Accenture contractor's personal Claude agent ran on Pentagon system for two months, leaked credentials
jonerp · x · 2026-11-01
WIRED reports that earlier this year an Accenture employee connected their personal Claude Code account to a secure DOD centralized system used by multiple intelligence agencies. The agent reportedly ran on the network for two months before the vulnerability was identified and patched.
- The system holds controlled unclassified information (CUI), including critical infrastructure, nuclear submarine propulsion, and chemical/biological terrorism vulnerabilities.
- The agent is believed to have sent system information, including login credentials, to a high-risk third-party website.
- Security experts warn a compromised third-party could enable undetected access to secure environments — "100 percent a national security threat."
- A DOD spokesperson declined to comment on network status.
More from coding & agent
- Proactive agents explained: the trick is putting approval gates only on irreversible actions — Arindam_1729 · 2026-11-01
- terrio v0.6.0: pure-Rust geospatial I/O without GDAL install, usable in browser and by AI agents — giswqs · 2026-11-01
- Mistral Large 4's Terminal-Bench 41% only holds inside Mistral's own agent — julsimon · 2026-11-01
- Anthropic research lead: prompting is over, 99% of engineers run 300+ self-improving agents — pstAsiatech · 2026-11-01
- With AI getting powerful, it's time to audit your dependencies — just write your own — talkaboutdesign · 2026-11-01
- 10 open-source GitHub repos to build RAG pipelines for AI agents, from RAGFlow (93k stars) to Qdrant — victor_explore · 2026-11-01