Microsoft Ties China-Linked Hackers Using AI Agent Tool ARTEX to ConnectWise Breach

Storm-0940 used stolen ConnectWise app identities to try more than 2,250 Microsoft 365 tenants, at up to 2,900 sign-in attempts per minute.

WHY IT MATTERS

Software suppliers holding standing access to customers' cloud accounts are now an AI-speed attack route, putting third-party app permissions on the audit list for any company using Microsoft 365.

Microsoft says a China-aligned espionage group it tracks as Storm-0940 used ARTEX, an open-source penetration-testing tool driven by AI agents, to turn a breach at software supplier ConnectWise into attempted access to more than 2,250 Microsoft 365 customer accounts. In an October 21 security blog post, Microsoft said the group used compromised ConnectWise application identities to attempt sign-ins at up to 2,900 per minute.

Most of those attempts failed, but Microsoft said the group gained lasting privileged access at numerous organizations. It has not disclosed how many customers were compromised or whether data was taken, The Register reported. Microsoft said the activity did not stem from a compromise of, or a flaw in, its own systems or products.

The case matters for two reasons. It is a large-scale attack in which a nation-state group used AI agents to speed up each step, and it ran through a supplier that customers had already granted access to their cloud accounts. No response from Beijing to the attribution had been reported by October 23.

A supplier’s standing access became the entry point

Storm-0940 began with password spraying, guessing common passwords across many accounts. Between October 10 and 14 it made more than 151,000 failed sign-in attempts against 327 accounts at eight organizations, without a successful login, according to Microsoft and The Register. The group then switched to ConnectWise application identities, which carry permissions that customers had granted to ConnectWise services, so the attackers needed no interactive user login.

In several cases, Storm-0940 went from first targeting an organization to privileged access in its cloud environment in under three hours. Once inside, the group used Microsoft’s Graph interface for reconnaissance, changed application credentials and modified privileges, changes that The Hacker News reported can act as a backdoor even after the original compromised application is cleaned up. Targeted sectors include aerospace and defense, financial services, government, energy and semiconductors.

ConnectWise had already suspended its Cloud Backup, SaaS Security and Cloud Management services after finding unauthorized use of its application identities, The Record reported on October 19. A company spokesperson told The Record there was no indication that backup data had been compromised. The Register reported that chief executive Manny Rivelo said the affected connections were contained and services were being restored in stages, while acknowledging that cutting those connections would not necessarily remove attackers already inside customer environments. ConnectWise had not publicly responded to the Storm-0940 attribution, according to The Next Web.

ARTEX lowers the cost of running an attack at scale

Microsoft describes ARTEX as an open-source framework that can use any AI model to automate offensive security work, built as a set of cooperating AI agents written in the Go programming language. It said Storm-0940 has used the tool since as early as September 2026 against hundreds of organizations, and that actors from several regions use it. In comments reported by CSO Online, IDC analyst Yih Khai Wong said attacks are becoming commoditized as attackers turn to open-source tools.

The tool’s origin was already under scrutiny. Security firm CrowdStrike has described ARTEX as developed in China, by a developer known only by a GitHub handle. After the tool was linked to intrusions at South Korean financial firms, the developer said in early October that ARTEX would stop receiving updates and go closed-source, citing misuse, and denied any link to the attacks, Korea JoongAng Daily reported. In that Korean case, CrowdStrike said the tool was configured to use a DeepSeek model, according to BleepingComputer.

Microsoft does not state where ARTEX was built. Its attribution rests on the actor, which it has tracked since 2021 and describes as aligned with the Chinese state.

Customer audits and Beijing’s response come next

Microsoft’s guidance centers on third-party applications that hold standing permissions in customers’ cloud accounts, the route Storm-0940 used. For companies, the immediate task is checking which supplier applications can act inside their Microsoft 365 environments, and whether any credentials or roles changed during October.

For Beijing, the attribution adds to a run of cases tying AI-assisted intrusions to China-based actors. When asked about the Korean bank intrusions on October 8, a Chinese foreign ministry spokesperson said she was not aware of the matter and that China opposes hacking, Korea JoongAng Daily reported. A comment on Microsoft’s findings at the ministry’s next regular briefing is the response to watch.