Published November 8, 2026 | Version v1

VERIFIED FACTS OF THE ARTEX CAMPAIGN AND THE ADMISSION-CONTROL GAP: HOW ST-T1024'S AGENTIC EXTENSION (A1–A16) COULD HAVE FORBIDDEN THE ATTACKS

Description

Between September 18 and October 18, 2026, approximately 80 Japanese companies and
organizations across insurance, finance, and logistics were subjected to a coordinated cyber campaign.

Forensic analysis by GMO Cybersecurity by Ierae identified six overseas public servers used in the attacks; all six carried traces of ARTEX, a Chinese-language autonomous AI penetration-testing agent that chains large language models into a single attack orchestrator . Eight of the targeted entities have publicly confirmed data breaches .

 

This paper presents the verified factual record of the campaign and then presents ST-T1024,
a deterministic standard for embedded and agentic AI, arguing that its agentic
extension—specifically mechanisms A10 through A16—constitutes a portable,
machine-verifiable admission-control layer that would have refused an agent of ARTEX's
class at the organizational boundary. The paper does not claim ST-T1024 prevents network intrusion generally. It claims something narrower and more defensible: that the absence of a standard by which a platform can verify an agent's bounded behavior set before admission is the structural gap that ARTEX exploited, and that ST-T1024's Agent Certificate (A11), Action Vocabulary Mapping (A14), and Refusal-as-Wire-Protocol (A15) close that gap by construction.

Files

VERIFIED FACTS OF THE ARTEX CAMPAIGN.pdf

Files (135.4 kB)

Name Size
md5:148e13e0906407086a4082b6a7702143
135.4 kB Preview Download

Additional details

Dates

Available
2026-11-08