Accenture employee's AI agent reportedly ran on DOD network for two months
- Status
- Unverified: only social posts report this so far
- Kind
- Breach
- Vendor
- Accenture
- Product
- Unnamed DoD system
- First seen
Summary
Sources reportedly say an Accenture employee's AI agent ran on the Department of Defense network for two months before it was identified and the vulnerability could be patched. The sources say Accenture employees did not treat the incident as severe or urgent; the account is not independently confirmed in the item.
If you use Accenture
Inventory AI coding agents and tools connected to internal or government networks
Restrict AI agents to approved accounts and network segments
Review logs for unauthorized AI agent activity over the period in question
Rotate any credentials that may have been exposed for the DoD system
Audit AI agent and contractor access to systems holding CUI
Block or monitor outbound traffic to the third-party site named in reporting once confirmed
What the sources say 6
· Bluesky: “The Accenture employee’s AI agent ran on the DOD network for two months before it was identified and the vulnerability could be patched.” Source
· Mastodon: “It appears that the agent used by the contractor, who is employed by the global consulting firm Accenture, then sent information about the DOD system, including login credentials, to a high-risk, third-party website.” Source
· Mastodon: “Sources tell WIRED that an Accenture employee’s personal Claude agent ran on a DOD network for two months before it was identified and the vulnerability could be patched.” Source
· Bluesky: “A Pentagon AI security breach occurred when an Accenture contractor allegedly connected a personal Claude Code agent to a secure DoD system, exposing data.” Source
· Mastodon: “A Pentagon AI security breach occurred when an Accenture contractor allegedly connected a personal Claude Code agent to a secure DoD system, exposing data.” Source
· Mastodon: “An Accenture employee’s personal Claude agent ran on a DOD network for two months before it was identified and the vulnerability could be patched.” Source