An Accenture employee reportedly connected a personal Claude AI agent to a secure Pentagon network. Sources told WIRED the agent ran on a Department of Defense system for two months. The vulnerability was identified and patched only after the agent was discovered.
The incident involved a personal AI agent, not an officially sanctioned government tool. It operated on a DOD network without authorization for an extended period. The breach raises questions about how personal AI tools access sensitive systems.
Accenture is a major defense contractor that works closely with federal agencies. The employee’s Claude agent was reportedly used without formal approval. Details on what data the agent could access remain unclear.
The agent went undetected for roughly 60 days before security teams flagged it. Officials then moved to patch the vulnerability. The timeline suggests monitoring gaps for unauthorized AI activity on secure networks.
The incident highlights growing risks as employees adopt AI assistants in workplace settings. Personal AI agents can bypass standard security reviews and oversight. Defense networks are especially vulnerable to unvetted tools.
The Pentagon has been expanding its use of AI while tightening security rules. Unauthorized AI deployments on sensitive systems remain a persistent concern. The Accenture case shows how easily such tools can slip through.
The vulnerability has been patched, but the incident underscores the need for stronger controls. Agencies must track and approve AI tools used on their networks. The case is likely to fuel calls for stricter AI governance in defense settings.
