Hackers let AI agents run a real attack. Microsoft says it hit 2,250 organizations
· 2 min read · Drafted with AI help, reviewed by Kepler Oates
For years, security people warned that AI would let attackers move faster. This week, Microsoft published a detailed account of it happening. A China state-aligned hacking group it calls Storm-0940 used a framework of AI agents to attack organizations "at machine speed," including through a breach of ConnectWise, a company whose software many IT providers use to manage and back up their customers' systems.
What happened
The tool. The group used ARTEX, an agentic AI penetration-testing framework: software that strings AI agents together to scan networks, guess passwords and abuse app permissions with little human input. Microsoft says it works with any AI model. The Hacker News reports its developers made it closed-source after it was abused against South Korean banks, and Zscaler found versions configured for DeepSeek, GLM and Grok.
The password spray. Between October 10 and 14, the agents tried common passwords against 327 accounts at eight organizations. Microsoft counted more than 151,000 failed logons, and saw no successful sign-ins in that phase.
The breach. On October 15, the agents took control of identities belonging to ConnectWise cloud backup applications and used them to reach customer environments. ConnectWise issued an advisory on October 16 and temporarily shut down services, restoring them in phases.
InteractiveHow the attack unfoldedTap a step, or use Next.
Agents start working Storm-0940 begins using ARTEX against organizations in defense, energy, finance, government and tech.
Password spray Agents try passwords on 327 accounts at eight organizations: 151,000+ failed logons, 111,000 of them on October 11.
Apps hijacked Agents take over ConnectWise cloud backup app identities and use them to reach customer environments.
Advisory ConnectWise warns customers and shuts down services, then restores them in phases.
Speed. The agents attempted sign-ins at more than 2,250 Microsoft 365 tenants, at up to 2,900 attempts a minute. In several cases, Microsoft says, attackers went from first contact to privileged cloud access in under three hours.
Persistence. Once inside, the agents changed app credentials, groups and admin roles. Microsoft warns that added app credentials can act as backdoors that survive after the original hole is closed.
It keeps going. Wiz reported sessions lasting 24 to 30 hours, with new IP addresses appearing within minutes when old ones were blocked. A tired human attacker stops; an agent doesn't.
The good news: most of the targeting failed. The basics stopped it.
What it means for you
Passwords that can be guessed will be guessed. A password spray is just trying common passwords on many accounts. Use a password manager and a unique, long password everywhere.
Turn on app-based multi-factor or a passkey. It's the single step that stops most of these attacks, even when the password leaks.
Review connected apps. In Google, Microsoft and social accounts, check which apps have access and remove the ones you don't use. That's exactly the kind of access these agents abused.
If you run IT for a business: Microsoft says to keep ConnectWise app connections disabled until you've reviewed them (don't delete the apps first, so you keep evidence), review sign-in logs from October 15 (October 1 if possible), audit app credentials and cut admin rights to the minimum.
Not financial advice. This summarizes public reporting for information only. It is not investment, tax or legal advice. Do your own research or talk to a licensed professional.
