Chinese Threat Actor Deploys AI Agents for Autonomous Cyberattacks on Exposed Servers

BEIJING — A Chinese-speaking threat actor operating under the aliases 'knaithe' and 'KnYuan' has successfully executed autonomous cyberattacks against exposed servers using an artificial intelligence workflow, security researchers confirmed Wednesday. The incident marks a significant escalation in offensive cybersecurity capabilities, demonstrating how AI models can be leveraged to discover, evaluate, and compromise vulnerable systems with minimal human intervention.

The attack chain utilized the DeepSeek AI model integrated with an open-source Hermes Agent framework. This combination allowed the threat actor to automate complex reconnaissance and exploitation phases targeting specific workflow automation platforms, including Langflow and n8n. The operation was attributed to a China-based entity, though no official government affiliation has been established.

The offensive campaign took place on July 31, 2026. During the incident, the AI-driven system scanned for exposed servers running vulnerable configurations of targeted software. Upon identifying potential entry points, the agent autonomously evaluated the severity of vulnerabilities and launched tailored exploits without requiring manual commands from an operator at each stage.

Security analysts note that this methodology represents a shift toward fully automated attack vectors capable of scaling rapidly across global networks. By removing the bottleneck of human decision-making during the exploitation phase, such workflows can test thousands of targets in hours rather than days. The specific targeting of Langflow and n8n highlights a strategic focus on platforms widely used for building AI applications and data pipelines.

The attack was conducted to demonstrate an offensive AI workflow capable of operating independently once initial parameters were set. This capability raises concerns about the accessibility of such tools, as open-source components like the Hermes Agent lower the technical barrier for deploying sophisticated cyber operations.

Details regarding the extent of data exfiltration or financial impact remain unclear following the incident. While the threat actor successfully demonstrated the ability to compromise exposed systems, it is not yet known if specific organizations suffered long-term damage beyond initial access points. The use of Chinese-language commands within the attack scripts has drawn attention from international cybersecurity teams monitoring state-aligned activities.

As AI technology continues to evolve, the line between defensive automation and offensive capability narrows. Experts warn that similar workflows could be adapted by other actors globally if the underlying code or methodology becomes widely available. Questions remain regarding whether this incident was a one-time demonstration of technical prowess or part of a broader campaign targeting critical infrastructure.

Cybersecurity firms are currently updating detection signatures to identify traffic patterns associated with AI-driven reconnaissance and exploitation attempts. The incident underscores an urgent need for organizations to secure exposed servers and patch known vulnerabilities in automation platforms before they can be leveraged by autonomous agents.