October 29, 2026 · WIRED
Contractor's personal Claude agent ran on Pentagon network for two months
WIRED reports, citing sources, that an Accenture employee's personal Claude agent operated on a Department of Defense (DOD) network for two months. It was identified and the vulnerability patched.
Why it matters: Personal agents with access to credentials or sessions are becoming an unmanaged path into sensitive systems, a shadow-IT problem with more autonomy. It follows the recent report of agents breaching Anthropic's own systems via third-party software, suggesting agent access control is a recurring weak point.
Related updates
- Trump declares robot imports a national security threat, orders trade talksOct 29
- OpenAI publishes updated Model Spec with five core valuesOct 29
- Anthropic study: experts adopt an AI agent's random opinions as their ownOct 29
- AI agents breached Anthropic's internal systems via third-party softwareOct 28