ConcludedRan 2 daysin the edition since 21 Oct 20267 sourcesupdated 22 Oct 2026

ARTEX tool in Chinese state intrusion campaigns

The gist

Latest: ARTEX AI Agent Used to Breach South Korean Banks, Japanese Firms

South Korean officials described the intrusions as among the first AI-powered attacks on the global financial system. An AI pentesting agent being operationalized for financially motivated data theft at this scale, across two countries and with a formal police probe, marks a concrete case of open-source AI tooling enabling large-scale intrusions.

When stories like this move, I write about them in an email each morning.

One free email each morning, around 8 AM US Eastern.

The full picture

ARTEX, an open-source AI penetration-testing agent developed in China, was used in breach campaigns against South Korean financial firms and Japanese companies in late September and early October 2026. The developer, using the handle Autumn-27, identified by the WSJ as Li Puhua, acknowledged on October 8 that the tool had been abused and announced it would cease updates and go closed-source; the GitHub repository now returns 'Not Found'. At least seven South Korean financial firms were breached, with personal data of 68,000 people stolen. South Korea's National Police Agency cyber terror unit opened a formal probe. A GMO affiliate's investigation found ARTEX was used against approximately 20 companies in Japan, nearly all in Japan and nine in the same industry. ARTEX integrates with multiple LLM providers including Claude, ChatGPT, and DeepSeek, and is designed to autonomously decompose targets into tasks and execute them with real tooling. It was published in late July 2026 under an AGPL-3.0 licence.

How it developed

22 October 2026

Microsoft's Security Blog disclosed that Storm-0940, a China-state aligned actor active since 2021, used ARTEX to abuse ConnectWise OAuth applications and reach more than 2,250 Microsoft 365 tenants at 2,900 sign-in attempts per minute.

ConnectWise reported unauthorized access to customer tenants on October 16, 2026; because ConnectWise's apps held pre-authorized access to Microsoft 365 environments, the attack bypassed MFA and interactive login. Some intrusions progressed from first targeting to Global Administrator privileges in under three hours via Microsoft Graph. The campaign is separate from earlier financially motivated ARTEX breaches at South Korean and Japanese firms.

First in the edition · 21 October 2026

20 October 2026

Bloomberg and Mainichi report GMO affiliate findings on Japan attacks targeting ~20 companies

10 October 2026in the 21 Oct edition

ARTEX AI tool breach in South Korea

The Wall Street Journal reported at least seven South Korean financial firms were breached and 68,000 people's data stolen using ARTEX, a Chinese-developed AI penetration-testing agent.

South Korea's National Police Agency opened a formal investigation, and CrowdStrike found the campaign through open directories at a Hong Kong IP address containing Claude Code session histories and ARTEX configuration files. Developer Li Puhua, identified by the WSJ and using the handle Autumn-27, posted a statement on 8 October 2026 acknowledging misuse and announced the tool would go closed-source; the GitHub repository subsequently returned 'Not Found'.

First citedRohan Paul on X mlq.ai mainichi.jp wsj.com+3 more

Sources

Rohan Paul on X