Chinese state actor uses AI agents to breach ConnectWise cloud services

The AI-powered breach

A Chinese nation-state actor, tracked by security researchers as Storm-0940, recently compromised Microsoft365 tenants belonging to ConnectWise. ConnectWise is an IT management software provider that helps other companies manage their technology infrastructure.

As reported by infosecurity-magazine.com, the breach forced ConnectWise to take several of its cloud services offline, including Cloud Backup and SaaS Security. The company discovered the suspicious activity on October 15 and identified unauthorized access to application identities within its Cloud Backup environment by October 17.

Microsoft determined that the attackers used a tool called ARTEX. This is an agentic AI penetration testing framework. Unlike standard software, agentic AI can use large language models (LLMs) to perform complex, multi-step tasks autonomously, such as conducting reconnaissance, finding vulnerabilities, and eventually stealing data.

How the attack scaled

The attackers used ARTEX to accelerate their scanning and reconnaissance of the software supply chain. By gaining access to the ConnectWise multi-tenant application, the group was able to move laterally to target the firm's downstream customers. This allowed them to hit hundreds of organizations across various industries, including defense, manufacturing, and financial services.

The speed of the attack was a primary concern for security experts. Microsoft noted that the group attempted to sign in to more than 2250 tenants at a rate of 2900 sign-in attempts per minute. This pace makes it extremely difficult for traditional security systems to flag the activity before the breach is complete.

Once the attackers gained initial access, they established persistence in the cloud environments. They did this by adding credentials to existing OAuth applications. OAuth is a protocol that allows applications to access resources on behalf of a user without sharing passwords. By abusing these trusted connections, the AI agents could authenticate to customer environments using the illicit access they had already acquired.

The ARTEX framework

ARTEX is designed to automate offensive security workflows. It can use various AI models to perform tasks that would normally require a human penetration tester. During a test, the AI agents identify assets and vulnerabilities, then organize that information into a graph to plan the next steps of the attack.

While the user interface for ARTEX defaults to English, the framework is versatile and can be switched to languages including Mandarin, Spanish, and Japanese. This flexibility suggests it is a sophisticated tool capable of being used in diverse geographic regions.

This is not the first time this technology has been spotted in the wild. In October 2026, reports surfaced that a China-based actor used ARTEX to discover vulnerabilities and exfiltrate data from financial organizations in South Korea. The current breach of ConnectWise shows how these tools are being moved from experimental use to active, large-scale operations.

Impact and restoration

ConnectWise has begun a phased restoration of its Cloud Backup service, though they warned customers that the process could take several days. The company stated that existing backup archives remain encrypted and isolated, with no evidence that the data itself was compromised.

The incident serves as a warning for Managed Service Providers (MSPs). Because MSPs manage the infrastructure for many different clients, they are high-value targets. As ConnectWise noted, one trusted connection can open the door to many organizations, making these providers a primary path for nation-state actors seeking to reach sensitive targets.

Customers who use ConnectWise have been advised to review their individual Microsoft 365 tenants. They are encouraged to look for any unauthorized changes to identities, groups, or privileged roles that might indicate the attackers are still maintaining a presence in their systems.