A sophisticated state-aligned cyber offensive has targeted North American IT infrastructure, exploiting trusted management software to penetrate enterprise environments at unprecedented velocities. This week, Microsoft Threat Intelligence published a critical disclosure detailing malicious activity orchestrated by Storm-0940, a China-linked threat group. The threat actors compromised privileged application credentials tied to ConnectWise’s Cloud Backup service (formerly SkyKick), leveraging an autonomous agentic artificial intelligence framework called “ARTEX” to systematically breach downstream business clients. The incident represents a watershed moment in automated corporate espionage, marking the practical realization of AI-accelerated supply chain compromises across the United States and Canada.

The intrusion vector targeted the multi-tenant architecture central to ConnectWise’s backup systems. After acquiring administrative application identities and OAuth tokens within the Cloud Backup network, the threat actors deployed ARTEX—a specialized multi-agent large language model platform designed to conduct autonomous penetration testing, reconnaissance, lateral movement, and data staging. According to Microsoft’s forensic telemetry, ARTEX launched credential attacks and tenant enumeration against more than 2,250 downstream organizations, executing up to 2,900 automated sign-in attempts per minute. In multiple documented instances, the attacker obtained administrative persistence inside corporate Microsoft 365 environments in less than three hours after the initial connection was forged.

For North American organizations, particularly Managed Service Providers (MSPs) and their enterprise clients, this operation underscores a radical shift in operational tempo. ConnectWise isolated the affected infrastructure by taking its Cloud Backup, SaaS Security, and Cloud Management platforms completely offline while incident responders investigated the blast radius. Although the firm affirmed that core encrypted backup archives remained untouched, the secondary risk was immediate: the adversary used valid, elevated identity permissions to inject new administrative credentials, alter service principal roles, and grant persistent API permissions inside corporate customer tenants. This enabled long-term data collection that could survive an upstream software disconnection.

The incident highlights why traditional, human-governed detection cycles fail against automated adversary toolsets. Defending against threats executing at machine speed requires organizations to rethink their trust models regarding third-party cloud management integrations. Enterprise incident response teams are actively reviewing their Microsoft Entra ID and Microsoft 365 audit logs, hunting specifically for unauthorized service principal registrations, modified application permissions, and unfamiliar OAuth grants created under the auspices of ConnectWise integration keys.

Moving forward, corporate security leaders must treat third-party vendor app registrations with zero-trust skepticism. Rather than relying on continuous blanket administrative privileges, enterprises are moving to enforce continuous conditional access policies, aggressive session lifetimes, and behavioral anomaly detection that alerts on sudden bursts of automated downstream API activity.