China's AI Ambition: Mythos-Class Models by 2027?

China's Zai chief suggests the country could launch its own "Mythos-class" AI models sooner than expected. This comes after the US government banned Mythos 5 and Fable 5 for foreigners, citing concerns about China-linked groups exploiting these models. Zai's founder, Tang Jie, believes China could have a similar AI model before 2027. Tech billionaire Elon Musk predicted this might happen by the first quarter of 2027, but Tang Jie responded, "Won't take that long." What's interesting is that Mythos-class models are reportedly powerful enough to hack software and have found tens of thousands of bugs. Zai's latest model, GLM 5.2, is already a leading open-weights model, outperforming Google's Gemini 3.5 Flash and Anthropic's Claude Sonnet 4.6 on a benchmark index. The bottom line is that if China develops its own advanced AI models, the US ban on certain models to prevent their use by China may not be effective in the long run.