ReadingResearchRadarInvestment framework
Sign in / Sign up
← Back to reading
NewsMaterial2 min readDisconfirmAI Editorial

Security firm Barracuda discloses AI-driven cloud breach: Attackers used ARTEX framework to compromise 10 accounts in 44 hours

Cybersecurity company Barracuda revealed that hackers used the AI penetration testing framework ARTEX to autonomously breach its cloud environment, stealing credentials across 10 accounts in just 44 hours. The attack originated from a stale access key left active for a third-party AI vendor.

2026-11-19 09:55 GMT+8

Cybersecurity firm Barracuda Networks has disclosed a significant security incident in which attackers used an artificial intelligence framework to autonomously breach its cloud environment. According to the company’s investigation, the attackers utilized the AI penetration testing tool ARTEX to traverse 10 cloud accounts within 44 hours, continuously harvesting credentials, exploring internal systems, and attempting to establish backdoors.

The initial entry point was a cloud access key that Barracuda had previously provided to a third-party AI company for a project but failed to revoke after the engagement ended. In October 2026, the development environment of this AI vendor was compromised, allowing attackers to steal the dormant key. On November 1, the attackers began using the key to explore accessible resources, extracting stored AWS access keys and Azure service principal secrets.

Over the next 24 hours, the AI-driven attack system leveraged these stolen credentials to progressively infiltrate data analysis systems, source code access tokens, and Kubernetes clusters. By early November 2, the attackers pushed short-lived branches to several code repositories, triggering build workflows that leaked sensitive information from the build system into job logs. The AI also searched for cryptocurrency wallet information on servers with shell access and attempted to create new cloud administrator identities, though this latter attempt failed.

Barracuda detected anomalous account activity on November 4 and connected the events after receiving notification of the security incident from the third-party AI company. While no direct access to Barracuda XDR customer data has been confirmed, incomplete event logs mean the company cannot rule out potential exposure. Barracuda has since rotated all potentially affected credentials and notified some customers to revoke and re-establish API integration keys.

Sign in / Sign up
ReadingResearchRadarInvestment framework

This incident highlights the growing risk of AI agents executing complex, multi-stage attacks at machine speed. Unlike traditional breaches reliant on manual exploitation, the use of ARTEX allowed for rapid lateral movement and credential harvesting across multiple cloud environments, underscoring the critical need for strict lifecycle management of third-party access keys.

Sources

  • ithome.com.twiThome↗
Continue researching →