Unit 42 Ties DeepSeek Agent to 460+ Autonomous Hack Attempts

Palo Alto Networks' Unit 42 has identified a Chinese hacker operator leveraging the open-source DeepSeek AI agent integrated with Hermes Agent to autonomously conduct over 460 hacking attempts, confirming three successful compromises targeting Citrix NetScaler vulnerabilities. The research highlights the efficacy of vendor-side AI safety controls as both OpenAI and Claude Code blocked offensive requests, unlike DeepSeek which operated unchecked through an open framework.

This Wire brief sits within Fusion42's coverage of Cybersecurity.