The FBI issued a FLASH alert on November 13, 2026, warning that attackers are using ARTEX, an open-source agentic AI penetration-testing framework, to automate cyberattacks against organizations worldwide2,3. The alert ties the framework to a recent ConnectWise supply-chain breach that compromised Microsoft 365 customer environments.
Microsoft previously linked ARTEX to Storm-0940, described as a China-state-aligned group targeting Microsoft 365 environments through compromised ConnectWise application credentials.
How ARTEX operates
ARTEX is written in Go and connects external large language models to agents that translate operator-set objectives into executable commands. Agents scan targets, run tools, review results, and adjust failed steps with limited human input throughout an attack campaign. The FBI describes intrusions in which ARTEX agents reached privileged access in under three hours.
Google observed individual automated bursts lasting two to 25 seconds, leaving defenders little time to assess connected events. The FBI's alert catalogs attack vectors including SQL injection, missing API authentication, exposed passwords, and abuse of trusted cloud connections. After initial entry, agents added credentials to existing applications, deployed web shells, and attempted credential theft.
A pattern across agentic tools
The FLASH follows Vector Wire's prior reporting that forks and copies of ARTEX remain publicly available and capable of executing credential attacks and cloud breaches at machine speed, even after the original developer made the project closed-source[2]. Separately, attackers have weaponized AntiFox, another open-source agentic penetration-testing platform, to steal AWS credentials from internet-facing services[3]. Days before the FBI alert, Unit 42 published research documenting AI agents powered by DeepSeek and Zhipu GLM autonomously breaching 10 organizations across the semiconductor, U.S. defense policy, and civil society sectors during August and September 2026[1].
ANALYSIS The FBI's decision to issue a FLASH — a format reserved for urgent, actionable intelligence — on a tool that is architecturally a legitimate pen-testing framework, not a malware family, marks a shift: the threat is the agentic orchestration layer itself, not a novel exploit. The sub-three-hour privilege-escalation timeline and two-to-25-second automated bursts described in the alert compress the detection window well below the response cadence of most security operations centers.