The sharpest story in cybersecurity news today is about speed and trust, not a new flaw in any one product. On October 21, 2026, Microsoft Threat Intelligence reported that Storm-0940, a China state-aligned actor it has tracked since 2021, used ARTEX, an open-source agentic AI penetration-testing framework, to automate attacks on hundreds of organizations. The framework sprayed passwords at a peak of 2,900 sign-in attempts per minute. Where spraying failed, the actor signed in through OAuth applications that already held granted access, a route that never asks a user for a password. Several intrusions reached privileged cloud access in under three hours. This page states what Microsoft confirmed, then gives you a working check: your own sign-in numbers, rated against the incident's own benchmarks, plus a review of the application identities an attacker would rather use than yours.
Put the spray in proportion. A person mistyping a password makes a handful of failed attempts against one account. A spray makes hundreds of attempts per account across hundreds of accounts, and nearly all of them fail. The signal is not any single failure; it is the shape: high volume, wide spread, near-total failure. The checker below measures exactly that shape in your numbers.
The identity you granted is the target. Password spraying is an old technique. What ARTEX changed is the labor: an agentic framework runs the scanning, the spraying, and the follow-on steps without a person at each keyboard, so the campaign's pace is set by infrastructure, not by staffing. Microsoft's figures show what that pace does to defenders. At 2,900 attempts per minute, a detection rule that waits for a human to review a daily report is a day late.
The second half of the campaign is the part worth sitting with. An OAuth application with granted access is an identity your organization created and trusted. It has no password to spray and no user to phish. When Storm-0940 authenticated as an application, the security controls aimed at people (password policy, sign-in risk, one-time codes) had nothing to evaluate. The controls that would have caught it are the ones Microsoft names in its guidance: watching service principal sign-ins and auditing unexpected application credential or privilege changes. If nobody in your organization can list your service principals and what each one may do, that list is the first thing to build.
Take a window from your identity provider's sign-in logs and enter what you find. Everything runs in your browser; nothing is sent anywhere. The verdict shows its math and names the benchmark it used.
Storm-0940's quiet route in was an application identity, not a person. Tick every statement that is true in your tenant today. The readout updates as you tick and names what you ticked.
Agent Auth Gateway, Ingenious AI's gateway for application and agent identities, exists for the second half of this campaign: it puts identities behind allowlists, budgets, and a kill switch, so a trusted identity that starts behaving like Storm-0940's tooling gets stopped instead of inherited. See Agent Auth Gateway.
One vendor, one security posture. Every platform below is live now on getingeniousai.com.
Storm-0940 is a China state-aligned hacking group Microsoft has tracked since 2021. In a report published October 21, 2026, Microsoft Threat Intelligence said the group used ARTEX, an open-source AI penetration-testing framework, to automate attacks on hundreds of organizations. The framework ran reconnaissance and password spraying at machine speed, then signed in through OAuth applications that already held trusted access, so no person had to type a password. Several intrusions reached privileged cloud access in under three hours.
Password spraying tries a small set of common passwords against many accounts at once, staying under per-account lockout limits. ARTEX removed the human speed limit: Microsoft measured a peak of 2,900 sign-in attempts per minute, more than 2,250 customer tenants reached, and over October 10 to 14 more than 151,000 failed authentication attempts against 327 accounts in 8 organizations, about 462 attempts per account.
Many cloud apps hold their own identities, called service principals, with permissions granted once and then trusted. Storm-0940 authenticated as those applications, so the sign-in never involved a user, a password, or a one-time code. Once inside as a trusted application, the group ran reconnaissance, changed application credentials, and tried to modify privileges to widen and keep access. Microsoft's guidance is to investigate service principal sign-ins and unexpected application credential or privilege changes.
Microsoft attributes the ConnectWise Cloud Backup supply-chain compromise to Storm-0940. ConnectWise Cloud Backup is a multitenant application, so one compromised application identity sits above many customer tenants at once. The campaign shows why a vendor application with broad tenant access is a beachhead: control the application identity and you inherit its reach.
Sources: Microsoft Threat Intelligence report, October 21, 2026 (the Storm-0940 attribution, the ARTEX framework, the 2,900 attempts-per-minute peak, the 2,250+ tenants, the October 10 to 14 failed-attempt counts, the under-three-hours intrusions, the ConnectWise Cloud Backup attribution, the targeted sectors, and the defender guidance on service principal sign-ins and application credential or privilege changes), as covered by Newsmax on October 25, 2026 and cross-checked against NHK, Business Standard, and Silicon UK reporting of the same Microsoft report.