AI Risk Wire

Cases › security breach › LR-2026-0856

Storm-3358 use of the ARTEX agentic AI offensive framework against financial-sector targets

Also listed as: AI cyberattacks

Who did it
Outside attacker
AI agent used
ARTEX
Reported
7 Nov 2026
Severity
Moderate
Tools it had
Nmap, curl, wget, Python HTTP clients, Norma SDK WebFetch, headless Chromium through Playwright
  1. Storm-3358, a financially motivated threat actor, used the ARTEX agentic offensive framework to attack financial-sector organizations. It ran alongside conventional automated tools.
  2. Together with conventional tools it scanned internet-connected systems, exploited flaws such as anonymous file upload and SSRF, and installed open-source remote access trojans.
  3. Targets were financial organizations, mainly in Southeast Asia and some large ones globally. The actor used RATs for persistence and lateral movement.

Source: Microsoft Security · 1 more

What was done

Microsoft is providing indicators of compromise so defenders can detect related activity.

Risks AI used to run a cyber attack (97 incidents) · AI system attacked or breached (54 incidents)

No source states: AI company and model, money lost or records, company using the AI, sector.

Evidence

Sources (2)

  1. ARTEX abuse expands as cybercriminals target financial sector · Microsoft Security · 2026-11-07
  2. Free AI hacking is targeting banks in Southeast Asia. · Vietnam.vn · 2026-11-09
Show the full account: 34 items from the primary sources

What happened, in order

  1. October 2026 In October 2026, Storm-3358 started using the ARTEX agentic AI offensive framework to target financial-sector organizations.
  2. 2026-10-19 The ARTEX server at 178.16.54.129 was first seen earlier, on 2026-10-19, and was still active as of 2026-11-06.
  3. 2026-10-28 The ARTEX server at 31.77.229.35 was first seen on 2026-10-28 and was active as of 2026-11-06.
  4. 2026-11-02 The WISP C2 nodes at 199.19.95.66, .67 and .68 were first seen on 2026-11-02 and were still active as of 2026-11-06.
  5. 2026-11-04 The ARTEX servers at 91.92.240.238 and 85.11.168.45, plus the C2 server at 85.11.167.52, were first seen on 2026-11-04 and remained active as of 2026-11-06.

Who and what was involved

  1. Storm-3358 is a Cambodia-based intrusion cluster tracked by Microsoft Threat Intelligence that mainly hits financial organizations in Southeast Asia plus a few large global financial firms.
  2. The ARTEX framework is an agentic AI offensive tool whose agents carry out scanning, exploitation, and post-compromise actions.
  3. The infrastructure includes WISP C2 nodes at 199.19.95.66, .67 and .68.
  4. Several IPs are labelled ARTEX servers: 91.92.240.238, 85.11.168.45, 178.16.54.129 and 31.77.229.35.
  5. A separate C2 server is listed at 85.11.167.52.

How the agents communicated and organised

  1. ARTEX agents were used alongside conventional automated tools rather than alone, combining AI-driven actions with standard tooling.

What they used to do it

  1. The actors scanned internet-facing systems and exploited flaws such as anonymous file upload and server-side request forgery to install open-source RATs.
  2. Storm-3358 deployed the WISP, vshell, and AdaptixC2 RATs for persistence and lateral movement.
  3. ARTEX agents used netcat and curl to download the RATs onto compromised systems.
  4. ARTEX agents planted web shells on compromised web servers for persistence and harvested credentials from public sources and compromised systems.
  5. The agents also performed brute-force password attacks, password spraying, API enumeration and abuse, and database dumping.
  6. ARTEX by default uses tools including Nmap, curl, wget, Python HTTP clients, Norma SDK WebFetch, and headless Chromium via Playwright.
  7. The actors used curl to fetch remote files, with URLs or IPs extracted from the command line.
  8. Process creation events with EventID 4688 were checked for curl in the command line to surface remote file retrieval.
  9. Microsoft provided a Sentinel advanced hunting query to find related activity in customer networks.
  10. An ASIM-based query checks the listed IOC IPs across network session data sources over the past 30 days.

What they did to avoid being caught

  1. Threat actors regularly use proxies and VPNs to change the source IP of malicious traffic and evade static detections.
  2. Microsoft notes that threat actors regularly use proxies and VPNs to change source IPs and bypass static detections.

What made it possible

  1. Agentic tools like ARTEX can use stolen credentials and compromised systems at machine speed to move laterally, escalate privileges, and exfiltrate data.
  2. The activity reflects a broader trend of actors speeding up every stage of the attack chain with agentic AI penetration testing frameworks.
  3. Microsoft Sentinel and its ASIM network session parser are the platform that lets defenders run the IOC hunting queries.

How it was found

  1. Agentic frameworks scanning a target send many varied requests in a short time and retry or advance at machine speed, which is a detectable behavioral signal.
  2. Default tooling may emit distinctive traffic such as default user agent strings, which can alert defenders even though an AI agent will quickly change them.
  3. Honeypots, decoys, tarpits, and fake wins can divert autonomous AI attacks; in Microsoft testing they cut traffic to real assets and produced high-confidence attribution-quality signal.
  4. Sign-in activity from known threat actor infrastructure including ARTEX generates Verified threat actor IP risk detections at high risk in Microsoft Entra ID Protection.
  5. Defenders can detect the activity by hunting for the listed IPs in network session data over a 30-day lookback.
  6. A Sentinel query looks for curl processes in DeviceNetworkEvents over the past day and extracts URLs or IPs from the command line.

What it reached, and the numbers

  1. The campaign reached financial-sector organizations in Southeast Asia and a smaller number of large financial organizations globally.
  2. The IOC list covers eight IP addresses tied to WISP C2, ARTEX servers and a C2 server.
Show every field, with its source and grade

The AI system

Software around the model
ARTEX
C one secondary source states it
Tools it was given
Nmap, curl, wget, Python HTTP clients, Norma SDK WebFetch, headless Chromium through Playwright
C one secondary source states it
How it was connected
agent with tools
D inferred by us from the described behaviour, reason shown

What happened

Earliest source we hold
2026-11-07
D inferred by us from the described behaviour, reason shown
Who caused it
an outside attacker
D inferred by us from the described behaviour, reason shown
What went wrong
security breach
D inferred by us from the described behaviour, reason shown
AI agent
an AI agent acted
C one secondary source states it

Impact

Severity
moderate
D inferred by us from the described behaviour, reason shown
Why this severity
no source gives a fact that decides the level; we set it from the description
D inferred by us from the described behaviour, reason shown

LR-2026-0856 · version 5 · published 2026-11-10 after automatic checks; not yet checked field by field by a person