DeepSeek Ran Autonomous Cyberattacks That Claude and OpenAI Safety Controls Blocked

A Chinese threat actor used the open-source Hermes Agent platform combined with DeepSeek, a Chinese AI model lacking safety controls, to autonomously perform cyberattacks, while safer Western AI models like Claude and OpenAI refused such requests. Palo Alto Networks' Unit 42 report demonstrates the operational impact of AI safety controls, with server-side protections blocking misuse and open-source models enabling unrestrained offensive capabilities.

This Wire brief sits within Fusion42's coverage of Cybersecurity, and 2 sources have reported it between 31 Jul 2026 and 1 Aug 2026.