Breach · Breaking

Barracuda says an ARTEX-style AI attacker used a forgotten vendor key to roam about 10 of its cloud accounts in 44 hours; Managed XDR customers told to rotate integration credentials

Barracuda disclosed that an attacker used a cloud access key it had given an AI vendor and never retired to move through about 10 of its cloud accounts in roughly 44 hours on Nov 1–2. One system it reached stores the credentials Barracuda Managed XDR uses to pull data from customers' Microsoft 365, Google Workspace and endpoint tools. Barracuda says the attack matches the ARTEX AI hacking framework and has told affected customers to revoke and recreate those credentials.

Nov 16, 2026 · 3 min read

What happened. Barracuda posted a Trust Center advisory on Nov 10, a technical write-up on Nov 11 and an update on Nov 13; SecurityWeek put it in front of a wider audience on Nov 16. The attacker made its first call at 04:33 UTC on Nov 1 with a valid cloud access key. Barracuda had issued that key to a third-party AI vendor for an earlier project and never retired it. "That is on us," the company wrote. The vendor's development environment was compromised in October, and Barracuda says the key sat in a configuration file in that vendor's GitHub repository. Barracuda has not named the vendor.

How it moved. The attacker read stored secrets such as AWS access keys and Azure service principal secrets, then used each one to reach the next system: cloud accounts, then a data analytics system, then a source-code access token, then Kubernetes clusters. Early on Nov 2 it pushed short-lived branches to a few repositories to trigger a build workflow that printed the build system's secrets into the job log, encoded so the log's secret masking would not hide them. It left a container in one Kubernetes cluster that mounted the host's file system, a way back in, and tried to create an admin cloud identity, which was blocked. On some servers it searched for crypto wallet files and seed phrases. Barracuda says none were there. In all it touched about 10 cloud accounts in roughly 44 hours. The last activity was at 23:31 UTC on Nov 2. An engineer spotted odd activity on Nov 4, and Barracuda removed the container that evening.

Why customers care. One system the attacker could reach stores the credentials Barracuda XDR uses to collect data from customers' connected systems, such as Microsoft 365, Google Workspace and endpoint security tools. Barracuda says it has found no direct evidence those customer credentials were viewed. Even so, it notified customers with potentially affected credentials and told them to revoke and recreate them and to review activity in connected systems from Nov 1 onward. The advisory covers only API integrations for Barracuda Managed XDR Cloud or Monitored Endpoint. If you got no notice from Barracuda, the company says you don't need to act.

The ARTEX link. The attacker named everything it created with "i" plus digits (i31911, i28324probe, i19engx1) and used GitHub workflow files i19ghci.yml and i19ci2.yml. Barracuda says those names, the order of steps and the built-in credential and wallet hunting all match ARTEX, the open-source AI pen-testing framework behind the FBI's Nov 13 FLASH alert. It has not attributed the attack, because many groups use ARTEX. It compares the intrusion to October's Railway breach.

Defender move. List every cloud key you have given a vendor or contractor and kill the ones tied to finished work. Lock CI so a new branch can't trigger a workflow that can read production secrets. Alert on cloud or Kubernetes resources created with machine-style names. Any security vendor that holds API tokens into your M365 or Google tenant is part of your attack surface, so know which ones do and how fast you can rotate them.

Sources