# ARTEX forks: ReversingLabs publishes detection for 11 families

> ReversingLabs maps 11 ARTEX fork families to a single engine and gives defenders behavior-based detection instead of file hashes.

- **Severity:** Info: Research, tools, evals or lab claims with no direct exposure. Context, not an alert.
- **Category:** Research, Incident
- **Published by:** ReversingLabs
- **Disclosed:** 2026-11-18
- **Affects:** ARTEX, ScopeWeaver, LiteLLM, One API
- **Research by:** Robert Simmons
- **Primary source:** https://www.reversinglabs.com/blog/threat-researchers-notebook-detection-strategies-for-artex-and-related-forks
- **Page:** https://theinjection.dev/items/reversinglabs-artex-fork-detection/

ReversingLabs traced 11 ARTEX fork families to one shared engine and published behavior-based detection guidance, with 3,096 actionable host indicators and 184 operator binaries across Linux, Windows and macOS.

## What happened

Robert Simmons of ReversingLabs published a threat research notebook on 18 November 2026 covering ARTEX, the agentic pentest framework, and its rebranded forks. The team counted 11 fork families sharing one engine, 184 operator binaries and 969 confirmed live consoles, and kept 3,096 of 3,147 host indicators as actionable after triage.

## How it works

Detection rests on behavior rather than file hashes, because many forks are rebrands with no code changes. Simmons describes an LLM-driven cadence of process and network bursts separated by short processing pauses, and ties the randomized scanning to the projectdiscovery/blackrock Go module, which appears as high-entropy, non-sequential connection attempts with many unanswered SYNs. The report also names DNS query bursts from dnsx and cdncheck, unusual SSH client identification strings, and brute-force and enumeration patterns against LLM gateway admin APIs.

## Who is affected

Organizations that expose internal LLM gateway and management panels are the described targets, with Sub2API, One API, New API and LiteLLM named. Hosting for actionable indicators concentrates at PEG Tech, UFO/B2-Net, Yancy, Tencent and Alibaba.

## What to do

Remove LLM gateway admin panels from the internet, enforce SSRF allowlists, and keep agent credential directories such as ~/.claude and ~/.codex out of web-accessible paths, container images and build artifacts, rotating any key that was exposed. Hunt the published indicators retroactively across July to November 2026, weighting alerts by hosting netblock.

## Quick facts

- **Fork families:** 11, sharing one engine
- **Host indicators:** 3,096 actionable of 3,147
- **Operator binaries:** 184, across Linux, Windows and macOS
- **Live consoles:** 969 confirmed
- **Found by:** Robert Simmons, ReversingLabs
- **Published:** 18 November 2026

## Sources

1. [RL Research](https://www.reversinglabs.com/blog/threat-researchers-notebook-detection-strategies-for-artex-and-related-forks)
2. [Unit 42 Census](https://unit42.paloaltonetworks.com/artex-and-ai-powered-red-teaming-tools)
3. [FBI FLASH](https://www.ic3.gov/csa/2026/261113.pdf)
