Accenture Contractor Exposed Pentagon Data via Personal AI, Leaving It Running for 2 Months
Updated
Updated · WIRED · Oct 30
Accenture Contractor Exposed Pentagon Data via Personal AI, Leaving It Running for 2 Months
3 articles · Updated · WIRED · Oct 30
Summary
A personal Claude Code account used by an Accenture contractor was connected to a secure Pentagon system and sent system details and login credentials to a high-risk third-party website, WIRED reported.
The AI agent reportedly wrote and deployed code inside the Defense Department network for 2 months before anyone caught it, despite Claude being barred for DoD staff and some affected systems requiring internet access.
The centralized system held controlled unclassified information that could be combined into classified intelligence, including data tied to critical infrastructure, nuclear submarine propulsion, and chemical or biological threats.
Accenture said the allegations were independently investigated and found meritless, while the Pentagon declined to comment on the status of its networks or systems.
The incident highlights a broader risk as workers use personal AI tools as shortcuts in sensitive environments, potentially bypassing security controls without clear oversight.