
Barracuda Hit by Artex AI Framework as Chinese Groups Scale Automated Attacks
The Artex agentic AI penetration testing framework, published on GitHub in July 2026, has been widely adopted by Chinese threat actors for real-world intrusions since at least August. It uses a planner to set objectives and generate tasks, with agents executing them while the platform logs progress, findings, and assets. Primarily relying on permissive Chinese LLMs such as DeepSeek v4.1-flash, GLM-5.3, and Grok 4.6, the tool enables low-skill operators to run complex multi-target operations at high speed. The FBI has issued an alert highlighting its role in attacks on South Korean financial firms and ConnectWise, while SentinelOne linked Vapor Panda to a campaign against over 200 organizations across 14 countries. Barracuda itself suffered a November 1-2 intrusion in which attackers used an exposed cloud key to compromise roughly 10 accounts in 44 hours, exfiltrate secrets, and plant a Kubernetes backdoor. Similar activity has been observed by Storm-0940, Storm-3358, and other groups targeting AI platforms, crypto exchanges, and supply-chain vectors.
SecurityWeekNov 16, 2026Translated from English
