Skip to content
Cyberttak.
Scams & safety· 3 min read

ARTEX, an open-source AI hacking tool, helped Storm-0940 try to sign in to 2,250 Microsoft 365 tenants

Microsoft says some attacks reached privileged cloud access in under three hours, while Huntress counted about 300 successful sign-ins among nearly 1,000 tenants checked.

By CyberttakPublished · 00:32

In brief

AI agents aimed at 2,250 Microsoft tenants

Microsoft says Storm-0940 used ARTEX to try sign-ins at more than 2,250 tenants. Some attacks reached privileged access in under three hours. Huntress saw about 300 successes.

Low angle view of server cabinets lining a hallway as hands test one handle, with cables and a blue crate on the floor
Low angle view of server cabinets lining a hallway as hands test one handle, with cables and a blue crate on the floorCredit: AI-generated · FLUX.2 klein 9B

Microsoft says a China-state-aligned group called Storm-0940 used an open-source AI tool named ARTEX to speed up a supply-chain attack linked to ConnectWise Cloud Backup. We know this from a Microsoft Threat Intelligence blog post published on October 21, 2026, and from reporting by outlets including The Register and Tom's Hardware.

A supply-chain attack means hitting a trusted supplier to reach its customers. Here, stolen credentials belonging to ConnectWise cloud applications let the attackers act as those apps. No user had to log in.

The numbers

  • More than 2,250 Microsoft 365 tenants (a tenant is one organisation's account space) saw sign-in attempts.
  • Attempts peaked at 2,900 per minute.
  • Some attacks reached privileged cloud access in under three hours.

Most attempts failed. Microsoft says the attackers still managed to gain access in some cases.

What ARTEX does

Tom's Hardware describes ARTEX as a system of AI-driven agents that run a penetration test on their own. A penetration test is a simulated attack that defenders normally use to find weak spots. Here it sped up scanning, password spraying (trying common passwords on many accounts) and abuse of OAuth apps, the permission links between services.

The agents can work around the clock with little human help. Microsoft says that leaves defenders little time to react. The framework can use any AI model, and Microsoft says actors in several regions use it.

Earlier password spraying

Between October 10 and 14, a spray campaign made more than 151,000 failed logons against 327 accounts at eight organisations. Microsoft saw no successes there. The attackers then switched to the ConnectWise app identities.

How many were actually hit?

The sources differ. Microsoft and most outlets speak of attempts against 2,250 tenants. One outlet says 2,250 tenants were breached, which Microsoft's post does not support. Huntress confirmed 162 compromised tenants by October 16. It later saw about 300 successful sign-ins among nearly 1,000 tenants checked.

Microsoft says its own systems were not compromised. It also says a successful sign-in does not prove data was stolen.

Reaction

ConnectWise suspended services and has begun a phased restoration.

The limits: it is uncertain how the ConnectWise credentials were stolen, and the true number of compromised tenants is not settled. Nothing here shows what data, if any, was taken.

Sources