Executive Summary
Between September and October 2026, a China-linked threat actor conducted a large-scale AI-driven offensive campaign targeting government ministries and critical industry sectors in Vietnam, Taiwan, Indonesia, and India. The operation utilized the Cairn autonomous exploitation framework, which dispatches LLM-based agents in Kali Linux containers to perform automated reconnaissance and exploitation. The campaign was discovered due to an operational security failure where the operator exposed their evidence directory on an unauthenticated server.
Technical artifacts indicate the operator is a Chinese speaker using Simplified Chinese in scripts and comments, though their primary motive appears financial rather than state-sponsored. Despite this, the victimology includes high-value national security targets such as Vietnam’s Ministry of Defence, Taiwan’s Legislative Yuan, and Indonesia’s State Intelligence Agency. The operator employed a sophisticated LLM backend, using a farm of 'CodeBuddy' identities and multiple subscription accounts to process over 36 billion tokens during the campaign.
This activity highlights a significant shift toward AI-automated tradecraft, where agents are tasked with reaching 'webshell or RCE' objectives while maintaining a 'highest red line' of suppressing non-English language signals on victim systems. The scale and speed of these intrusions—achieving access in 48 of 66 projects—demonstrate the growing threat posed by machine-accelerated exploitation against both legacy vulnerabilities and modern cloud environments.