- The Chinese government is reportedly working with Alibaba, ByteDance, and Z.ai to consider restricting international access to certain advanced AI models, marking a potential shift in the country’s open-source AI strategy.
- The author argues that it would be a mistake for the US to respond by banning the use of Chinese AI.
- Instead, the US and its allies should invest in developing open-weight models and release them regularly to build trust within the developer community.
- The article explains that open-weight models disclose their “weights,” allowing users to download, run them on private infrastructure, and fine-tune them as needed—unlike closed models like OpenAI’s GPT, which only allow access via online services.
- According to the author, Chinese open-weight models like Qwen and Kimi are expanding rapidly in the international market. Alibaba states that Qwen is currently the most downloaded AI model family in the world and has been used to build over 100,000 specialized models.
- Data cited from OpenRouter shows that Chinese open-weight models accounted for only about 1% of weekly text processing traffic in 2024, rising to over 20% in some weeks of 2025 and exceeding 40% in some weeks this spring.
- The article cites examples of many US businesses using Chinese AI, such as Airbnb deploying Qwen for customer care assistants, Cursor building Composer 2 on the Moonshot platform, and DoorDash using Kimi for basic source code evaluation tasks.
- According to the author, China’s strategy is to release AI models for free to attract global developers, thereby promoting its hardware ecosystem and technological standards, similar to how it previously expanded influence in solar energy, telecommunications, and rare earths.
- The article argues that once a business has built a system based on an open-weight model, switching to another is not simple due to data, infrastructure, and deployment costs.
- The author proposes that the US government fund the development of an advanced open-weight AI model, commit to long-term public availability, and support chips, storage infrastructure, and computing power through financial agencies like the Development Finance Corporation and the Export-Import Bank.
- The article concludes that the US’s advantage should not rely on banning Chinese AI but on building a more attractive AI ecosystem so that global developers choose it voluntarily.
📌 If China begins to restrict advanced AI models, the US should not respond with bans but must accelerate the development of its own open-weight AI ecosystem. According to the author, AI success depends not only on the model but also on creating an open, stable platform accompanied by computing infrastructure to attract developers worldwide, thereby maintaining a long-term technological competitive advantage.
