DeepSeek AI Enters The Hacker’s Toolkit As Threat Actor Automates Offensive Operations

A Chinese-speaking threat actor has been observed using DeepSeek AI as an autonomous offensive operator to identify exposed infrastructure, research vulnerabilities, obtain public exploits and launch attacks with minimal human involvement.

Palo Alto Networks Unit 42 tracked the activity to an actor known as knaithe and KnYuan.

Researchers described it as an early but functional example of end-to-end AI-assisted cyberattack automation.

The actor primarily deployed DeepSeek through the open-source Hermes Agent framework.

The setup provided terminal access, a skills system and Telegram-based command-and-control capabilities.

DeepSeek served as the reasoning engine for target selection, code generation, vulnerability assessment and follow-up actions.

The threat actor customised Hermes Agent with red-team capabilities, including a ‘godmode’ jailbreak feature, an unauthenticated WebSocket exploitation workflow and a FOFA asset-search tool.

An MCP server enabled FOFA searches, Nuclei scan generation and natural-language conversion into FOFA queries.

Investigators found configurations for Qwen, GLM, Kimi and MiniMax, while Claude Code and Codex had more limited roles.

In May 2026, DeepSeek targeted a critical Langflow vulnerability, CVE-2026-33017. It found 84 exposed instances but failed to exploit the selected system.

The AI then assessed Langflow as low value and shifted towards n8n, identifying more than 647,000 exposed instances.

Although the automated attacks produced no confirmed compromise, Unit 42 linked the same actor to successful manual operations. These included data theft from three Citrix NetScaler targets and command execution on 11 Marimo notebook instances.

The actor targeted more than 460 systems using automated and manual methods.

Also Read: Pixel 11 In India: Is Google Betting Big On AI And Local Manufacturing?