AI Agents Accelerate Supply Chain Breach, Granting Privileged Access in Hours

The most recent development in this escalating cyber incident reveals that a China-state-aligned hacking group, designated Storm-0940, utilized an artificial intelligence framework to accelerate its intrusion into corporate networks. According to Microsoft, the group deployed ARTEX, an open-source agentic AI penetration testing tool, to streamline reconnaissance and credential theft. This automated approach allowed the attackers to move with unprecedented speed, with some operations achieving privileged access in less than three hours after initial targeting.

The attack, which centered on the ConnectWise multi-tenant application, began with the theft of application credentials. These credentials allowed the AI agents to authenticate as applications rather than individual users, bypassing standard interactive sign-ins. The group attempted sign-ins at more than 2,250 Microsoft 365 tenants. Simultaneously, the attackers conducted password spraying campaigns, targeting 327 accounts across eight organizations with over 151,000 failed logon attempts. Despite the volume, no password guesses were successful, with nearly 98% of the activity resulting in account lockouts. The use of ARTEX, described as a system of LLM-driven agents, enabled continuous scanning and abuse of OAuth applications with minimal human intervention, leaving defenders with a narrow window for detection.

Storm-0940, active since 2021, has no known industry aliases and targets sectors including aerospace, defense, semiconductors, and government. The breach did not stem from a vulnerability in Microsoft’s own systems but rather from compromised ConnectWise credentials. At its peak on October 11, the password spray executed 2,900 attempts per minute, all of which failed. However, the AI agents leveraged four specific application credentials to perform over 200 sign-ins in under four minutes from a single IP range. Within affected tenants, the agents performed directory operations, changed app credentials, and modified privileged roles.

Security firm Huntress observed specific behaviors in one tenant on October 15, including an application assigning itself to the top administrator role. By October 17, Huntress confirmed unauthorized sign-in activity for almost 1,000 Entra tenants, with over 300 sign-ins succeeding. The sequential, rapid-fire nature of the attacks suggests the use of a script working from a pre-defined target list. This incident highlights a shift toward autonomous attacks, a trend also seen in recent breaches at South Korean financial institutions, where the same ARTEX framework was reportedly used.

The developer of ARTEX, identified by the handle 'Autumn-27,' has announced that the project will cease updates and transition to closed source. Meanwhile, ConnectWise has suspended its Cloud Backup, SaaS Security, and Cloud Management services as a precautionary measure. A phased restoration process has begun, with recovery expected to take several days. ConnectWise has not yet found indicators that partner backup data was compromised, though Microsoft advises customers to disable affected app connections and review recent logs until formal restoration instructions are issued.

The incident underscores the growing reliance on AI in the cyber-attack lifecycle. While the exact number of compromised tenants remains uncertain, the speed at which the AI agents operated suggests a new baseline for threat actors. With the open-source nature of the tool now shifting to closed source, the landscape for automated penetration testing may change, but the immediate impact on the 2,250 targeted tenants remains a critical concern for IT security teams across the region.