The FBI issued a FLASH advisory on Friday warning that threat actors have weaponized ARTEX, an open-source autonomous penetration testing framework, to execute automated cyberattacks globally since at least August 2026. Built in Go, the tool uses external large language models—including Anthropic Claude, OpenAI GPT, DeepSeek, GLM, Grok, and Kimi—to scan vulnerabilities, escalate privileges, and extract data with minimal human supervision. SecurityWeek also reported that Chinese threat actors have widely embraced the agentic AI software, contributing to recent intrusions such as a breach at Barracuda. [2][3]
According to the FBI, ARTEX lets attackers with low technical skills automate complex multi-target operations, occasionally advancing from an initial probe to privileged system access in under three hours. Although developer Autumn-27 removed the original GitHub repository on October 8 following reports of abuse, the framework continues circulating through archived copies, localized variants, community forks, and a paid version dubbed ARTEX PRO. Researchers discovered over 1,600 exposed ARTEX consoles by early November. [3]
The utility has fueled prominent breaches, including an October supply-chain incident at ConnectWise Cloud Backup that affected Microsoft 365 environments. Microsoft linked that intrusion to Chinese state-aligned group Storm-0940, which used the tool to target more than 2,250 downstream tenants. In South Korea, police arrested seven suspects accused of leveraging ARTEX in intrusions across over 90 organizations, stealing roughly 8.2 million records. To defend against the automated bursts, researcher MalwareMustDie has released an open-source detection signature pack. [3]