Skip to content
See the World Through ScienceA project of ALLATRA

A Security Firm Says It Found the Attackers' Own AI Chat Logs on an Open Server

AI & Technology

Republish this story

Our work is licensed under Creative Commons BY-NC 4.0. You may republish this piece for free — with credit to ALLATRA Media and a link to the original, unedited beyond length trims, and not for commercial use.

Read the full license

Daytime view of high-rise office towers on Yeouido island in Seoul, seen from across the Han River.
Yeouido, the island district where much of Seoul's financial industry has its offices, seen across the Han River (illustrative)."Yeouido" by pixabay: @heecheol890102, via wikimedia, CC0

Security company CrowdStrike Intelligence reported on Oct. 7, 2026, that it had traced the servers behind a run of intrusions at South Korean financial organizations in which data was taken, and that the operator had left its own working files on one of those servers, in folders anyone could browse.

CrowdStrike's published account of the intrusions says those files gave it direct insight into how the operator worked. They included session logs from Claude Code, an AI coding agent, setup files for ARTEX, a freely available tool that automates security testing, and the agent's stored memory. The campaign was active from late September to early October 2026, and how many organizations were affected remains unconfirmed.

CrowdStrike says the activity has not been attributed to a named adversary. With moderate confidence, it assesses that the actor is likely a Chinese speaker and financially motivated, and it gives two reasons: the use of ARTEX, which the report describes as developed in China, and Chinese-language prompts it found in the files. The assessment is the vendor's own.

According to the report, the ARTEX instance used DeepSeek v4.1-flash as its main model backend, and the operator added GLM-5.3, from Zhipu AI, and Grok 4.6 for further agent sessions. The operator also asked the AI agent where stolen Korean data is usually sold, and for help finding Telegram groups that trade it.

As for what was breached, the report cites industry reporting rather than its own observation: several South Korean financial firms had data breaches over that period, and at two of them the systems involved were a loan progress inquiry service used by brokers and an employee mobile work-support system.

CrowdStrike Intelligence says it expects attackers to keep trying AI tools to work faster. The report lists the network indicators the company tied to the activity and maps them to ATT&CK, MITRE's public catalog of attacker techniques, so defenders can search their own records.

Sources

Spot an error?

Spot an error?

Report an error

Spotted a mistake on this page? Tell us what's wrong and our editors will take a look.

What kind of problem?

Only if you'd like us to be able to follow up. We won't use it for anything else.

We correct mistakes openly. Select any text to flag it. Fixes are logged under our Corrections Policy.

Report an error

Reporting on

A Security Firm Says It Found the Attackers' Own AI Chat Logs on an Open Server

What kind of problem?

Only if you'd like us to be able to follow up. We won't use it for anything else.

We read every report. Corrections are logged publicly.