ARTEX attacks on Japanese companies:GMO, JPCERT/CC and a 1,000-server probe.
In October 2026 GMO Cybersecurity by Ierae said it had confirmed ARTEX on attack infrastructure aimed at Japanese companies. JPCERT/CC later added AI pentest platforms, including ARTEX, to its alert, and Kyodo News reported a wider probe in November. This page follows those sources.
- When
- GMO 2026-10-19; JPCERT/CC update 2026-10-23; probe reported 2026-11-15
- Reported by
- GMO Cybersecurity by Ierae, JPCERT/CC, Kyodo News via EconoTimes
- Scope
- About 20 Japanese companies on one server; about 1,000 suspected servers worldwide
- ARTEX role
- ARTEX found on attack infrastructure
What happened
GMO's cyber crime countermeasure center wrote on 2026-10-19 that, while studying attack infrastructure behind breaches of Japanese companies, it confirmed ARTEX in use. Earlier it had only inferred AI involvement from forensic logs. It said the case resembled the campaign against South Korean financial institutions that CrowdStrike had reported.
From what GMO could observe, the infrastructure it studied targeted almost only Japan: about 20 Japanese companies, nine of them in the same industry. GMO then found six more AI attack servers aimed mostly at Japanese companies. Of the targets it confirmed, eight had already announced data leaks in September and October 2026, which GMO says points to one group behind several incidents. It shared the information with law enforcement, and Impress Watch reported the same findings on 2026-10-20.
JPCERT/CC updated its alert JPCERT-AT-2026-0030 on 2026-10-23 with a new case E: attack attempts using AI pentest platforms such as Cairn, CyberStrikeAI and ARTEX, mainly against internet-facing web applications. They include attempts to find and use vulnerabilities, attempts to steal configuration and backup files, and attempts against internet-exposed WSUS servers.
On 2026-11-15 EconoTimes, relaying Kyodo News, reported that JPCERT/CC and overseas security firms flagged about 1,000 servers as suspected AI attack infrastructure, that JPCERT/CC is tracking 600 to 700 of them, and that most hosted AI platforms including ARTEX. Traces of an AI coding tool from Anthropic were found on one server.
How ARTEX was involved
GMO describes ARTEX as an AI autonomous penetration test system developed in China, found on attacker infrastructure. In a note added on 2026-10-21 GMO says that the open-source developer being a Chinese engineer has nothing to do with who is using the software to attack.
JPCERT/CC lists ARTEX beside Cairn and CyberStrikeAI, so it is one of several tools named in the alert. The sources describe ARTEX on attacker-controlled infrastructure. They do not publish a step-by-step record of any single breach.
Confirmed and alleged
Reported as observed
- GMO confirmed ARTEX on attack infrastructure used against Japanese companies (GMO, Impress Watch).
- JPCERT/CC added case E, AI pentest platform abuse including ARTEX, to JPCERT-AT-2026-0030 on 2026-10-23.
- About 1,000 suspected servers were flagged by JPCERT/CC and overseas firms, and JPCERT/CC tracks 600 to 700 (Kyodo via EconoTimes).
Alleged, assessed or unclear
- That one group is behind several Japanese incidents is GMO's reading, and it is limited to what GMO could observe.
- GMO said it was still investigating whether other countries are attacked from the same infrastructure.
- Investigators called the Anthropic tool traces a possible sign that attackers used AI to strengthen operations. That is a suspicion, not a finding.
Takeaways for defenders
Drawn from what the sources themselves recommend or observed. Not a complete checklist.
- JPCERT/CC recommends rate limits on API requests, and separate limits on login, password reset, SMS sending and search functions.
- It recommends access control on every API endpoint, including non-public ones, with least-privilege API tokens that have an expiry and can be revoked quickly.
- It also recommends region limits where the service area is limited, applying fixes for known vulnerabilities, reviewing defenses against lateral movement after a web server compromise, and reviewing detection and first response.
- JPCERT/CC lists suspicious source IP addresses with observation dates and asks organizations to check whether they were accessed from them.
Sources
Every claim above comes from these pages or from this site's news data. Open them for the full reports.
- 相次ぐ不正アクセスの攻撃元の多くがAIであることを確認
GMO Cybersecurity by Ierae, 2026-10-19
- 日本企業への不正アクセス、中国のAIツール「ARTEX」使用を確認
Impress Watch, 2026-10-20
- JPCERT-AT-2026-0030 (updated 2026-10-23, case E)
JPCERT/CC, 2026-10-23
- Japan-Led Probe Flags 1,000 Servers as Suspected AI Attack Infrastructure
EconoTimes (relaying Kyodo News), 2026-11-15
Related pages
Related incidents
Follow the ARTEX story.
New forks, releases and reports, dated and sourced.
