Free Chinese AI Models Are Getting Scary Good at Hacking
Wired

Free Chinese AI models like DeepSeek and GLM have demonstrated advanced hacking skills that rival leading private models.
Recent experiments testing AI models on simulated networks reveal that free Chinese models, including DeepSeek and GLM, possess hacking capabilities nearly matching OpenAI's advanced cyber model. This indicates that competitive cybersecurity capabilities are becoming accessible through public and open models. The findings highlight the rapidly narrowing gap between public AI tools and proprietary top-tier models in sensitive technical domains.
Why it matters: The emergence of highly capable hacking skills in accessible Chinese AI models demonstrates that advanced cyber capabilities are no longer restricted to exclusive proprietary systems.





