Threat Actors Deploy ARTEX and PentAGI to Automate Real-World Intrusions

Google Threat Intelligence Group and Mandiant reported that financially motivated attackers and suspected cyberespionage operators have used agentic offensive frameworks, including ARTEX and PentAGI, in real-world intrusions since at least late August 2026. The frameworks automated reconnaissance, privilege escalation, lateral movement, persistence, and data collection or exfiltration preparation. An ARTEX-associated campaign tracked as UNC7228 affected organizations across the United States and Europe and overlaps with Microsoft’s PRC-nexus STORM-0940 activity.

Observed automation compressed individual execution sequences into seconds but frequently produced conspicuous logs, repeated errors, and recognizable artifacts. The findings indicate early-stage adoption of multi-stage automation, rather than evidence of fully autonomous attacks. Recommended defenses include isolating control planes, enforcing short-lived and least-privileged credentials, segmenting networks, restricting application abuse, and correlating telemetry to detect suspicious behavior across attack stages.