China’s AI Models Take the Open-Source Route: Building Influence Between Closed and Open Approaches

After listening to this episode of Keji Lang, I see China trying to build a different kind of open-source influence from that of the United States in the AI era. China still trails the United States in AI models. Hence Anthropic’s claim that Chinese AI model developers have conducted distillation attacks against it. This can also be seen as Chinese AI companies trying every way to break through under comprehensive U.S. restrictions, including legal, illegal—violating terms of service—and gray-area approaches. But how long will their models remain behind? Could they eventually overtake the leaders, or remain very close technologically—sometimes a trick loses its mystery once it is revealed—while creating distinctive advantages elsewhere? Chinese AI model providers are concentrating their efforts on the open-source world. Even in the United States, it appears that many AI developers choose Chinese providers’ open-source models to develop products and business applications within open-source communities. Developers in the AI world have repeatedly been warned not to build products or business models on major providers’ AI models with only a wrapper and a little UI/UX optimization. Such a thin layer of added value can be replaced in an instant. Will Chinese AI model providers move from open source to closed approaches in the future? Unknown. For now, major U.S. model providers are choosing closed approaches, while Chinese providers are taking the open-source, open-access route to build industry visibility and influence—and are therefore welcomed and supported by AI developers worldwide.