Protect your company against ARTEX:the basics, done faster and everywhere.
ARTEX does not use new tricks. It runs known attack steps at machine speed, against many systems at once. This guide turns what the FBI, Microsoft, Google, CrowdStrike and JPCERT/CC recommend into one ordered checklist for a security team.
Checked
What changes with AI agent attacks
Speed and scale. The FBI says some ARTEX intrusions reached privileged access in under three hours, and that small groups can now run many campaigns at once. Microsoft saw several Storm-0940 cases under three hours from first targeting to privileged access. ARTEX traffic is noisy, but the FBI warns that the speed leaves little time to respond.
The weaknesses are old ones. TechCrunch notes that many attacks linked to ARTEX used guessed weak or known passwords and missing multi-factor authentication. Google's threat intelligence group lists long-standing practices as the answer: control-plane isolation, short-lived credentials, least privilege, segmentation and behavioral detection.
The checklist, in priority order
Work from the top. Each step names the source behind it, so you can read the full advice.
- 01
List everything that faces the internet
Include side systems, not only the main site: partner and supplier portals, mobile backends, employee apps, sales-support systems, legacy services, third-party integrations and apps other business units built (FBI). Require a review before anything new goes public.
- 02
Patch internet-facing systems fast
Apply fixes for known vulnerabilities (JPCERT/CC), with internet-facing appliances and long-lived Linux servers first (CrowdStrike). If a system cannot be patched, take it offline or put it behind access control.
- 03
Turn on MFA and clean up credentials
Require multi-factor sign-in, remove default and weak passwords, and use short-lived credentials (Google GTIG). API tokens should have least privilege, an expiry and quick revocation (JPCERT/CC). Scan code, images and pipelines for leaked secrets and rotate any you find (FBI, StepSecurity).
- 04
Take admin panels off the internet
Put cloud admin, network device management, publishing interfaces, hypervisors and backup platforms behind trusted networks, private links or Zero Trust access (FBI).
- 05
Add rate limits and volume alerts
Rate limit APIs, and set separate limits on login, password reset, SMS sending and search (JPCERT/CC). On portals, alert on sequential lookups and on too many records per session, and check authorization on the server for every record (FBI).
- 06
Show less to strangers
Do not return stack traces, SQL errors, internal paths or framework details to users who are not signed in. Keep them in internal logs (FBI).
- 07
Control outbound traffic
Servers should only reach the destinations they need. The FBI lists unexpected calls to LLM APIs and unknown LLM proxies in egress logs as signs to investigate, which is easier when outbound traffic is limited. CrowdStrike also advises requiring IMDSv2 on cloud instances.
- 08
Log and watch every server
Run endpoint detection on every server, including ones you think are unreachable. In one case the operator told the agents to look for hosts without a sensor (CrowdStrike). Keep identity and audit logs long enough to look back, and review your first response plan (JPCERT/CC).
- 09
Limit vendor, MSP and app access
Give multi-tenant apps and consent grants least privilege, and watch for new credentials, privilege changes and unusual app-only activity (Microsoft, FBI). See the ConnectWise case below.
- 10
Segment the network and protect backups
Segment so one broken server does not open the rest; DIVD credits segmentation with stopping its attacker from going deeper. Keep backup platforms on trusted networks only (FBI), and keep at least one copy that everyday admin accounts cannot change or delete.
Start with the side systems
In the South Korean bank campaign, CrowdStrike describes the entry points as a broker-facing loan inquiry service at one bank and an employee mobile support system at another, not core banking. The FBI says ARTEX is good at exactly these overlooked systems.
JPCERT/CC's alert on AI pentest platforms, ARTEX included, describes attempts to find and use vulnerabilities in internet-facing web apps, attempts to steal configuration and backup files, and attempts against internet-exposed WSUS servers. Those are good first places to check.
Vendors and MSPs: the ConnectWise case
Microsoft says Storm-0940 used ARTEX to speed up the compromise of ConnectWise's Cloud Backup app, then signed in to customer tenants through apps those customers had already approved. Sign-ins were attempted at more than 2,250 tenants. Attempts are not confirmed breaches.
- Keep a list of every vendor app with access to your tenant, including ones no longer in use. Search by application ID, not display name (Microsoft, FBI).
- Review audit records for new credentials on apps, new service principals or owners, consent grants and privileged role changes (Microsoft).
- Give vendor apps only the permissions they need, and remove what they no longer use.
- Ask your MSP and key vendors how they protect their own admin access and how fast they will tell you about an incident.
Test yourself before someone else does
The same kind of tool can find your gaps first, if you run it with written permission, on your own systems, with its safety controls turned on. The guide to authorized AI pentesting on this site covers how. If you think you were already hit, go to the incident response guide.
Questions
- How do I protect my company from ARTEX?
- Know every system you expose, patch fast, require MFA, keep admin panels off the internet, add rate limits, limit outbound traffic, log everything and review vendor access. The checklist above puts these in order.
- Can I block ARTEX with a firewall rule?
- Not reliably. ARTEX runs from many servers and proxies, and the FBI says IP indicators can go stale. Blocking known indicators helps a little; fixing exposure and watching behavior helps more.
- Does MFA stop AI agent attacks?
- Not all of them, but it closes one of the most common doors. TechCrunch notes many ARTEX-linked attacks used weak or known passwords and missing MFA.
- Is there a patch for ARTEX?
- No. ARTEX is an attack tool, not a bug in your systems. The fix is to close the weaknesses it looks for: unpatched services, weak credentials, open admin panels and loose API limits.
- Should small companies worry about ARTEX?
- Yes. The FBI says ARTEX lowers the skill needed for complex attacks, and Check Point describes mostly opportunistic, money-driven use. Small companies with exposed systems are easy targets.
Sources
Every claim on this page comes from these sources or from this site's own sourced pages. Open them for the full reports.
- Threat Overview of ARTEX Autonomous Penetration Testing Framework (FLASH-20261113-001)
FBI (IC3 FLASH, TLP:CLEAR), 2026-11-13
- Storm-0940 uses ARTEX agentic AI in ConnectWise supply chain compromise
Microsoft Threat Intelligence, 2026-10-21
- ARTEX and Friends: Enter Agentic Offensive Frameworks
Google Threat Intelligence Group + Mandiant, 2026-10-30
- VAPOR PANDA Uses AI-Driven ARTEX to Grow Global Espionage Campaign
CrowdStrike OverWatch, 2026-10-29
- JPCERT-AT-2026-0030 (updated 2026-10-23, case E)
JPCERT/CC, 2026-10-23
- Hundreds of Artex AI hacking tools found publicly exposed to the web
TechCrunch, 2026-11-06
- ARTEX: Attackers Adopted This AI Pentesting Tool in Three Months. Leaked Secrets Are Its Fastest Way In.
StepSecurity, 2026-11-12
Read next
Follow the ARTEX story.
New forks, releases and reports, dated and sourced.
