China-linked hackers turned AI agents and a ConnectWise breach on 2,250+ Microsoft 365 tenants
The rough week in security got rougher. Microsoft says a China-state-aligned group it tracks as Storm-0940 was behind the recent ConnectWise Cloud Backup compromise, and that it used an open-source AI hacking framework called ARTEX to try signing in to more than 2,250 Microsoft 365 tenants, at a pace of up to 2,900 attempts a minute (Microsoft Security Blog).
What happened
ConnectWise is an IT management and remote-access company that many businesses and IT providers use. The attackers got hold of credentials for its cloud applications, the ones customers had already granted access to their Microsoft 365 environments for backups. With those, they could log in as the trusted app, with no user password or sign-in prompt needed (Microsoft, TechCrunch).
Most attempts failed, but Microsoft says the group still got privileged, persistent access at many organizations, in some cases going from first target to privileged cloud access in under three hours. Inside some tenants it ran reconnaissance through Microsoft Graph, changed app credentials and tried to change roles, including an app assigning itself Global Administrator. Those changes can stay behind as a backdoor even after the original app is fixed (Microsoft).
What we don’t know yet
Microsoft hasn’t said how many customers were actually compromised, and it stresses that a successful sign-in doesn’t necessarily mean data was taken. It also says the activity didn’t come from a flaw in Microsoft’s own systems (Microsoft, The Register).
Where ConnectWise stands
ConnectWise first posted an advisory on Oct 16, paused access to the affected cloud services and says it is restoring them. CEO Manny Rivelo told The Register the affected connections are contained, though shutting them off doesn’t automatically remove attackers who already set up a foothold (ConnectWise Trust Center, The Register).
If you run IT
- Keep the affected ConnectWise app connections disabled until ConnectWise says otherwise.
- Don’t delete the enterprise apps first; preserve logs and evidence.
- Review service-principal sign-ins back to at least Oct 15 (ideally Oct 1), plus any new app credentials, role assignments and mailbox rules.
Microsoft’s post has the app IDs, IP indicators and hunting queries (Microsoft Security Blog).
Koffee take: AI tools that test defenses are now being used to break them at machine speed. Multi-factor sign-in and tight app permissions matter more than ever.
Filed Oct 22, 2026 ~4:30 PM CT — Afternoon pass. Sources: Microsoft Security Blog (Oct 21, 2026) · TechCrunch (Oct 22, 2026) · The Register (Oct 22, 2026) · ConnectWise Trust Center advisories.