Chinese open-weight AI models have emerged as a low-cost alternative to proprietary US models, with companies including Coinbase, Shopify, and Airbnb using models such as GLM, Qwen, and Kimi. On OpenRouter, the share of Chinese open-weight models rose from 20% in September last year to 55% as of last month. Alibaba’s Qwen surpassed 700 million cumulative downloads on HuggingFace earlier this year, overtaking Meta’s Llama, and the Qwen3 family alone generated more monthly downloads than OpenAI, Mistral, NVIDIA, Zhipu, Moonshot, and MiniMax combined. Discussions held in May suggest legal experts are proposing a tiered regulatory framework rather than a blanket restriction on open-source AI. Basic open-weight models would remain accessible under filing or registration requirements, while increasingly capable models would undergo greater security scrutiny, and the most advanced frontier models would face the strictest controls. The released documents emphasise balancing openness with security, warning that once frontier model weights are publicly released, they cannot be recalled. Analysts suggest China’s policy may be a response to US treatment of frontier AI models as export-controlled assets.