Storm-0940, a state-aligned threat group, began using the ARTEX agentic AI penetration testing framework to automate attacks on hundreds of organizations in September 2026 (Microsoft Security Blog). Microsoft says the China-aligned group leveraged the autonomous framework to gain privileged cloud access across multiple targets in under three hours from initial engagement. The campaign is linked to a recent supply chain compromise involving a ConnectWise multi-tenant application.
Deploying ARTEX signals a move toward operationalizing autonomous security software in offensive campaigns. Autonomous agents allowed operators to run without human schedule limits, testing multiple compromise routes simultaneously across diverse global industry verticals. These mechanisms echo traditional supply chain compromise techniques, where attackers manipulate products or software before delivery to reach a broad victim surface (Mitre).
After establishing an initial foothold, the actor abused pre‑authorized application permissions to authenticate directly into Microsoft 365 customer environments. Because the compromised applications already had approvals, the activity bypassed interactive user sign‑in. Adversaries often use supply chain footholds to launch follow‑on attacks against downstream components and connected systems.
Following automated entry, Microsoft observed AI‑accelerated post‑exploitation activity routed through Microsoft Graph, including target reconnaissance and credential modifications. Operators also tried to alter application privileges to secure persistent, elevated access. Historically, Storm‑0940 has maintained persistence in cloud networks by attaching new credentials to existing OAuth applications to reach organizational resources.
Microsoft confirmed that the malicious operations conducted with ARTEX did not stem from any direct vulnerabilities or compromises within Microsoft corporate systems or products. Detecting Storm‑0940 authentication or an affected application does not automatically prove internal customer data was exfiltrated. The full extent of data accessed across affected multi‑tenant environments remains undisclosed.
Security teams should inspect service principal sign‑in logs, track unauthorized credential additions to enterprise applications, and follow vendor‑published containment procedures while preserving forensic logs. Investigators can use hunting queries targeting anomalous Microsoft Graph activity to spot unauthorized access paths established by the threat actor.
The integration of the ARTEX framework by Storm-0940 suggests that agentic offensive tools are significantly compressing dwell time windows for defenders. By reducing initial compromise to privileged cloud access to under three hours, the actor appears capable of outrunning manual incident response workflows. This operational shift likely indicates that cloud security teams must increasingly rely on automated identity posture management and real-time credential revocation rather than traditional reactive log reviews.
Spotted an error? See our corrections policy or email editorial@lordofclaude.com.
Discussion & Comments (0)
Join the conversation