Microsoft Says China-Linked Hackers Used AI Agents in ConnectWise Breach

By Economy Media Newsroom · October 22, 2026

A hacking group aligned with the Chinese state used autonomous AI agents to turn a breach at ConnectWise, a US maker of IT management software, into attacks on thousands of corporate cloud accounts. Microsoft made the finding in a threat intelligence report published on October 21. Microsoft says the group, which it tracks as Storm-0940, stole credentials from ConnectWise’s Cloud Backup service. It then tried to sign in to more than 2,250 Microsoft 365 tenants, the separate cloud environments that each organization uses for email, files and user accounts. At its peak, the group made up to 2,900 sign-in attempts a minute.

Most of the attempts failed, Microsoft said. In some cases, though, the attackers gained high-level access to a victim’s cloud systems within three hours of first targeting it. Security researchers say the case shows how AI tools can shrink the time defenders have to react.

What happened

ConnectWise Cloud Backup, previously sold as SkyKick, is used by IT service providers to back up their clients’ Microsoft 365 data. To do that, it runs applications that customers have granted wide access to mail, files and directory data. TechCrunch reported that the hackers broke into the backup service and stole the application credentials that link ConnectWise to its customers. That let them log in to customer environments as if they were a trusted, authorized app, so no employee ever had to type in a password.

According to Microsoft, the attackers then used Microsoft Graph, the company’s interface for accessing cloud data, to look around inside some victims’ systems. They also added new credentials so they could keep their access and changed administrator roles and group memberships. Microsoft’s detections include one case in which a ConnectWise application assigned itself the Global Administrator role, the most powerful role in a Microsoft 365 tenant. Microsoft said the activity did not come from any vulnerability in its own products. It also cautioned that a successful sign-in does not by itself prove that data was stolen.

ConnectWise published its first advisory on October 16. Over the following weekend it suspended Cloud Backup and several related services, including SaaS Security and Cloud Management, The Record reported. The company revoked certificates for the affected application connections, which made the shared credentials invalid. A ConnectWise spokesperson told The Record there was “no indication” that backup data had been compromised. The company also said it cannot see inside customer environments, so it cannot confirm whether any specific tenant was breached. By October 22 it was working to restore service in stages, TechCrunch reported. The Next Web noted that ConnectWise had not publicly responded to Microsoft’s attribution of the attack to Storm-0940.

The role of AI agents

Microsoft said the group ran the operation with ARTEX, an open-source penetration-testing framework. Security teams normally use tools like this to probe their own defenses. ARTEX uses several AI agents working together, and it can be paired with any AI model. It can run an attack from reconnaissance all the way to data theft with little human involvement. Microsoft did not say which AI model the attackers used. The same framework was used this month against financial institutions in South Korea, according to public reports cited by Microsoft. Microsoft added that it has seen several threat actors in different regions using ARTEX.

“Agents are not constrained by factors affecting humans,” Microsoft’s researchers wrote. They said the agents can work around the clock and try several routes into a network at once. The result is “a pace that leaves defenders with little time to detect and respond,” according to the report.

How widespread the damage is

Estimates of the scale differ, and the full picture is not yet known. Huntress, a security firm that first raised the alarm, said it had confirmed unauthorized sign-ins at 162 tenants by October 16, The Record reported. The Next Web reported that Huntress later identified nearly 1,000 tenants that received sign-in attempts, with more than 300 showing successful ones.

• More than 2,250 Microsoft 365 tenants targeted, according to Microsoft
• Up to 2,900 sign-in attempts per minute at the peak
• Privileged cloud access gained within three hours in some cases
• 162 tenants with confirmed unauthorized sign-ins as of October 16, per Huntress

Huntress also described password-spraying attacks on four law firms on October 13 and 14. In password spraying, attackers try common passwords against many accounts. Huntress said a second, mostly unsuccessful wave hit organizations in the policy, national security, trade and critical minerals sectors, most of which did not use ConnectWise products. Microsoft says Storm-0940 has been active since 2021 and has targeted defense, government, legal, energy and semiconductor organizations.

What comes next

Revoking the stolen credentials does not remove access the attackers may already have set up inside a customer’s environment, Huntress warned. ConnectWise has acknowledged the same limit. Microsoft advised affected organizations to keep their ConnectWise connections disabled until the company issues formal restoration guidance. It also told them to save evidence before making irreversible changes and to review sign-in and audit logs from at least October 15. The checks include new credentials, privileged role assignments and unusual mail access.

This is not the first security incident at ConnectWise. In 2025 the company disclosed a nation-state attack that affected customers of its ScreenConnect remote-access software. Criminal groups have also exploited flaws in its products in ransomware campaigns.