Compromised credentials fed to ARTEX let agentic tools reach deep network access in hours
Google Threat Intelligence and Mandiant say attackers supplied compromised credentials to agentic pentesting frameworks such as ARTEX, letting the tools autonomously move through victim networks; one Wiz-tracked campaign reached deep system access in about four hours, once in 15 minutes.
ARTEX ran in compressed bursts of two to 25 seconds, once producing 27 AWS permission events in 23 seconds. It used an application-specific SQL API to point MySQL logs at publicly accessible JSP paths, iterating webshells until commands executed.
The framework also created SSH keys, Elastic IPs, t3.micro jump hosts and plaintext-password administrator accounts, showing credential theft combined with autonomous persistence.
Mandiant Consulting’s attributed defensive implication is long-standing hygiene: control-plane isolation, short-lived credentials, least privilege, segmentation and behavioral detection.
Sources & technical scope 1 source
Google Threat Intelligence — original source ↗
Exploitation: Observed in Mandiant incident-response engagements since late August 2026; KEV catalog exploitation not established by supplied evidence.
Scope & limitations: Product versions, patch availability, patch mechanics, and breadth beyond observed engagements are not established by supplied evidence.
External findings belong to the credited researchers.