ThreatBeaconXSubscribe
CriticalThreat Intelligence · 9 min read

Storm-0940 Uses Agentic AI to Accelerate ConnectWise Supply-Chain Attack

Microsoft Threat Intelligence reported that Storm-0940, a China-state-aligned threat actor, used the open-source agentic AI penetration-testing framework ARTEX to automate attacks against organizations worldwide. Microsoft assesses that Storm-0940 was responsible for the ConnectWise Cloud Backup supply-chain compromise disclosed in October 2026. The attackers abused compromised application credentials associated with ConnectWise to authenticate to downstream Microsoft 365 customer environments. ARTEX accelerated reconnaissance, password spraying, and OAuth application abuse, enabling attacks to proceed at machine speed. Microsoft reported that some AI-assisted operations reached privileged cloud access in less than three hours. Follow-on activity included Microsoft Graph reconnaissance, application credential modifications, and attempts to change privileges to establish persistent access. Microsoft emphasized that successful authentication or the presence of an affected application does not automatically prove that customer data was accessed or exfiltrated. The incident demonstrates how agentic AI can accelerate existing attack techniques and increase the scale of supply-chain compromises.

Written by Jakka Lakshmi Koumudi·Published Oct 22, 2026

1. Description

Storm-0940 is a China-state-aligned threat actor active since 2021. Microsoft reports that the group has historically targeted organizations across multiple industries for data theft, using techniques such as password spraying, exploitation of internet-facing applications, and persistence through cloud application credentials.

In September 2026, Microsoft observed Storm-0940 using ARTEX to automate attacks against hundreds of organizations globally. The campaign included organizations in aerospace and defense, finance, government, manufacturing, technology, energy, software, and other sectors.

Microsoft connected the actor's activity to the ConnectWise supply-chain compromise. By abusing trusted application identities that already had access to customer Microsoft 365 tenants, the attackers could operate without requiring interactive sign-ins from individual users.

The incident illustrates the risk of highly privileged third-party applications: compromising one trusted integration can create opportunities to access multiple downstream customer environments.

2. Technical Description

ARTEX is an open-source, multi-agent penetration-testing framework that uses large language models to automate offensive security workflows. Written in Go with a Next.js frontend, it can coordinate activities from reconnaissance through exploitation and potential data exfiltration with limited human involvement.

Microsoft reported that Storm-0940 used ARTEX to accelerate several attack stages:

  • Automated reconnaissance: Identifying targets and available cloud resources.

  • Password spraying: Attempting common or selected passwords across multiple accounts.

  • OAuth application abuse: Using compromised application credentials to access connected Microsoft 365 environments.

  • Microsoft Graph operations: Enumerating cloud resources and investigating available permissions.

  • Credential modification: Adding or modifying application credentials to create further access paths.

  • Privilege modification: Attempting to assign elevated roles and permissions to service principals.

ARTEX is the automation framework, not a malware payload. The documented activity primarily involved cloud identities, authentication, application permissions, and Microsoft Graph rather than a conventional executable-based infection.

Microsoft stated that the activity did not result from a compromise of or vulnerability in Microsoft's own corporate systems, products, or services.

3. Attack Overview

The combined attack sequence described by Microsoft is as follows:

  1. Storm-0940 uses ARTEX to automate reconnaissance and select targets.

  2. The actor performs password spraying against Microsoft 365 accounts and uses compromised application credentials where available.

  3. The ConnectWise supply-chain compromise provides a route to authenticate to downstream customer tenants through trusted application identities.

  4. Automated agents conduct Microsoft Graph reconnaissance to identify resources, permissions, and potential follow-on opportunities.

  5. In some affected environments, the attackers attempt to modify application credentials and privileged roles.

  6. These changes can establish additional access paths that may persist even after the original compromised application is disabled.

  7. Defenders must investigate service-principal sign-ins, directory changes, Graph activity, and resource access to determine the actual scope of compromise.

Microsoft observed some attacks achieving privileged cloud access in less than three hours. However, the majority of the broader targeting activity was unsuccessful, and the presence of an affected application does not by itself establish compromise.

4. Technical Analysis

4.1 ARTEX-assisted attack automation

ARTEX coordinates multiple AI-driven agents to perform offensive security tasks. Its workflow can combine reconnaissance, asset discovery, attack-path evaluation, and subsequent operations without requiring an operator to manually initiate every step.

Microsoft reported that Storm-0940 used this automation to increase the speed and scale of its operations. AI-driven agents can continue working outside normal business hours and attempt multiple attack paths concurrently, reducing the time defenders have to investigate alerts and revoke access.

The security concern is not simply the use of AI, but the acceleration of existing attack techniques through automated decision-making and execution.

4.2 Large-scale password spraying

Between October 10 and October 14, 2026, Microsoft observed Storm-0940 infrastructure at 144.172.64[.]99 conducting a cross-tenant password-spraying campaign against Microsoft 365 and Exchange Online.

Microsoft reported the following figures:

  • More than 151,000 failed logon events.

  • 327 accounts targeted across eight organizations.

  • Peak activity of approximately 2,900 failed attempts per minute.

  • More than 111,000 failures concentrated into one day.

  • Forty-eight accounts receiving more than 1,000 failed attempts each.

  • Nearly 98% of recorded activity represented lockout responses.

Despite the scale, Microsoft reported no successful authentication, token issuance, Graph API calls, or directory changes during this particular password-spraying activity.

This distinction is important: the earlier password-spraying campaign and the subsequent ConnectWise application-credential abuse were related in Microsoft's investigation, but they were not the same authentication event.

4.3 ConnectWise application credential abuse

After reports of the ConnectWise incident emerged on October 15, Microsoft observed overlapping network indicators and subsequent authentication activity using compromised ConnectWise application credentials.

The attackers exploited the trust already established between the multi-tenant applications and downstream customer environments. This enabled access attempts through application identities without requiring individual users to sign in interactively.

The activity demonstrates how compromise of a shared service-provider application can expand the potential impact beyond the original vendor environment.

4.4 Microsoft Graph reconnaissance

In multiple affected environments, Microsoft observed follow-on operations through Microsoft Graph. Depending on the permissions granted to the application, Graph can provide access to directory information, mail, files, and other Microsoft 365 resources.

The reported activity included reconnaissance and changes involving application credentials, identities, groups, and privileged roles.

Defenders should correlate service-principal authentication with Graph request records, source IP addresses, requested scopes, response codes, and subsequent directory changes. Authentication alone does not prove that sensitive resources were accessed.

4.5 Credential changes and cloud persistence

Microsoft reported attempts to add or modify application credentials and alter privileged roles. Such changes can provide attackers with alternative authentication paths, potentially allowing access to continue after the original compromised application has been disabled.

For this reason, containment must include reviewing application secrets, certificates, role assignments, consent grants, application owners, and other identity changes made during the suspected compromise window.

4.6 Detection and investigation

Microsoft recommends using Microsoft Defender XDR, Microsoft Defender for Cloud Apps, Microsoft Sentinel, and Microsoft Entra logs to investigate the activity.

Relevant detection opportunities include:

  • Unusual Microsoft Graph activity associated with the compromised ConnectWise application.

  • An application assigning itself a Global Administrator role.

  • Suspicious service-principal sign-ins from known threat infrastructure.

  • Unexpected application credential or permission changes.

  • Newly added OAuth applications accessing mailbox or Teams data.

  • Unusual SharePoint enumeration or other high-volume resource access.

A match against an application ID or IP address should initiate an investigation, not be treated as conclusive evidence of compromise. The findings must be correlated with the tenant's actual sign-ins, permissions, and resource-access records.

5. Indicators of Compromise (IOCs)

Microsoft published the following indicators for investigating the ConnectWise-related activity. IP addresses are defanged where appropriate.

IOC TypeIOCAssociated Activity / Purpose
IP Address144.172.64[.]128Sign-ins and Microsoft Graph activity involving ConnectWise applications
IP Address144.172.64[.]99Earlier password-spraying activity and related ConnectWise investigation
IP Address144.172.64[.]60Sign-ins and Graph API activity associated with affected applications
IP Address187.15.88[.]186Graph API access through compromised applications; VPN address requiring contextual investigation
IP Address187.15.90[.]217Graph API access through compromised applications; VPN address requiring contextual investigation
IP Address187.15.89[.]186Graph API access through compromised applications; VPN address requiring contextual investigation
User AgentPython-urllib/3.12Sign-ins and Graph API activity using affected ConnectWise applications
User Agentpython-requests/2.31.0Observed client identifier in the incident
User Agentpython-requests/2.34.2Observed client identifier in the incident
User Agentcurl/7.88.1Observed client identifier in the incident
User Agentazure-toolkit/1.0Observed client identifier in the incident

Associated application IDs

IOC TypeApplication IDAssociated Application
Application Client ID506e08f9-6e9a-4da4-942b-a099a03644aeConnectWise Cloud Backup
Application Client ID24205314-2b89-4e07-a2a4-579e898a5640ConnectWise – MS Teams
Application Client ID0de14c77-b9b0-4137-9360-34878a929827ConnectWise – MS Graph
Application Client ID2399d300-6fca-4e6a-a686-399fefe75b83ConnectWise Cloud Management

IOC handling notes:

  • Microsoft notes that the three 187.15.x.x addresses were VPN IPs. Their relevance is time-sensitive, and they should be investigated in the context of the affected applications and activity on October 15–16, 2026.

  • User-agent strings are supporting indicators, not proof of malicious activity.

  • Application IDs identify affected integrations; their presence alone does not prove that an application was abused.

  • The Microsoft report does not provide a confirmed malware file hash for this activity.

  • Search for the published indicators, but also investigate unfamiliar IPs and application behavior beyond this list.

6. Potential Impact

  • Cross-tenant compromise: Abuse of a privileged multi-tenant application can create exposure across numerous customer environments.

  • Cloud privilege escalation: Unauthorized changes to application roles may provide elevated access to Microsoft 365 resources.

  • Persistence: Additional application credentials can preserve access even after the original integration is disabled.

  • Data exposure: Email, SharePoint, Teams, and other cloud resources may be at risk depending on permissions and observed activity.

  • Rapid attack execution: AI-assisted automation can shorten the time between initial targeting and privileged access.

  • Operational disruption: Organizations may need to disable integrations, review service-principal permissions, rotate credentials, and validate affected workloads.

  • Increased defensive workload: High-volume authentication and automated reconnaissance can generate large volumes of alerts and logs in a short period.

Microsoft explicitly cautioned that successful authentication does not necessarily mean that customer data was accessed or exfiltrated. Impact must be determined from tenant-specific evidence.

7. Recommendations

Immediate Actions

  1. Keep affected ConnectWise application connections disabled until ConnectWise provides formal restoration instructions.

  2. Inventory the affected application IDs in every Microsoft 365 tenant where the applications are currently or were previously installed or authorized.

  3. Preserve evidence from Microsoft Entra service-principal sign-in logs, directory audit logs, Microsoft Purview Audit, Microsoft Graph activity, and SIEM records.

  4. Review activity from at least October 15, 2026, and extend service-principal sign-in investigations to October 1 where logs are available.

  5. Search the listed IP addresses and user agents, while also investigating unexpected source infrastructure beyond the published indicators.

  6. Inspect application credentials and permissions for newly added secrets, certificates, owners, consent grants, role assignments, and other changes.

  7. Investigate Microsoft Graph activity for unusual directory enumeration, permission discovery, mail access, file downloads, and changes to identities or groups.

  8. Check for persistence by reviewing service-principal credentials and privileged roles that may have been modified during the incident.

  9. Remove confirmed unauthorized changes after preserving the relevant evidence and coordinating with an authorized incident responder.

  10. Assess backup and restoration workflows before re-enabling affected services.

Preventive Actions

  1. Apply least-privilege permissions to enterprise applications and service principals, especially multi-tenant integrations.

  2. Monitor application credential additions, consent grants, privileged role assignments, and unusual app-only authentication.

  3. Use alerts for anomalous Microsoft Graph requests and unexpected access to mailboxes, SharePoint, and Teams.

  4. Establish clear processes for emergency revocation of third-party application permissions and credentials.

  5. Maintain centralized retention of Entra, Microsoft 365, and Graph telemetry to support investigations.

  6. Review vendor access regularly and remove unused integrations and unnecessary application permissions.

  7. Use Microsoft Defender XDR, Defender for Cloud Apps, and Microsoft Sentinel detections where available.

  8. Ensure that AI-enabled security governance includes visibility into automated agents and their access to sensitive enterprise resources.

8. Conclusion

The Storm-0940 campaign illustrates how agentic AI can accelerate established attack techniques and increase the scale of supply-chain compromises. By using ARTEX to automate reconnaissance, password spraying, and OAuth application abuse, the actor was able to attempt access across many organizations and, in some cases, achieve privileged cloud access rapidly.

The ConnectWise incident highlights the risk posed by highly privileged third-party applications. Defenders must investigate not only the initial authentication but also subsequent Graph activity, credential changes, role assignments, and potential persistence mechanisms.

Microsoft attributes the ConnectWise supply-chain compromise to Storm-0940, a China-state-aligned actor. However, successful authentication alone does not prove data exfiltration, and each organization must establish its actual level of exposure from its own telemetry.

MITRE ATT&CK Mapping

T1580 — Cloud Infrastructure Discovery