"ARTEX is a system of LLM-driven agents that run an autonomous penetration test from reconnaissance through data exfiltration with little needed human interaction," Microsoft Security Research wrote — a description that sits at the center of a sprawling supply‑chain intrusion that targeted ConnectWise customers' Microsoft 365 environments.
Storm-0940 and the rise of AI‑driven tooling
Microsoft researchers attribute the campaign that compromised a ConnectWise Cloud Backup service to a China‑state aligned threat actor they track as Storm‑0940. According to Microsoft, the group — active since 2021 — began using the agentic AI penetration testing framework ARTEX as early as September 2026 to automate attacks across hundreds of organizations and a broad range of industries, including aerospace and defense, automotive, chemicals, energy, financial services, government, legal, manufacturing, media, military, semiconductors, software, technology, think tanks and non‑profits.
Microsoft Threat Intelligence warned that multiple threat actors from various regions are adopting ARTEX to "strengthen capabilities and speed up the tempo of attacks." The company said ARTEX works with any large language model to automate offensive security workflows and accelerates scanning, password‑spray attacks, and abuse of OAuth applications in downstream victims.
Timeline: password‑spraying, an intense login surge, and the ConnectWise hit
Microsoft detailed a narrow window of activity that preceded the ConnectWise compromise. During a four‑day stretch ending Oct. 14, Storm‑0940 targeted 327 Microsoft 365 and Exchange Online accounts across eight organizations, generating more than 151,000 failed logon attempts and peaking at 2,900 failures per minute. The day after that burst of activity, the same malicious infrastructure and AI agents focused on ConnectWise, using ARTEX to accelerate reconnaissance and exploitation against ConnectWise Cloud Backup customers' Microsoft 365 tenants.
Microsoft said Storm‑0940's agents reached more than 2,250 tenants and attempted sign‑ins at a rate of up to 2,900 attempts per minute — a tempo that Microsoft characterized as leaving defenders little time to detect and respond.
This site is the portfolio.
OSINTSights runs on Cloudflare Workers, D1, R2, and Vectorize, with an AI pipeline on Hetzner ARM. Nubivance designed, built, and operates it. We do the same for clients.
See what we buildTechniques observed: OAuth abuse, Microsoft Graph, and credential changes
Researchers reported that ARTEX‑driven agents in some environments progressed past initial authentication to Microsoft Graph and directory operations. Observed follow‑on activity included reconnaissance, modifications of application credentials to establish persistence, and alterations involving identities, groups, and privileged roles.
Microsoft warned that a compromised third‑party application with high privileges could provide broader access than originally intended. While the company did not disclose a final count of compromised customers, it emphasized that successful authentication does not necessarily mean Storm‑0940 accessed or exfiltrated data. Nonetheless, Microsoft noted evidence in multiple customer environments that the threat group's AI agents could gain persistent access and establish backdoors by modifying credentials in victims' directories.
ConnectWise response and customer impact
ConnectWise temporarily shut down services Saturday following the attacks and began restoring services in phases on Wednesday. The company advised customers to keep connections disabled during the recovery and said it expects the process to take several days. Microsoft described the ConnectWise supply‑chain incident as "one of many" in a wider wave of AI‑driven cyberattacks; Igor Tsyganskiy, executive vice president at Microsoft, wrote on LinkedIn that "Since early October, a wave of AI‑driven cyberattacks has been ongoing."
What this means for technologists, affected enterprises, and threat actors
- Technologists and security teams: Microsoft’s account centers on highly automated, LLM‑driven agents that can scan large numbers of tenants and attempt authentication at machine pace. Teams will need to prioritize detection controls that flag rapid, high‑volume sign‑in failures and monitor for post‑authentication directory activity such as application‑credential changes and privilege modifications.
- Affected enterprises and procurement leaders: The incident underscores the risk that a high‑privilege third‑party application can serve as a pivot. ConnectWise customers were urged to keep connections disabled while services are restored, and organizations that rely on managed backup or third‑party integrations may reassess permissions and recovery plans during the ongoing restoration.
- Threat actors and tool maintainers: Microsoft noted that multiple actors are adopting ARTEX. The availability of an open‑source, LLM‑driven penetration testing framework that can be repurposed for offensive campaigns creates new incentives for both attack automation and for defenders to harden automation‑related controls.
Microsoft’s reporting leaves several concrete facts plainly visible: Storm‑0940 used ARTEX at scale beginning in September 2026, generated more than 151,000 failed logons in a compact campaign that peaked at 2,900 failures per minute, and leveraged the same infrastructure to target ConnectWise Cloud Backup customers after that burst of activity. ConnectWise paused services, began phased restoration days later, and advised customers to keep connections disabled while recovery continues.
The episode maps a clear pattern: powerful automation, rapid attack tempo, and the strategic value of third‑party privileges. How defenders adapt to detect and contain AI‑driven agentic tooling — and how quickly backup and integration providers can resume services without exposing new attack paths — will be the immediate tests following this breach.
