Silent Lynx Used an AI Agent to Scale Cyber Espionage Against Tajikistan

A threat group tracked as Silent Lynx used an AI-assisted operating system to support espionage against Tajik government and transportation targets, according to research from CloudSEK.

The researchers found exposed attacker infrastructure that appeared to contain material stolen from a Tajik railway environment, including thousands of documents. CloudSEK assessed the Silent Lynx attribution with medium confidence, meaning the evidence is meaningful but not conclusive.

The unusual element was a messaging-based interface connected to a DeepSeek-powered agent. CloudSEK said the setup helped the operators perform reconnaissance, search collected data, organize results and remove artifacts after access had been established.

This does not mean an AI model independently breached the railway. The available evidence points to a human-directed campaign in which an agent accelerated tasks that normally require repeated commands and manual review. That distinction matters as vendors and researchers debate what qualifies as an autonomous cyberattack.

Even with human oversight, agentic tooling changes the economics of espionage. Operators can process larger document collections, translate material, generate queries and maintain multiple victim workflows with fewer people. The advantage is particularly useful for campaigns spanning government agencies, infrastructure operators and regional partners.

For defenders, the case reinforces the need to monitor what happens after initial access. Unusual archive creation, large document searches, scripted cleanup and outbound connections to AI services may reveal the collection stage even when the original intrusion is missed.

CloudSEK summarized the investigation in its research updates. Organizations in government and transportation should review privileged access, segment document repositories and alert on high-volume access patterns that do not match a user’s normal role.