CrowdStrike says the China-linked group VAPOR PANDA used AI agents to target more than 200 organizations in at least 14 countries over 11 days, with intrusions confirmed at nine.
The activity began on October 17, 2026. The group used ARTEX, a framework for AI agents that ran mainly on DeepSeek, a model developed in China, and a second commercial AI model that was not identified. Targets included foreign ministries in several regions, U.S. organizations in areas such as defense and semiconductors, and telecommunications providers.
CrowdStrike said the nine confirmed cases involved access to systems. They also involved installing the BLOODALCHEMY backdoor. The activity then included further information gathering and attacks on login credentials.
Targeting an organization alone did not mean it was breached. CrowdStrike also reported attempts to guess passwords and scan for login details at at least 12 other organizations. CrowdStrike's account says agents carried out work in parallel.
A human operator directed them and stepped in at key decision points. The operation did not run entirely on its own. CrowdStrike said the campaign's control system left its tools, target files and agent transcripts exposed.
That helped investigators link the activity to VAPOR PANDA. These findings are CrowdStrike's assessment. CrowdStrike said this campaign was separate from other ARTEX activity attributed elsewhere to Storm-0940.