Storm-0940 & ARTEX: China’s AI Agents Breach 2,250+ Microsoft 365 Tenants via ConnectWise Supply Chain

On 21 October 2026, Microsoft’s threat-intelligence team released what may be the most consequential supply-chain attack disclosure of the year — and it comes with a twist that should alarm every Indian IT leader: the attackers didn’t just use a stolen credential. They deployed an AI agent to do the heavy lifting. Welcome to the era of autonomous, nation-state cyberattacks powered by open-source artificial intelligence.

The Chinese state-aligned threat actor known as Storm-0940 used an agentic AI framework called ARTEX to compromise credentials from ConnectWise Cloud Backup (formerly SkyKick) and then attempted unauthorized sign-ins across more than 2,250 Microsoft 365 tenants — at a rate of 2,900 login attempts per minute. In some cases, attackers gained privileged cloud access in under three hours from initial targeting. If your organization is among the tens of thousands of Indian enterprises that rely on Microsoft 365 — or works with a Managed Service Provider who uses ConnectWise — this affects you.

Storm-0940: Who Is Behind the Attack?

Storm-0940 is a China state-aligned threat actor previously tracked by Microsoft under overlapping clusters and linked by other researchers to the broader Silk Typhoon activity set. The group has historically focused on credential theft, cloud-tenant compromise, and espionage targeting government, defence, and critical infrastructure organizations in North America, Europe, and the Asia-Pacific region — including India.

Before the ConnectWise phase, Storm-0940 conducted a targeted password-spraying precursor campaign from late September to 14 October 2026, attempting to brute-force 327 Microsoft 365 and Exchange Online accounts across eight separate organizations. That campaign was largely unsuccessful — but it seeded the intelligence the group needed to pivot to a far more effective vector: a trusted SaaS tool already inside the network perimeter.

ARTEX: The AI Agent That Did the Dirty Work

ARTEX (Agentic Red-Teaming and Exploitation) is an open-source agentic AI pentesting framework that chains together multiple AI-driven sub-tasks: reconnaissance, credential testing, OAuth application abuse, and lateral movement. What makes ARTEX significant — and dangerous — is that it operates with minimal human intervention. Once pointed at a target, the framework can autonomously:

• Enumerate Microsoft 365 tenant metadata via the Graph API
• Spray stolen or leaked credentials against authentication endpoints
• Abuse OAuth delegated permissions to persist without an interactive login
• Escalate privileges by modifying directory roles
• Establish persistent backdoors through consented application registrations

The speed is staggering. Microsoft observed ARTEX submitting 2,900 sign-in attempts per minute — a rate that would take a human red team hours to achieve manually. In documented intrusions, the actor went from initial targeting to privileged cloud access in fewer than 180 minutes. This isn’t a future threat. It is operational reality, right now.

Phase	ARTEX Action	Observed Impact
Precursor (Sep–Oct 14)	Password spraying — 327 M365/EXO accounts	Intelligence gathering; limited breach
Supply chain pivot (Oct 14–16)	Compromise of ConnectWise Cloud Backup credentials	Access to downstream tenant sign-in without MFA prompt
Automated intrusion (Oct 14–16)	ARTEX — 2,900 sign-in/min across 2,250+ tenants	Privileged access achieved in <3 hours in some cases
Post-compromise	Graph recon, credential modification, role elevation	Persistent backdoors via consented app registrations

Why Supply Chain + AI = Dangerous Multiplier

The genius — and menace — of this attack lies in combining two vectors that independently are well-understood, but together create an exponential risk multiplier.

Supply chain compromise through a trusted SaaS tool like ConnectWise bypasses traditional perimeter controls because the malicious sign-ins look legitimate — they come from a known application with delegated permissions rather than an unfamiliar IP address. IT teams and SIEM rules trained to flag anomalous logins often miss these because the application identity is trusted.

Layer in ARTEX’s AI-driven automation, and you eliminate the time that defenders rely on: that window between the first anomalous event and human escalation. When an attacker can probe, pivot, and persist across thousands of tenants in under three hours, your mean time to detect (MTTD) becomes your Achilles heel.

This also has implications for Indian Managed Service Providers. India is among the world’s fastest-growing Microsoft 365 markets, and ConnectWise-class tools are standard in mid-market MSP stacks serving BFSI, manufacturing, and government clients. A single compromised MSP credential can give an adversary simultaneous access to dozens of client environments — exactly what Storm-0940 exploited.

For a deeper look at how a prior supply-chain attack on an AI platform unfolded, see our earlier analysis: LiteLLM 1.107.0 Supply Chain Breach: AI Credential Stealer Hit 2 Hours After Deployment.

What You Should Do Right Now — Sanjay Seth’s Defence Playbook

This attack is containable with the zero-trust controls that should already be in your roadmap. Here is a prioritized action list:

1. Audit Entra ID sign-in and audit logs immediately. Go back to at least 15 October 2026. Look for:
   • Sign-ins from ConnectWise, SkyKick, or any cloud-backup application
   • Newly consented OAuth applications, especially those with Directory.ReadWrite.All or RoleManagement.ReadWrite.Directory permissions
   • Unexpected Global Administrator or Privileged Role Administrator assignments
   • Service principal sign-ins from unusual IP ranges or geolocation
2. Revoke and rotate all ConnectWise Cloud Backup credentials. If you were a Cloud Backup, SaaS Security, or Cloud Management customer, treat all associated service accounts as compromised until proven otherwise.
3. Enforce Conditional Access on all service principals. Many organizations apply MFA only to human identities. Service accounts and app registrations must also be bound by Conditional Access policies — restrict sign-in by trusted IP ranges and require compliant devices.
4. Apply least-privilege to third-party app permissions. Run a quarterly review of all Microsoft Entra app registrations. Any application with directory-write or role-management permissions that cannot be justified should be removed.
5. Enable Microsoft Entra ID Protection alerts for impossible travel, anomalous token usage, and bulk sign-in events. Configure real-time alerts to your SOC or SIEM (FortiSIEM, Sentinel, or similar).
6. Notify CERT-In if you find indicators of compromise. Under the 2022 CERT-In directions, reportable incidents must be notified within 6 hours of discovery. Supply-chain breaches triggering unauthorized access to customer data qualify.

Embedding zero-trust network segmentation alongside cloud-identity controls ensures lateral movement is constrained even if a tenant-level breach occurs. Our case study on Zero-Trust Across a 15-Building Campus illustrates the principle at scale.

Frequently Asked Questions

Is ConnectWise Cloud Backup safe to use again?

ConnectWise suspended affected services on 16 October 2026 and has been conducting a forensic investigation. As of the time of writing, the services remain suspended or in a restricted state for affected customers. Do not reconnect unless you have received explicit written clearance from ConnectWise confirming remediation, and even then, rotate all credentials and re-review app permissions before restoring access.

Does ARTEX represent a new class of threat, or is AI just a speed boost?

Both — and the distinction matters. ARTEX primarily accelerates existing attack techniques (password spraying, OAuth abuse, Graph enumeration). But the speed multiplier is qualitatively different: it collapses the time defenders have to react from hours to minutes. Future ARTEX-class tools will likely add autonomous decision-making for target selection and evasion, making AI-augmented attacks a new threat category, not just a faster version of the old one.

Are Indian organizations specifically at risk?

Yes. India ranks among the top five global markets for Microsoft 365 adoption, and India’s rapidly growing MSP sector commonly uses ConnectWise-class platforms. Storm-0940 has previously targeted APAC organizations including those in the financial services and government sectors. Indian enterprises with operations in North America or Europe are doubly exposed through multi-tenant management relationships. The Telangana state government’s TG-iPASS portal was separately hit by ransomware on 22 October 2026 — a coincidence that underscores October 2026 as a period of heightened targeting of Indian digital infrastructure.

What are the MITRE ATT&CK techniques I should hunt for?

Key techniques associated with this campaign: T1078.004 (Valid Accounts: Cloud Accounts), T1110.003 (Password Spraying), T1098.003 (Account Manipulation: Additional Cloud Roles), T1550.001 (Use Alternate Authentication Material: Application Access Token), and T1069.003 (Permission Groups Discovery: Cloud Groups). Map these to your SIEM detection rules and ensure you have coverage before the next campaign begins.

Sources and Further Reading

• Microsoft Security Blog — Storm-0940 Uses ARTEX Agentic AI in ConnectWise Supply Chain Compromise (October 21, 2026)
• The Hacker News — China-Linked Hackers Use ARTEX AI Agents in ConnectWise Supply Chain Attack
• CyberScoop — Microsoft Pins ConnectWise Supply-Chain Attack to China-State Aligned Threat Group
• Brandefense — ConnectWise Supply Chain Attack Checks (IOCs and Entra Hunting Queries)

---