China-Linked Hackers Use AI Agents to Attack Over 200 Organisations at Once

CrowdStrike's OverWatch team has identified a campaign by a group it tracks as VAPOR PANDA, which used an AI-driven framework called ARTEX alongside the Chinese-developed DeepSeek large language model and a second commercial model. Starting on October 17, 2026, the group targeted more than 200 organisations worldwide. Victims included foreign ministries in Latin America, Europe, Asia and the Middle East; semiconductor, defence and think tank organisations in the U.S.; and telecommunications providers in multiple regions.

CrowdStrike says the campaign marks an escalation in how China-linked actors use AI agents. Instead of a single intrusion at a time, a human operator directed multiple simultaneous engagements, which ARTEX calls "tasks". Exploitation activity from one ARTEX controller reached entities in at least 14 countries over an 11-day window. The controller intermittently exposed tooling, target files and agent transcripts, which helped researchers see the scale of the operation. Investigators also found the group deploying a backdoor known as BLOODALCHEMY.

The operation also built security evasion into its automated workflow, including instructions to identify hosts that do not have the CrowdStrike Falcon sensor. CrowdStrike notes this campaign is separate from ARTEX activity reported elsewhere under the Storm-0940 name, and that ARTEX has previously been used by financially motivated and unattributed actors. The source excerpt covers only the opening of the report, so further technical detail is not included here.