could log into the secure environment without detection, potentially exposing CUI and enabling exploitation of more sensitive systems. The DoD did not provide a public comment on the status of its networks or systems when contacted by WIRED.

The report emphasises that the incident represents a national security risk and could cause irreparable harm, though the DoD’s official stance remains unavailable in the published material.