FBI Warns AI Hacking Tool Artex Is Fueling Attacks Worldwide
A new breed of AI penetration-testing frameworks is letting unskilled criminals hit companies, banks and governments at machine speed. The FBI says Artex, the most prominent of them, has been used against financial, defense, semiconductor and health care targets worldwide. Attacks that once took days now finish in hours.
- Original source: wired.com.
- The source published it on Nov 18, 2026 at 4:35 PM.
- No second outlet has reported it yet. Treat it as a single-source claim.
The FBI issued an alert last week naming Artex, an open-source AI attack framework, as a tool used by a range of malicious actors against organizations across financial services, government, technology, defense, semiconductor, energy, health care, manufacturing, telecommunications and legal industries. Since roughly the end of August, companies, government agencies and banks around the world have been targeted by criminals running this new wave of AI-powered penetration testing setups, according to alerts from about half a dozen security firms.
Artex is written in the Go programming language and was first published on GitHub in July by a pseudonymous developer using the handle Autumn-27. It has since been used across every stage of an intrusion, from scanning networks and gaining access to moving inside systems and staging data for export. Jacob Larsen, an offensive security researcher and founder of the firm Cypen, calls it an all-in-one product and a point-and-shoot weapon, noting earlier AI pen-testing tools handled only reconnaissance or only testing.
The framework is not itself an AI product. It is a harness that points external frontier and open-source models, connected through APIs, at existing penetration testing tools and processes. Operators type plain-language prompts — Check Point researchers recorded one set of hackers asking Artex to launch a penetration attack on a redacted victim to reach its admin panel, database or server. Multiple agents then divide the plan into tasks, attempt them and report back. Guardrails in frontier models can block hacking activity when detected, but free Chinese and abliterated models are getting better at breaking in.
The results have been fast and loud. China-linked hackers tracked as Vapor Panda used Artex in October to target more than 200 organizations, including US think tanks, semiconductor and defense firms, and government foreign ministries in Europe, according to CrowdStrike. Microsoft said Chinese hackers it calls Storm-0940 used the tool to try signing in to Microsoft 365 systems after a supply chain attack, gaining privileged access in under three hours. Reports also link Artex to the theft of personal information from thousands of South Korean financial institutions, potential cryptocurrency theft and scans for exposed login details on GitHub.
Autumn-27 pulled Artex from GitHub on October 8, writing that they are in no way associated with the malicious attacks and that the tools were built for educational and research purposes. WIRED got no response to an email sent to an address linked to the developer. Copies of the framework keep circulating anyway: Guillaume Valadon of GitGuardian says his firm has documented 45 different AI attack frameworks, with CyberstrikeAI, Hexstrike, CAIRN and PentAGI frequently named by researchers.
Defenders have one advantage — the tools are noisy. Silas Cutler, principal security researcher at Censys, says they are not stealthy right now, and many attacks fire thousands of probes in quick succession; agentic coding platform Dosu reported Artex sending 80,000 requests in 90 minutes. Censys and the Shadowserver Foundation found thousands of Artex instances exposed to the open internet, some left unsecured enough for researchers to watch operators at work. Two exposed servers Cutler examined last week appeared to show Chinese speakers targeting other Chinese entities, including a university hit with a password-spraying attack that he says had the vibe of a student trying to change their grades. Cutler describes a mix of organized, sophisticated criminal activity running alongside what he calls the juvenile crime realm.
TopicsArtex · FBI · CrowdStrike · Microsoft · AWS · Censys · GitGuardian · Vapor Panda
Source: wired.com
Written by AI from verified reporting.
More stories
- Tech & AI heise.de 0 viewsMicrosoft rolls out seven homegrown AI models and a weights licensing plan
- Tech & AI breakingthenews.net Caught by @hendra.setiawan99 0 viewsMicrosoft Taps Intel Foundry Packaging for Its Own Azure Chip
- Tech & AI cryptobriefing.com 0 viewsNvidia ships Vera Rubin NVL72 as AI labs sound the alarm
- Tech & AI theverge.com 0 viewsOpenAI paused a frontier training run; chipmakers never looked back

