Microsoft has attributed a supply-chain attack on IT management software company ConnectWise to a China-state-aligned hacking group it calls Storm-0940, saying the attackers used AI agents to break into customers' systems at "machine speed". Microsoft disclosed the findings on 21 October 2026, according to CyberScoop and The Hacker News.
How the attack worked
- The group used ARTEX, a Go-based "agentic" AI penetration-testing framework that can plan and carry out attack steps on its own. It was made closed-source after being abused against South Korean banks, The Hacker News reported.
- On 15 October, the AI agents took over the identities of several ConnectWise cloud applications and used them to get into the environments of ConnectWise Cloud Backup customers.
- Sign-in attempts reached more than 2,250 Microsoft 365 tenants, at about 2,900 attempts per minute.
- Earlier, from 10 to 14 October, a password-spraying campaign targeted 327 accounts at eight organisations, producing more than 151,000 failed log-ins.
- Some intrusions reached privileged access in under three hours.
ARTEX was configured to use AI models including DeepSeek, with GLM and Grok in other setups, with models checking each other's work, according to The Hacker News.
Who was targeted
Targeted sectors included defence, government, energy, finance, semiconductors, legal services and think tanks. Microsoft did not say how many customers were compromised and noted that a successful sign-in does not necessarily mean data was stolen.
"A wave of AI-driven cyberattacks has been ongoing," said Microsoft executive vice president Igor Tsyganskiy, describing this attack as "one of many," CyberScoop reported. Microsoft added that several threat groups in different regions now use ARTEX.
ConnectWise response
ConnectWise shut down affected services on Saturday and began restoring them in phases on Wednesday, a process it expects to take several days. The company said nation-state actors are "increasingly targeting the software supply chain."
Why it matters
The case is one of the clearest examples yet of AI agents being used to automate real attacks, compressing work that once took days into hours. It follows this week's scramble over AI-discovered PHP vulnerabilities; see our report on PHP2Shell.
What organisations should do
- Review sign-in logs for unusual activity from third-party cloud apps.
- Limit the permissions granted to connected apps and rotate their credentials.
- Enforce multi-factor authentication and block password-spraying with lockout and conditional-access rules.
