Microsoft links Storm-0940 to supply chain operation in ConnectWise with ARTEX

Microsoft has confirmed that a group linked to the Chinese State, to which it calls Storm-0940, was responsible for the operation of the supply chain that affected the ConnectWise Cloud Backup application. According to the company, the actor used an artificial intelligence framework agent called ARTICLE to automate at high speed tasks of recognition, password scanning and abuse of OAuth applications, and thus jump quickly from the compromised application to the environments of your customers. These facts, disseminated by Microsoft, place the incident among the most clear cases where agentic AI tools have accelerated offensive operations to scale in cloud environments.

The elements confirmed by Microsoft include that: Storm-0940 used ARTEX at "machine speed" to attack hundreds of organizations since September 2026; that in a specific period (10-14 October 2026) the actor's infrastructure generated more than 151,000 failed authentication attempts, with a peak of 2,900 errors per minute; and that on October 15, agents managed to control the identities of several ConnectWise cloud applications and began to use these identities to authenticate themselves in client environments. Microsoft also specified that in several ARTEX-driven incidents, the actor reached privileged access to cloud infrastructure within less than three hours of the start of the attack. The company has also related Storm-0940 to other previous methods, such as a botnet called Quad7 / CovertNetwork-1658 used in password spray campaigns.

Technically, what happened combines three vectors that are especially dangerous today: operation of the supply chain (commitment to an application with delegated permits), extreme automation via AI agents and abuse of OAuth authorisation models. ARTEX, according to Microsoft, is a Go framework for agentic operations that was closed as a public code after previous abuses. In this incident the actor took advantage that the applications sold (the service identities and their credentials) already had confidence in the customers' tenants; by stealing or controlling those credentials the threat could authenticate without human interaction and operate with the permits already granted. With that door, the actions followed with recognition through Microsoft Graphh, changes in application credentials to establish persistence and modifications on privileged identities, groups and roles.

Who does this affect? Confirmed: managed service providers (MSP) and hundreds of ConnectWise client organizations in very diverse sectors (airlines and defense, automotive, energy, finance, manufacturing, semiconductors, software, NGOs, etc.). More broadly, the attack pattern shows that any supplier with delegated application credentials can serve as a vector to engage multiple customers: the chain is as strong as its weakest link. Microsoft also notes that other bands have started integrating ARTEX into their operations, which amplifies the risk of similar attacks on other vendors and cloud platforms.

Real and concrete consequences. Confirmed and reported: exhaustion of detection windows by high speed of attempts; in some cases, acquisition of privileged access and establishment of back doors through modifications to credentials and roles that may persist even after mitigating the original compromised application. Reasonable estimate (not fully confirmed by Microsoft): The combination of AI + applications with privileges can increase the probability of side movements and exfiltration in less time than human processes take to react, reducing the effectiveness of traditional controls that depend on manual analysis or latency-generated alerts. Still uncertain information: the exact number of customers with effective privileged access, the extension of exfiltered data and the actual duration of back doors in affected environments.

What organizations and users should do now. Some measures are immediate defensive and others are strategic to reduce exposure to such attacks in the future. In practical and concrete order: 1) Review and rotate the credentials of third party applications- revoke client secrets and certificates of suspicious apps and force new administrative authorizations; 2) Review and minimize application permits using administrative consent rather than delegated consent where possible; 3) Enable OAuth consent and application governance policies(control which apps can obtain permits in the tenants); 4) Implement Conditional Access and block authentication from IPs and ranks associated with the actor in addition to restricting access to Graph and other sensitive APIs through risk-based policies; 5) Review sign-in records, changes in service records and Microsoft Graphh operations with priority and keep logs outside the committed environment; 6) Adopt PIM (Privileged Identity Management) practices and periodic role review to minimize accounts that can raise privileges; and 7) Prepare playbooks to respond to incidents focused on abuse of OAuth applications and the complete remediation of committed service principals.

For more technical administrators: audit all applications and service principals in Azure AD, look for recent changes to secrets or certificates, remove unused permits and apply the admin consent policy for third-party applications. Use anomaly detection and cloud application protection solutions such as Microsoft Defender for Cloud Apps and configure real-time alerts about changes in credentials and role assignments. For MSP users and customers: require transparency and mitigation vouchers to your suppliers and change passwords / critical credentials if your MSP confirms exposure.

Useful tools and public resources to implement these recommendations are available in Microsoft's official documentation on OAuth permissions and consent and in Microsoft's technical security coverage: https: / / learn.microsoft.com / azure / active-directory / develop / v2-permissions-and-consent and https: / / www.microsoft.com / security / blog /. It is also recommended to follow the guidelines for the governance and management of suppliers and MSPs published by national cybersecurity organizations and internal access control policies.

In short, what Microsoft has confirmed is two convergent trends: the exploitation of implicit confidence in application identities and the growing integration of AI agents into attack chains operating at machine speed. While some operational details and the total scope of the intrusion remain to be clarified, the lesson for security equipment is clear: minimize permits, monitor and govern delegated applications and accelerate detection with automated controls are essential steps to reduce the opportunity window that tools like ARTEX give to well-funded attackers.