Storm-0940 Uses ARTEX AI Agent to Probe IT and Software Vendors for Supply Chain Access
Do Son 
| At a glance | |
| Actor | Suspected Storm-0940, which Microsoft describes as Chinese state-aligned |
| Activity | AI agent-driven reconnaissance, password spraying, and authentication bypass attempts |
| Targets | IT and software vendors, their identity providers, and their customer data |
| Scale | Multiple vendors; hundreds of user accounts sprayed, per Wiz |
| Status | No law-enforcement action reported |
| Source | Wiz Research, with reference to Microsoft and Zscaler findings |
TL;DR
Throughout October 2026, Wiz tracked an actor using the ARTEX AI agent to attack IT and software vendors. The activity overlaps with Microsoft’s reporting on Storm-0940. The agent hit Okta, Entra ID, and AWS Cognito logins in search of a path to vendors’ customers.
What happened
ARTEX is an open-source autonomous penetration testing tool. A Chinese developer released it on GitHub in July 2026. It then won Baidu’s “Agent+” red team challenge in September. An operator gives it a target and a goal. A planner agent then chooses each step, while worker agents run tools such as nmap and Impacket.
Attackers adopted it fast. ARTEX artifacts appeared in attacks on South Korean banks in late September and Japanese firms in October. The developer closed the source on October 8, but by then the code had more than 12,000 forks.
A repeatable playbook
Wiz saw the same steps across many victims. First, the agent downloaded JavaScript bundles, source maps, and API docs. Minutes later, it called internal API routes that appear only in those source maps.
Next, it hammered password-reset and account-recovery pages. Where sign-up was open, it registered with a throwaway email. It then “immediately tried to look at someone else’s data.” It also tested forged, unsigned login tokens against more than 60 SSO routes in two minutes.
Password spraying ran in tight bursts against Okta and Entra ID. Each burst used a new user agent and paced attempts one second apart. The agent mostly targeted software engineers’ accounts.
Who is behind it
Attribution remains suspected. Wiz says its observations overlap with Microsoft’s recent report on Storm-0940. Microsoft calls that group Chinese state-aligned and links it to a supply-chain attack using stolen vendor credentials. However, Wiz refers to the actor it tracked as an “ARTEX-using actor.” Microsoft and Zscaler have also tied ARTEX to two separate actors.
Impact and scale
Most attempts failed, according to Wiz. Still, the agent never stopped. Some sessions ran 24 to 30 hours. When defenders blocked an IP, a new one appeared within minutes. When a page rate-limited it, the agent moved to another endpoint.
Wiz warns that “an autonomous agent does not give up. It simply continues to iterate until it finds a single weak point.” For vendors, one weak point could expose every downstream customer. That matches the supply chain approach Microsoft has tied to Storm-0940.
How to stay protected
Defenders should consider these steps:
- Remove public source maps and lock down API documentation in production.
- Rate-limit password-reset and account-recovery endpoints, and alert on bursts.
- Reject unsigned tokens and validate OAuth state parameters.
- Block disposable email domains at sign-up where possible.
- Test tenant isolation so one account cannot reach another tenant’s data.
- Watch Okta and Entra ID for one-second, steady-rate login attempts.
In short, the ARTEX AI agent turns old techniques into a tireless campaign. Vendors that guard customer access should treat these patterns as a top priority.
Support Our Threat Intelligence
Find our threat intelligence and malware analysis helpful? Support our work today and unlock a 100% ad-free reading experience!
Buy Me a Coffee
