# ConnectWise breach and AI agents hit Microsoft 365 tenants

> Microsoft says a China-aligned group used stolen ConnectWise application credentials and an AI tool, ARTEX, to target more than 2,250 Microsoft 365 tenants. How many were breached is not disclosed.

Published 2026-10-23, updated 2026-10-23. Confidence: medium. When it could matter: Immediate for affected customers; wider effects timing unknown.

## In plain words
A company that provides IT management and backup services, ConnectWise, was breached. Microsoft says a China-aligned hacking group then used stolen access keys, and an AI tool that automates attacks, to try to get into more than 2,250 business accounts on Microsoft's cloud. Most attempts failed, but some succeeded. Nobody has said how many companies were affected or whether data was taken. If your organisation uses ConnectWise cloud services, check your Microsoft 365 settings for changes you did not make, because switching off the original connection may not remove an intruder who is already inside.

## What happened
Microsoft says the attackers, tracked as Storm-0940, used an open source AI penetration-testing framework called ARTEX. The Register describes it as written in Go and built to automate reconnaissance, vulnerability discovery, credential attacks and data theft. Microsoft says the group began using it as early as September against organisations worldwide, including government agencies, defense contractors, banks, manufacturers and tech firms. The Register adds that Storm-0940 has operated since at least 2021.

Between October 10 and 14, the group made more than 151,000 failed sign-in attempts against 327 accounts at eight organisations, by password spraying. Microsoft saw no successful logins from that campaign. The attackers then moved to compromised ConnectWise application identities. These let them use permissions that customers had already granted to ConnectWise services, with no interactive user login. The Register reports attempts against more than 2,250 Microsoft 365 tenants, at times reaching 2,900 attempts a minute. Microsoft also saw AI-accelerated activity through Microsoft Graph, including reconnaissance, application credential changes and attempts to modify privileges. In some cases the group went from picking a target to privileged cloud access in under three hours.

ConnectWise pulled its Cloud Backup, SaaS Security and Cloud Management services after finding unauthorised access to application identities. Its CEO, Manny Rivelo, said the affected application connections were contained and services were being restored gradually. TechCrunch reports the same pause and investigation.

## What is confirmed and what is not
According to the reports, Microsoft states the following: the attribution to Storm-0940, the use of ARTEX, the volume of sign-in attempts, the use of compromised ConnectWise application credentials, and persistent privileged access at numerous organisations. ConnectWise confirms it paused services, contained the affected connections and is restoring them.

Not established: how many customers were compromised, and whether any data was stolen. The Register says Microsoft cautions that a successful login does not necessarily mean information was accessed. ConnectWise did not initially name who was responsible, and the sources show no ConnectWise statement accepting Microsoft's attribution.

The two outlets differ in tone and in some figures. TechCrunch says the hackers used AI agents "to rapidly hack into around 2,250 organizations" and gives the rate as almost 3,000 attempted sign-ins per minute. The Register says 2,250 was the number of tenants targeted, at times 2,900 attempts a minute, and that most attempts failed. The more careful reading is attempts against that many tenants, not that many breaches. TechCrunch also says the same software was used in a recent compromise of South Korea's financial systems. The Register does not mention this, so it rests on one report here. The sources also differ on how the access was gained: TechCrunch describes theft from ConnectWise's cloud backup service, while The Register refers to application identities. These may be the same thing described differently, but the sources do not say.

## Analysis
### Customers who may still have an intruder inside
The most direct consequence falls on organisations that connected ConnectWise cloud services to Microsoft 365. The attackers did not need to break each customer's defences. They used credentials that customers had already authorised, so the sign-in looked like a trusted service. Microsoft's observation that the group changed application credentials and tried to modify privileges is the key detail. If a new credential or permission was added, then revoking the ConnectWise connection removes only one door. The Register reports that Microsoft warns this could let attackers remain after the original credentials were revoked, and that ConnectWise itself said shutting down the connections would not necessarily remove attackers already established. Because Microsoft has not said how many customers were compromised, an organisation cannot treat silence from ConnectWise as a clean result. This is TOSWO Research's reading of the mechanism, not a finding that any particular reader is affected.

### Security and the supply-chain pattern
The incident follows a familiar chain: compromise a vendor, take the application credentials it holds, then reach many customers at once. The credentials mattered more than the AI. Microsoft's own summary, as reported, is that the techniques are not new: stolen credentials, password spraying and application abuse. Its claim is that AI let Storm-0940 run them faster. That claim is worth testing. The sources show speed figures, such as under three hours to privileged access, but give no comparison with the group's earlier campaigns. A sceptical expert would ask how much of the pace came from ARTEX and how much from the ease of using valid credentials against many tenants. The sources do not answer this.

The password-spraying phase is also instructive. More than 151,000 failed attempts produced, by Microsoft's account, no successful logins. The switch to vendor-held application identities is what changed the result. If that holds beyond this case, then strong user-level controls will not by themselves protect against a trusted third-party connection that has been hijacked. That is general reasoning, and the sources do not test it.

### Technology and industry
ARTEX is open source and, per The Register, freely available for penetration testing. Microsoft says attackers from several geographic regions have used it. The mechanism is dual use: a tool written to stress-test defences automates the same steps an attacker needs. Defenders can use it too, and the sources give no evidence on whether that offsets the risk. For IT service vendors, the incident raises the cost of holding long-lived application credentials that reach customer cloud environments. ConnectWise has been here before: in May 2025 it admitted a nation-state intrusion affecting a small number of customers using ScreenConnect. The Register says there is nothing to suggest the two attacks are linked, so the sources do not support reading this as one continuing campaign.

### Law and policy
The sources report no regulator action, lawsuit or disclosure deadline. The open question for affected organisations is whether they must treat this as a reportable incident. That depends on whether data was accessed, which Microsoft has not said. Nothing here should be read as legal advice.

### Weakest links in the story
Three links are weaker than the headlines suggest. First, the number 2,250 measures attempts, not victims. Second, the attribution to a China-aligned group comes from Microsoft, and ConnectWise did not initially name anyone. Third, the AI speed-up is asserted by Microsoft without a published baseline in these reports. The strongest reason this view could be wrong is that Microsoft may hold telemetry that shows a larger or smaller breach count than the reports convey. Those figures would change the assessment.

Readers who want context on how such incidents are logged can see the [incident record](/record) and how [AI risk is measured](/method).

### What would settle it
Three documents would settle most of this: Microsoft's full technical write-up on Storm-0940 and ARTEX, ConnectWise's own incident notice with indicators for customers, and any later statement giving the number of compromised customers and whether data was taken. A statement from ConnectWise accepting or contesting the attribution would also help.

## What can be done
For customers of ConnectWise cloud services: review the list of applications and service principals in your Microsoft 365 tenant, look for credential or permission changes made through Microsoft Graph, and review sign-in logs from early October onward. Revoke and reissue credentials rather than only disabling the connection. Ask ConnectWise for indicators and confirmation of when restored services are safe to reconnect. Multi-factor authentication matters for user accounts, though it would not have stopped a hijacked application identity. TechCrunch notes that weak security and missing multi-factor authentication marked the South Korean financial incident it links to the same tool.

For vendors: limit the permissions that application identities hold and rotate them. For readers wanting a wider plan for service failures, see the [Day Zero plan](/day-zero).

## Questions readers ask
### Was our company hacked in the ConnectWise breach?
The sources do not say. Microsoft says attackers attempted sign-ins against more than 2,250 Microsoft 365 tenants, but it has not disclosed how many customers were compromised. Only customers of the affected ConnectWise services need check first.

### Who did it?
Microsoft says a China-linked espionage group it tracks as Storm-0940 did it. The Register says ConnectWise did not initially identify who was responsible. The reports show no independent confirmation.

### What is ARTEX?
The Register describes ARTEX as an open source penetration-testing framework in Go that uses AI agents to automate attack steps. It says it is freely available and that Microsoft has seen attackers from several regions using it.

### Is it safe to turn the ConnectWise services back on?
The sources do not answer that. ConnectWise says it contained the affected connections and is restoring services gradually. Its CEO said shutting the connections would not necessarily remove attackers already inside customer environments, so customers should check their own tenants.

## Who is affected

- Organisations using ConnectWise cloud backup and security services with Microsoft 365 (relative exposure 10 of 10)
- IT teams and managed service providers that hold customer-linked application credentials (relative exposure 8 of 10)
- Government agencies, defense contractors, banks and manufacturers named as targets (relative exposure 6 of 10)
- Microsoft 365 administrators reviewing app permissions and sign-in logs (relative exposure 7 of 10)
- Security vendors and defenders facing AI-automated attack tools (relative exposure 4 of 10)

## Risks

- Attackers who gained persistent privileged access may remain in customer tenants even after ConnectWise connections were shut down.
- Customers may assume they are safe because the number of compromised organisations has not been disclosed.
- Stolen vendor-held application credentials can bypass user-level protections such as password policies.
- Freely available AI penetration tools lower the effort for other attackers to copy the method.

## What can be done

- Audit Microsoft 365 applications, credentials and privilege changes made through Microsoft Graph, and revoke and reissue them.
- Review sign-in logs from early October onward for unfamiliar application activity.
- Ask ConnectWise for indicators of compromise and its restoration status before reconnecting services.
- Limit and rotate permissions granted to third-party application identities.

## Sources

1. Chinese hackers spied on thousands of companies after breaching IT giant ConnectWise, TechCrunch, 2026-10-22: https://techcrunch.com/2026/10/22/chinese-hackers-spied-on-thousands-of-companies-after-breaching-it-giant-connectwise/
2. Microsoft says ConnectWise breach let Chinese spies loose on 2,250 cloud tenants, The Register, 2026-10-22: https://www.theregister.com/security/2026/10/22/microsoft-says-connectwise-breach-let-chinese-spies-loose-on-2250-cloud-tenants/5304687

---
Source: https://www.toswo.com/signals/connectwise-breach-and-ai-agents-hit-microsoft-365-tenants
Prepared by Toswo Research (https://www.toswo.com/research) from the sources above. Free to cite with attribution: https://www.toswo.com/cite
