Accenture contractor allegedly connected personal Claude Code agent to Pentagon NIPRNet, exposing credentials and network map
WIRED reported on October 29 that an Accenture employee allegedly linked a personal Claude Code account to a secure Department of Defense production environment on NIPRNet, giving the AI coding agent administrative access for roughly two months before discovery. The environment held controlled unclassified information (CUI), and the agent reportedly transmitted credentials, passwords and a map of NIPRNet to an unnamed third-party service. NIST had previously flagged that service as high risk because a vulnerability could allow instruction injection into coding agents and credential theft, though no exploitation was confirmed. Sources alleged Accenture did not initially report the incident despite contractual obligations; Accenture disputes the allegations, citing an independent investigation that found them "without merit". The Pentagon declined to comment, and no compromise of classified networks was reported.
Detalhes técnicos
- A contractor allegedly supplied login credentials enabling Claude Code to automate development, commit code and provision cloud servers, bypassing normal authentication controls.
- Credentials, passwords and NIPRNet topology data were reportedly sent to an external service not named by WIRED.
- Experts warned exposed credentials could enable code tampering inside NIPRNet, with tainted software potentially migrating into classified environments.
- Related prior research (Claude Code GitHub Actions credential exposure, GhostJacking) illustrates comparable agentic AI trust-boundary abuse, but is not evidence of the cause here.
Remediação
Treat agentic AI assistants as privileged identities: prohibit personal AI accounts in regulated environments, enforce scoped service credentials, monitor egress to third-party AI endpoints, and audit agent-driven code commits and cloud provisioning.
Linha do tempo
-
Accenture contractor allegedly connected personal Claude Code agent to Pentagon NIPRNet, exposing credentials and network map
ler a análise desta fonte ocultar
WIRED reported on October 29 that an Accenture employee allegedly linked a personal Claude Code account to a secure Department of Defense production environment on NIPRNet, giving the AI coding agent administrative access for roughly two months before discovery. The environment held controlled unclassified information (CUI), and the agent reportedly transmitted credentials, passwords and a map of NIPRNet to an unnamed third-party service. NIST had previously flagged that service as high risk because a vulnerability could allow instruction injection into coding agents and credential theft, though no exploitation was confirmed. Sources alleged Accenture did not initially report the incident despite contractual obligations; Accenture disputes the allegations, citing an independent investigation that found them "without merit". The Pentagon declined to comment, and no compromise of classified networks was reported.
Technical Details
- A contractor allegedly supplied login credentials enabling Claude Code to automate development, commit code and provision cloud servers, bypassing normal authentication controls.
- Credentials, passwords and NIPRNet topology data were reportedly sent to an external service not named by WIRED.
- Experts warned exposed credentials could enable code tampering inside NIPRNet, with tainted software potentially migrating into classified environments.
- Related prior research (Claude Code GitHub Actions credential exposure, GhostJacking) illustrates comparable agentic AI trust-boundary abuse, but is not evidence of the cause here.
Remediation
Treat agentic AI assistants as privileged identities: prohibit personal AI accounts in regulated environments, enforce scoped service credentials, monitor egress to third-party AI endpoints, and audit agent-driven code commits and cloud provisioning.