CrowdStrike: China-linked VAPOR PANDA used AI agents against 200 organizations
The company confirmed post-exploitation at nine organizations; the larger figure describes targets, not successful breaches.
CrowdStrike says the China-linked VAPOR PANDA group used the ARTEX agent framework against more than 200 organizations across at least 14 countries during an 11-day campaign. The figure refers to targets, not breaches: the company confirmed post-exploitation at nine victims, where attackers deployed the BLOODALCHEMY backdoor, and password-spraying or credential-scanning against at least 12 more.
Targets included foreign and interior ministries, telecom providers, and U.S. companies in pharmaceuticals, semiconductors, defense, nuclear energy and finance. In one intrusion, ARTEX chained an unsigned SAML assertion with an XXE flaw to reach cloud metadata and recover an enterprise administrator account. In another, unrotated keys from an old mail-server export were used to forge an administrator token.
ARTEX ran primarily on DeepSeek, supplemented by a second unidentified commercial model. CrowdStrike says planning agents in one intrusion consumed roughly 1.4 billion input tokens. Recovered transcripts show a Chinese-speaking operator directing planner and worker agents to write web shells, extract domain administrator passwords and find Windows machines without CrowdStrike’s Falcon sensor.
CrowdStrike says this campaign is separate from the ARTEX activity Microsoft attributed to Storm-0940. Together, the reports suggest the framework is spreading among China-linked groups. The findings are CrowdStrike’s assessment based on its own telemetry.
DeepSeek
