Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
ReleaseCode / SDK · ActionableTechCrunch→ original

FBI: AI Tool Artex Breaches Companies in Under Three Hours

The FBI has warned that hackers are using Artex, an open-source tool that connects to a language model and hunts for holes in a company's network on its own. In some cases, less than three hours pass between picking a target and breaching it, which leaves the on-call team too little time to react. The attackers stole data on at least 68 000 people in South Korea and got into Microsoft 365 at no fewer than 2250 ConnectWise customers. Small oversights are enough: a contractor's forgotten password or missing two-factor login. Close those gaps now.

AI-processed from TechCrunch; edited by Hamidun News
FBI: AI Tool Artex Breaches Companies in Under Three Hours
Source: TechCrunch. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceTechCrunch ↗
Editor-in-Chief
Key metric #1Artex appeared in July; its author closed the source code, but the old version is still circulating online
Key metric #2Barracuda did not revoke a contractor's key and later changed all customer credentials
Key metric #3Researchers found hundreds of Artex servers online with no password

The FBI has issued an urgent warning: hackers around the world are using Artex, an open-source tool that uses a language model to scan a company's network, find a weak spot, and break in, with almost no human involvement. "Less than three hours" is how long it takes, in some cases, for an attack to go from choosing a target to an actual breach. Imagine your company still has a working access key belonging to a former contractor that nobody revoked.

Until recently, only a patient human could find it. Now an automated tool does it within hours, while the on-call engineer is still reading the first alert. If you are responsible for infrastructure, cloud access, or a product that holds customer data, this story belongs on your to-do list for the coming weeks.

What Artex is and how it works

Artex belongs to a class of AI platforms for penetration testing (pentesting). Such tools were originally built to test the defenses of your own network: you hire a "friendly" hacker who finds the holes before real criminals do. But the same tool works just as well in the hands of attackers.

It works in a simple way. Artex connects to a large language model. According to the FBI, this can be a cloud model from Anthropic or OpenAI, an open model, or a non-American one. It then quickly scans the victim's network and looks for the easiest targets: weak protection, and exposed or stolen passwords. Once it gets a foothold through one such hole, it spreads across the network and reaches databases, mailboxes, and other sensitive information.

The FBI candidly notes that Artex attacks can be "noisy": they generate a lot of traffic and may trigger alarms. But speed decides the outcome: defenders simply don't have enough time to work out what is happening. According to the FBI, the targets are financial companies, government agencies, technology and defense firms, telecom operators, and law firms.

What has already been breached

Since its first public appearance in July, Artex has been linked to several high-profile incidents.

  • South Korean financial organizations. Weak security and a lack of multi-factor authentication (a login that requires confirmation through a second channel) helped the attackers. Personal data of at least 68 000 people was stolen.
  • ConnectWise. The FBI confirmed Microsoft's earlier findings: attackers whom Microsoft believes are likely Chinese gained access to Microsoft 365 cloud environments of at least 2250 customers of this IT company and exfiltrated data. ConnectWise did not respond to a request for comment.
  • Barracuda. The company said it was breached by financially motivated criminals. The tool found exposed credentials for a third-party service that had not been revoked after the company finished working with a contractor. Through them, the hackers reached the password vault Barracuda uses to collect data from customer systems. The logs were not detailed enough to rule out access, so the company "assumed the worst" and changed all customer credentials. Chief Information Security Officer Arve Kjølen wrote that defenders who first notice suspicious activity and only then respond often arrive after the attacker has already left.

Barracuda did not disclose how many of its customers were affected.

Closing the code didn't help.

What can you do?

After the wave of attacks, Artex's developer made the program closed-source so as not to help criminals. However, according to the FBI, the earlier open version is still circulating online and remains "widely available." There is also an ironic detail: the hackers' own platforms are poorly secured, too. Last week, researchers found hundreds of Artex systems exposed to the internet without a password, probably because of a configuration error. One of them was the system used to attack South Korea.

The main lesson: Artex does not invent new vulnerabilities. It quickly exploits old, simple ones. So the defense is down-to-earth:

  • enable multi-factor authentication everywhere, especially in admin panels and cloud consoles;
  • when you finish working with a contractor, revoke their keys and tokens immediately, and keep a list of them;
  • set up detailed logging: without it, as in Barracuda's case, you will have to rotate all secrets "just in case";
  • don't expose internal and test services to the internet without a password, including your own AI tools;
  • prepare automated detection and access restriction: a human may have hours to respond, not days.

A caveat: the FBI warning does not include complete statistics on the number of attacks, and the ConnectWise data relies on Microsoft and has not yet been commented on by the company itself. But the direction is clear: AI shortens the time between "spotted a hole" and "got inside," and the basic security hygiene that many have put off for years is becoming a question of survival.

Story

Key Questions Answered

Closing the code didn't help. What can you do?

After the wave of attacks, Artex's developer made the program closed-source so as not to help criminals. However, according to the FBI, the earlier open version is still circulating online and remains "widely available." There is also an ironic detail: the hackers' own platforms are poorly secured, too. Last week, researchers found hundreds of Artex systems exposed to the internet without a password, probably because of a configuration error. One of them was the system used to attack South Korea.

What Artex is and how it works?

Artex belongs to a class of AI platforms for penetration testing (pentesting). Such tools were originally built to test the defenses of your own network: you hire a "friendly" hacker who finds the holes before real criminals do. But the same tool works just as well in the hands of attackers. It works in a simple way. Artex connects to a large language model. According to the FBI, this can be a cloud model from Anthropic or OpenAI, an open model, or a non-American one.

What has already been breached?

Since its first public appearance in July, Artex has been linked to several high-profile incidents. South Korean financial organizations. Weak security and a lack of multi-factor authentication (a login that requires confirmation through a second channel) helped the attackers. Personal data of at least 68 000 people was stolen. ConnectWise.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…