Free Chinese AI Models Are Getting Scary Good at Hacking

An experiment showed that the free Chinese AI models DeepSeek and GLM demonstrated hacking skills close to OpenAI's most advanced non-public cyber model when attempting to break into simulated networks.