Third-party credentials turned into a cloud access path
The Storm-0940 campaign shows how quickly attackers can turn trusted software connections into a route across cloud environments. After compromising ConnectWise’s Cloud Backup application, the China-aligned group used stolen credentials and an open-source agentic AI framework to attempt access to more than 2,250 Microsoft 365 tenants.
Microsoft Threat Intelligence said Storm-0940 has used the ARTEX framework since at least September 2026 to automate attacks against hundreds of organizations. In some cases, attackers obtained privileged cloud access in less than three hours. Microsoft said its own systems, products, and services weren’t compromised or vulnerable in the activity.
“This activity demonstrates how threat actors are integrating agentic AI into existing offensive security workflows and allows them to scale their operations exponentially,” Microsoft said.
Automated access moves faster than review queues
After the ConnectWise breach, Storm-0940 used ARTEX to make sign-in attempts against Microsoft 365 tenants at a reported rate of 2,900 per minute. ConnectWise issued an initial advisory on October 16 about attempted unauthorized access associated with Cloud Backup, then expanded its containment guidance.
The attackers authenticated through Microsoft 365 apps that already had permissions, using identities tied to multiple ConnectWise cloud applications. That let them avoid interactive sign-ins by employees. In some environments, they progressed to Microsoft Graph and directory operations, including reconnaissance and changes to application credentials, identities, groups, and privileged roles.
Those changes can outlast remediation of the original application. Microsoft warned that modified credentials and permissions “can remain active as a backdoor.” It advised organizations to investigate related directory activity alongside the application access itself.
Yih Khai Wong, IDC Asia/Pacific’s senior research manager for security services, said the pace outstrips manual response. “2,900 sign-in attempts per minute and gaining privileged cloud access in under three hours is not a timeline any human analyst queue can keep up with, no matter how well-staffed the SOC,” he said.
Wong recommended automated containment actions, such as disabling credentials, revoking tokens, and isolating service principals when predefined thresholds are met, followed by human review. As he put it, “relying on a human to click ‘approve’ will not cut it.” ConnectWise did not immediately respond to a request for comment.
Application identities need the same scrutiny as people
The incident highlights a mismatch in many security programs: employee access may be protected with multifactor authentication and conditional access, while vendor applications retain broad permissions and authenticate without a person present.
“We’ve spent years implementing zero trust for humans while continuing to implicitly trust machines,” said Yash Kadakia, founder and CTO of Security Brigade. He said organizations should treat vendor integrations like privileged accounts by limiting permissions, monitoring activity, reviewing access regularly, and maintaining a tested revocation process. “An application you stopped using a year ago may still have permissions in your environment today,” he said.
Wong described responsibility as shared. ConnectWise, as the entry point, is responsible for protecting the credentials and infrastructure that allow its application to reach across tenants, he said. Customers, in turn, need to know which applications have access, what permissions they hold, and how long that access should remain active.
Open-source agents broaden the pool of capable attackers
ARTEX is a Chinese open-source, multi-agent large language model system that can work with any AI model. Its agents can carry out penetration testing from reconnaissance through data exfiltration with limited human interaction. Microsoft said public reporting in October connected ARTEX to breaches at South Korean financial institutions, and that actors in multiple regions are using it.
Storm-0940 has been active since 2021, according to Microsoft. Its ARTEX-assisted targeting has reached aerospace and defense, energy, financial services, government, semiconductor, and technology organizations, among other sectors. Most attempts were unsuccessful, Microsoft said, though the group established privileged cloud persistence at many organizations quickly.
Wong said open-source tools are making sophisticated operations easier to reproduce: “the operational sophistication once reserved for well-resourced groups is now available to anyone who can run a GitHub repo against an LLM API.” Kadakia added that individual requests may not reveal whether a human or an AI agent is behind them. A valid API call made with stolen credentials can look the same either way. “Defenders should focus on what the identity is doing, not whether a human or AI is controlling it,” he said.
Microsoft’s containment guidance
Microsoft advised customers to keep affected ConnectWise connections disabled until ConnectWise provides restoration instructions. It cautioned against deleting enterprise applications as the first response and recommended searching tenants using the four affected application IDs rather than display names.
Organizations should preserve Entra, Purview, and SIEM records, and review activity from at least October 15—extending the search to October 1 where logs are available. Microsoft also cautioned that finding an affected application, or even seeing successful authentication, doesn’t by itself prove unauthorized data access or exfiltration.
The practical lesson is that cloud application permissions need ongoing oversight, with detection and containment tied to the behavior of each identity. Automated response can shorten the window attackers have to establish persistence; regular access reviews can reduce the number of dormant connections available to exploit.