China-linked VAPOR PANDA uses the ARTEX AI agent framework to break into 200+ government, pharma, defense and telecom organisations at once, then plants the BLOODALCHEMY backdoor

CrowdStrike (2026-10-29) reports that from 2026-10-17 a single ARTEX controller linked to the China-nexus group VAPOR PANDA ran parallel AI-agent intrusions against more than 200 organisations in at least 14 countries. The agents used DeepSeek plus a second, unconfirmed commercial LLM, with a human operator steering them in Chinese. Targets: foreign ministries in Latin America, Europe, Asia and the Middle East, a Middle East interior ministry, US pharmaceutical, semiconductor, defense, nuclear and financial firms and think tanks, and telecoms in the US, Belgium and the Philippines. Post-exploitation is confirmed at 9 organisations, where the BLOODALCHEMY backdoor was installed, and password spraying hit at least 12 more. Methods seen: an unsigned SAML assertion chained with an XXE to query the cloud instance metadata service for an admin role able to run commands on thousands of instances; JSP/PHP web shells; an unauthenticated orchestration service giving root in production containers; LAPS password reads and resource-based constrained delegation; mail-server admin tokens forged with keys from a 2022 directory export that were never rotated; VPN access using configs found in mail and keys from older leaks. The agents were told to look for Windows hosts without the CrowdStrike Falcon sensor. This is separate from the Storm-0940 ARTEX/ConnectWise activity. No new CVE is named; the exact products and flaws behind the SAML/XXE and orchestration-service entry points are unknown.

Summary written from the sources below. Check the originals before acting.