Cloud computing security company Barracuda suffered a cyberattack carried out by malicious actors using the AI penetration testing framework named Artex. This attack comes amid a noticeable escalation in automated cyberattacks relying on this framework to accelerate operations and expand their scope.
The Artex framework was first published on GitHub in July 2026 and is primarily designed for security testing by relying on a planning system to define objectives and generate tasks, along with agents to execute them while recording progress, results, and assets. The framework used Chinese large language models such as DeepSeek and GLM, in addition to other models, due to their high flexibility compared to US-made models.
Nature of Attacks and Targeted Entities
Although the framework was designed for security purposes, malicious actors began using it in real intrusion operations last August, and these attacks have seen a notable increase in recent weeks. The US Federal Bureau of Investigation (FBI) issued a new warning confirming that this framework allows users with limited technical experience to execute complex, multi-target operations simultaneously, targeting sectors such as financial companies in South Korea and ConnectWise.
Additionally, the Chinese cyberespionage group known as Vapor Panda used the Artex framework in October during a campaign targeting over 200 organizations across 14 countries, including foreign ministries, defense and nuclear organizations, and semiconductors. Reports from other security companies such as Cisco Talos and Amazon indicate that other groups have adopted the same framework for financial purposes, stealing smart model credentials, and cryptocurrencies.
Barracuda Attack Details and Practical Impacts
In the case of Barracuda, attackers exploited a valid, exposed cloud access key to map systems and reuse discovered secrets to access a widespread system and steal credentials. The systematic attack lasted for approximately 44 hours between November 1 and 2, targeting sensitive materials across secret vaults, account configurations, and server metadata services, leaving a backdoor in a Kubernetes cluster.
Technical analyses clarified that the attack resembles previous breaches targeting Railway and several AI development platforms. Although no official accusations have been leveled against a specific entity due to the framework's use by multiple groups, security companies confirm that attackers relied on basic traditional security vulnerabilities such as default credentials and unauthenticated interfaces, but executed them at ultra-high speed thanks to AI automation.