Home/Blog/FBI Artex warning: AI tool hits networks in under 3 hours
blog

FBI Artex warning: AI tool hits networks in under 3 hours

November 17, 2026 · Security News

Terminal screen illustrating the FBI Artex warning: an AI attack tool breaching networks in under three hours

Artex, an open-source AI penetration-testing tool, has carried attackers from first targeting to active compromise in under three hours, the FBI says in a new flash advisory. The FBI Artex warning ties the tool to the ConnectWise breach, which reached the Microsoft 365 environments of at least 2,250 customers, and to a Barracuda intrusion through a vendor credential nobody revoked. If an MSP manages your tenant, or old vendor access still works in it, you are exposed through the same doors.

Key takeaways

  • The FBI says Artex-linked attacks can move from targeting a system to active compromise in under three hours.
  • Microsoft found, and the FBI confirmed, that Artex was used to breach ConnectWise and reach at least 2,250 customers' Microsoft 365 environments.
  • Barracuda was breached through an unrevoked third-party service credential and rotated all customer credentials.
  • Attackers stole personal data on at least 68,000 people from South Korean financial institutions where MFA was missing.
  • Artex is now closed-source, but the old open-source version remains widely accessible.

What does the FBI Artex warning say?

The FBI says attackers are running autonomous AI hacking tools against live networks, and it singles out Artex, one of several open-source platforms built to probe defenses. The public reporting carries no advisory number or issue date. The FBI frames the activity as happening over recent months.

Named targets include financial services, government, technology, defense, telecom and internet providers, and law firms. The advisory names no CVE and no affected product version. No single patch closes this: the intrusions run on identity and configuration gaps.

How does an autonomous attack tool work?

Artex automates the tedious half of a penetration test. The operator connects it to a large language model, either a hosted frontier model from Anthropic or OpenAI or an open-source or non-U.S. one, and the model drives the job: scan the network fast, pick the easiest opening, push deeper.

The FBI names weak security settings and exposed or stolen passwords as the usual foothold. From there the tool helps the operator spread toward databases, mailboxes and other sensitive data. Artex debuted publicly this past July and has been used in several attacks since.

South Korea shows the pattern. Attackers used it against a number of financial institutions where weak defenses and missing multi-factor authentication (MFA) were major factors, and stole personal data on at least 68,000 people.

Why doesn't the noise save defenders?

Artex is not stealthy. The FBI calls its intrusions “noisy,” generating enough traffic and data to trip a target’s alerts. In practice, expect bursts of scanning followed by attempts on whatever services answer. That helps only if someone reads the alert and acts before the tool reaches a mailbox.

Barracuda CISO Arve Kjoelen put it bluntly: “Defenders who rely on spotting malicious activity and then responding to it will often arrive after the attacker has finished.” For a small IT team with no overnight cover, managed detection and response turns a 2 a.m. alert into a locked account instead of a morning surprise.

How did Artex reach ConnectWise customers and Barracuda?

Through the vendors. Microsoft found, and the FBI has confirmed, that attackers used Artex to break into IT management company ConnectWise and reach the Microsoft 365 environments of at least 2,250 customers, then stole data from affected ones.

Microsoft assesses the actors as likely Chinese; the reporting carries no separate FBI attribution. ConnectWise did not respond to TechCrunch’s request for comment.

Barracuda’s intruders were financially motivated criminals with a mundane way in. Artex found an exposed credential for a third-party service that Barracuda never revoked when the vendor relationship ended. That credential led to stores of credentials Barracuda used to collect data from customer systems.

Barracuda’s logs were too thin to rule out access, so the company chose to “assume the worst” and changed all customer credentials. It told a minority of XDR customers directly to revoke and recreate integration credentials, without saying how many. A secrets and credential exposure assessment hunts for that kind of forgotten credential before a scanner finds it.

Is Artex still available after going closed-source?

Yes. The developer closed the source after the attacks, but the FBI says the earlier open-source release remains “widely accessible,” so closing the repository did not take the tool out of attackers’ hands. Researchers recently found hundreds of Artex systems exposed online without a password, likely through misconfiguration, including the one used against South Korea.

If your own team runs similar tools, keep them behind authentication and rotate the model API keys they hold.

What should an SMB check this week?

Start by finding out whether you were touched. Pull Microsoft Entra ID sign-in logs for MSP and vendor accounts and look for unfamiliar locations or odd hours. Ask your MSP in writing whether it uses ConnectWise and what it has been told. The reporting does not say who is notifying the 2,250 affected customers, so don’t wait for a letter.

Then close the openings Artex looks for:

  • Remove app registrations and OAuth grants tied to vendors you no longer pay, and rotate any credential a vendor cannot account for.
  • Put phishing-resistant MFA, such as FIDO2 security keys, on email, VPN and admin accounts.
  • Block legacy authentication, which can sidestep MFA. New MFA does not revoke sessions already issued or passwords already stolen.
  • Confirm sign-in and audit logs are on and retained. Logging switched on after an incident cannot reconstruct what came before.

If access already looks wrong (new inbox rules, an admin nobody created, unexplained vendor sign-ins), treat it as account takeover. Export logs first, then revoke sessions and rotate credentials that reach finance or customer data, working from a clean admin device. Our account takeover recovery team runs that sequence when the in-house bench is thin.

FAQ

Does the FBI Artex warning name a patch I can install?

No. The reporting names no CVE, product version or patch. The attacks rely on exposed or stolen passwords, missing MFA and leftover vendor credentials, so the fix is identity and access cleanup.

How fast can an Artex attack compromise a network?

The FBI says some Artex-linked cases went from targeting a system to active compromise in under three hours.

Was ConnectWise breached using Artex?

The FBI confirmed Microsoft's earlier finding that Artex was used to break into ConnectWise and reach the Microsoft 365 environments of at least 2,250 customers. ConnectWise did not respond to TechCrunch's request for comment.

Did Barracuda rotate customer credentials after the Artex attack?

Yes. Its logs were not detailed enough to rule out access, so Barracuda changed all customer credentials and told a minority of XDR customers to revoke and recreate integration credentials.

Is Artex still open source?

Its developer made it closed-source after the attacks, but the FBI says the earlier open-source version is still widely accessible.

Are small companies at risk, or only banks and government?

The FBI lists finance, government, technology, defense, telecom and internet providers, and law firms, but SMBs are exposed through their MSPs and vendors, as ConnectWise's customers were.

A three-hour attack leaves no time to find a responder. An emergency incident response retainer puts one on call before you need them.


← All articles

get in touch

Tell us what you're running

Scoping is free. We reply within one business day, and under 30 minutes for active incidents.