How Storm-0940 Used ARTEX AI to Abuse ConnectWise Credentials

Microsoft disclosed that Storm-0940, a China state-aligned threat actor, compromised the ConnectWise Cloud Backup application to launch a rapid supply chain attack. The group utilized ARTEX, a Go-based agentic AI penetration testing framework, to accelerate reconnaissance and credential abuse. This campaign targeted hundreds of organizations across multiple sectors, including aerospace, energy, and financial services, since September.

How did the attackers bypass standard authentication?

The attackers did not require users to enter passwords. Instead, they abused stolen application credentials and permissions that customers had already approved for the ConnectWise service. According to Microsoft, this allowed the threat actor to authenticate as trusted applications within target environments. This method enabled the group to operate without an interactive user sign-in, effectively leveraging existing trust relationships to gain entry.

What speed and scale did the campaign achieve?

Microsoft reported that Storm-0940 targeted more than 2,250 Microsoft 365 tenants with a velocity of 2,900 sign-in attempts per minute. This pace, described as "machine speed," gave defenders a very short window to detect and respond to the intrusion. In several instances, the time from initial targeting to gaining privileged access to cloud infrastructure was less than three hours. Zscaler and Microsoft published separate reports on October 21 detailing this ARTEX abuse.

What actions did the actor perform inside networks?

After gaining control of ConnectWise application identities, the attackers used Microsoft Graph to inspect environments and modify application credentials. They attempted to change privileges and add credentials to existing OAuth applications to ensure persistent access. Microsoft stated that these AI-accelerated activities included reconnaissance and attempts to expand access even after the original connection might be disabled. This behavior aims to preserve entry points through other applications.

What is the history of this threat actor?

Storm-0940 was first detailed by Microsoft in November 2024, with activity assessed since at least 2021. The group has previously used the Quad7 botnet to conduct evasive password spray attacks. In this incident, however, they shifted focus to exploiting internet-facing applications and abusing OAuth permissions. Microsoft observed no successful authentication during a separate password-spray activity between October 10 and 14, which generated over 151,000 failed logons.

What analysts watch for

Security teams usually examine sign-in logs for high-frequency authentication attempts from known service principals or OAuth applications. Analysts should look for unusual changes to application permissions, such as new secrets or certificate updates, especially if made by automation. Persistence often involves adding new credentials to existing trusted apps. Check for Microsoft Graph API calls that inspect directory roles or modify group memberships. Rapid privilege escalation following a service compromise is a key indicator.

Background: Agentic AI

Agentic AI uses a large language model as a reasoning engine to plan, call tools, and observe results in a loop. It does not just generate text; it executes functions. You control it by defining available tools, setting strict output schemas, and monitoring the feedback cycle for drift.

Read the full guide: How Agentic AI Works: The Loop, The Limits, And The Leak

What to do and how to stay safe: Storm-0940

• Review OAuth application permissions for any recent changes to secrets or certificates that you did not authorize.
• Monitor sign-in logs for high volumes of authentication attempts from service accounts or application identities.
• Verify that no new credentials have been added to existing trusted applications in your cloud infrastructure.
• Check for unauthorized modifications to privileged roles or group memberships using Microsoft Graph API logs.

Step-by-step guide: 7 Privilege Escalation Mistakes That Leave Systems Exposed

General security guidance from the Sandbox Dispatch newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Did the attackers steal user passwords in this campaign?

No, the attackers abused pre-approved application credentials and OAuth permissions, bypassing the need for interactive user sign-ins or password entry.

How many tenants were targeted by this attack?

Microsoft reported that Storm-0940 attempted access to more than 2,250 Microsoft 365 tenants following the ConnectWise Cloud Backup breach.

Is ARTEX a new malware family?

No, ARTEX is an AI-powered penetration-testing framework that was made closed-source after being abused against South Korean financial institutions.