Artex Is Criminals’ New Favorite ‘Point-and-Shoot’ AI Hacking Tool
Cybercriminals are using the open-source Artex harness to automate multi-stage network intrusions, sending 80,000 requests in 90 minutes against live targets.
In short
Network defenders, security auditors, and system administrators · 80,000 requests sent in 90 minutes against Dosu · GitHub release July, deleted October 8, active forks · Patch systems, eliminate default credentials, monitor scans
By Dillip Chowdary • Nov 19, 2026 • Source: Wired

At a glance
- Who it is for
- Network defenders, security auditors, and system administrators
- The number
- 80,000 requests sent in 90 minutes against Dosu
- Available
- GitHub release July, deleted October 8, active forks
- What to do
- Patch systems, eliminate default credentials, monitor scans
- Language base
- Go programming language framework
- Tracked tools
- 45 automated attack frameworks documented by GitGuardian
In this post
Cybercriminals began launching automated intrusion campaigns against global targets in late August using Artex, an open-source Go penetration testing harness first published on GitHub in July, according to Wired's report.
This operational review examines the automated framework for enterprise security teams and network defenders evaluating emerging agent threats. Systems operators can audit network perimeters, patch exposed vulnerabilities, and inspect high-volume scanning traffic to stop unauthorized intrusions.
01Automated Penetration Harnesses in Real Attacks
Security researchers have documented a marked shift in cybercriminal operations as threat actors deploy autonomous frameworks against corporate networks, banks, and public agencies. These systems combine off-the-shelf testing scripts with external machine learning models to identify perimeter weaknesses and execute structured attack sequences. Operators direct the software using conversational instructions while the software executes tasks at machine velocity.
Multiple cybersecurity firms confirmed that the underlying utilities do not introduce novel exploit methods or sophisticated evasion mechanisms. Instead, the software compresses multi-day network probing sequences into operations that conclude in several hours. Unskilled attackers now initiate complex engagements that previously required advanced technical proficiency, transforming basic intrusion attempts into rapid operational breaches.
Enterprise defenders face expanding operational pressure as autonomous toolkits compress the window between initial scanning and perimeter compromise. Amazon Web Services chief information security officer CJ Moses confirmed that agentic frameworks have progressed from experimental testing into sustained offensive operations running with minimal human intervention. Organizations now confront persistent intrusion campaigns executed without manual pauses between phases.
02Architecture of the Go Framework
The Artex toolkit functions as an orchestration layer written in Go that routes structured instructions to external models through commercial and open interfaces. Created by a pseudonymous developer using the handle Autumn-27, the repository acts as an operational bridge between natural language prompts and established testing processes. Operators direct external models to scan target properties, test admin panels, and stage retrieved databases for unauthorized extraction.
- Plan decomposition breaks broad intrusion commands into structured technical tasks across target network endpoints.
- Action execution directs connected language models to probe system entry points and run target exploits.
- Findings presentation compiles harvested vulnerability data and credential discoveries for the human operator.
Frontier AI laboratories maintain internal safety filters designed to intercept malicious cyber instructions, but attackers frequently bypass these restrictions using unaligned or regional models. Jacob Larsen, founder of security firm Cypen, described the toolkit as an all-in-one point-and-shoot weapon compared to earlier single-purpose scanners. GitGuardian researcher Guillaume Valadon observed that the simplified workflow lowers technical barriers enough for inexperienced operators to launch functional attacks.
03Global Campaigns Against Enterprise Targets
Advertisement
Tech Pulse Daily
Get tomorrow's pulse first
Join engineers who read Tech Pulse before stand-up. Free, weekday mornings.
Malicious actors have deployed the framework across diverse economic and strategic sectors to execute financial theft and credential harvesting. Recent campaigns compromised personal records across thousands of South Korean financial institutions, attempted cryptocurrency theft, and scraped exposed developer credentials from public code repositories. The Federal Bureau of Investigation published an alert warning that adversaries are targeting defense, semiconductor, healthcare, energy, telecommunications, manufacturing, and government institutions worldwide.
State-aligned hacking groups quickly integrated the automation into broader corporate espionage and administrative intrusion operations. Security firm CrowdStrike tracked a China-linked threat group designated Vapor Panda that used the framework in October to target more than 200 organizations, including European foreign ministries and American defense firms. The automation enabled rapid scanning across hundreds of external targets without requiring individual manual configurations.
Enterprise infrastructure providers observed similar intrusions compromising identity systems and administrative controls. Microsoft security researchers identified Chinese hacking cluster Storm-0940 using the framework to access Microsoft 365 environments following an initial supply chain intrusion. The automated sequencing allowed operators to secure privileged administrative access across corporate environments in less than three hours.
04Repository Removal and Proliferating Variants
| Framework Name | Implementation Scope | Primary Operational Role | | Artex | End-to-end framework | Autonomous reconnaissance, exploitation, and data staging | | CyberstrikeAI | Attack harness | Scripted vulnerability probing and exploit execution | | Hexstrike | Modular harness | Multi-stage automated penetration workflows | | CAIRN | Agent framework | External model task routing and infrastructure testing | | PentAGI | Autonomous harness | Structured multi-objective penetration assignments |
Security analysts tracking developer ecosystems emphasize that open-source distribution prevents complete containment once code spreads across independent servers. Autumn-27 deleted the primary repository on October 8, publishing a disclaimer asserting the code was created exclusively for educational and research demonstrations. The author did not respond to inquiries regarding whether the removal followed formal law enforcement coordination.
GitGuardian identified 45 distinct autonomous attack frameworks actively circulating across developer repositories and underground channels. Alternative projects such as PentAGI and Hexstrike replicate similar orchestration mechanics, connecting external reasoning engines to standard security testing tools. Multiple independent developers continue maintaining mirrors and functional forks that circulate across developer communities despite public deletions.
Note: Autumn-27 removed the original repository from GitHub on October 8, but functional forks and derivative codebases remain active across private channels.
05High Noise Signatures and Defenses
Network defenders hold clear operational advantages because current autonomous frameworks generate extreme traffic volumes during scanning phases. Censys principal security researcher Silas Cutler noted that the utilities lack stealth controls and create distinctive forensic signatures across target firewalls. Agentic coding platform Dosu recorded an incident where an automated setup transmitted 80,000 distinct requests against its infrastructure in 90 minutes.
Researchers inspecting publicly accessible instances discovered that numerous operators fail to secure their own attack controllers. Censys and the Shadowserver Foundation discovered thousands of internet-exposed installations, enabling analysts to observe real-time operator behavior. One exposed controller revealed native Chinese speakers executing password spraying against a university network, displaying amateur operational patterns alongside sophisticated state-sponsored activity.
Enterprise mitigation relies on foundational defensive controls rather than specialized agent-blocking solutions. Security teams protect network perimeters by disabling default passwords, rotating administrative credentials, and applying software patches promptly. Automated rate limiting and network traffic monitoring allow defenders to detect high-frequency scanning bursts and terminate unauthorized connections before attackers achieve network persistence.
Frequently asked questions
What is Artex?
Artex is an open-source penetration testing harness written in Go that routes plain-language operator instructions through external artificial intelligence models to automate network scanning and exploitation.
Who created Artex and when was it removed?
A developer using the handle Autumn-27 published the framework on GitHub in July and deleted the repository on October 8 after stating it was built for research.
Which organizations have been targeted using Artex?
Attackers targeted over 200 entities including US defense contractors, think tanks, European foreign ministries, South Korean financial institutions, and Microsoft 365 environments.
How can system administrators defend against Artex attacks?
Administrators can block intrusions by enforcing regular software patching, eliminating default credentials, removing exposed secrets, and configuring rate limits to catch noisy scanning traffic.
Sources
Author
Dillip Chowdary
Writes Tech Bytes coverage of AI, engineering, and the tools that actually ship. Editor of Tech Pulse Daily.
Related on Tech Bytes
Advertisement