Research
Free Chinese AI Models Are Getting Scary Good at Hacking
WIRED · By Will Knight & Lily Hay Newman ·
The Brief
- An experiment asked AI models to break into simulated networks.
- DeepSeek and GLM models showed hacking skills not far behind OpenAI's most advanced cyber model, which is not public.
- The source describes the free Chinese models as approaching frontier models in this test.
Why it matters
Openly available models are getting closer to frontier models at offensive cyber tasks, which matters for anyone assessing AI-related risk to their systems.
Who’s affected
- Security teams assessing AI-driven attack risk
- #ai
- #cybersecurity
- #llm
- #hacking
- #deepseek
- #benchmarks
An experiment in which AI models were asked to break into simulated networks showed the hacking skills of DeepSeek and GLM to be not far behind OpenAI’s most advanced cyber model, which isn’t public.