Microsoft says a China-aligned group used AI agents in the ConnectWise supply-chain attack

Microsoft Threat Intelligence attributed the recently disclosed compromise of ConnectWise’s cloud backup application to Storm-0940, an actor it described as aligned with the Chinese state and active since 2021. It said the group used ARTEX, an open-source, multi-agent penetration-testing framework that can drive any AI model, to automate attacks on customers downstream of the ConnectWise breach. The attribution is Microsoft’s own; the part of the post that supports the link to ConnectWise is a match between network infrastructure.

Microsoft’s timeline began with a password-spraying campaign between 10 and 14 October against 327 accounts at eight organisations. It recorded more than 151,000 failed sign-ins, peaking at 2,900 failures a minute, and no successful logins. ConnectWise’s advisory followed on 16 October, and one of the IP addresses it listed matched the spraying infrastructure. After taking control of several ConnectWise cloud-application identities, Microsoft said, the group’s agents tried to authenticate to more than 2,250 Microsoft 365 tenants through the permissions those applications already held, so no interactive user sign-in was needed. In some tenants the agents went on to use Microsoft Graph for reconnaissance, adding credentials to applications and changing privileges, which could leave a foothold after the original credentials were revoked.

Most attempts failed, Microsoft said, but the group still obtained privileged cloud access at a number of organisations, including several cases in under three hours from first targeting. Microsoft did not say how many customers were compromised, and cautioned that a successful sign-in does not show data was accessed or taken. It said the activity did not stem from a vulnerability in its own products. The Register reported that ConnectWise had switched off several cloud services after finding unauthorised access to its application identities.

The same tool had appeared two weeks earlier in a separate case, in which CrowdStrike linked ARTEX to attacks on South Korean banks by a financially motivated actor. Microsoft said it had seen ARTEX used by actors in several regions. Its argument was about speed: agents work around the clock and try several routes at once, which leaves defenders little time to respond to the alerts that high-volume activity generates.