AI development should be innovative, inclusive and cooperative, not zero-sum

There is now a division in the artificial intelligence development track. Closed-source or open-weight? The choice of AI model is really a choice between two opposing paradigms — on innovation, regulation and who gets a share of the AI pie.

The closed-source model seeks to build technical barriers relying on exclusive model weights, locking the dividends of AI progress within a handful of enterprises. Companies such as OpenAI and Anthropic have adopted this path. While this model may deliver quick profits, it shrinks room for innovation and weakens competition.

On the other hand, the core value of open-weight models lies in inclusiveness and cooperation. For the industry, such models can break technological monopolies, and make AI more accessible to small and medium-sized businesses, unleashing innovation across different sectors. Most Chinese AI models are open-weight, an approach that is a major driver of AI development in China.

This has led some United States policymakers to tighten oversight of foreign open-weight AI models after Kimi K3 was launched this month. Some have even proposed requiring built-in “kill switches” so that the US government can shut down these systems remotely if they are considered a “national security risk”.

A group of 25 US high-tech enterprises, including Nvidia, Microsoft, IBM and Meta published a joint letter on Friday calling on the US administration not to impose restrictions on open-weight AI models. The letter stresses that open-weight AI models are safer than closed ones, and keeping AI models open is essential for the US’ national security, economic competitiveness and innovation.

The collective call for openness and transparency in AI development in the letter reflects a rational assessment of the industrial reality. There is growing consensus among major stakeholders that the long-term competitiveness of AI stems from a thriving open ecosystem, rather than technological monopolies of certain entities, as technological restrictions will inevitably harm the AI ecosystem and an insular approach hinders technological breakthroughs and industrial upgrading.

But for a long time, some US politicians have viewed AI advancements as exclusive competition, attempting to erect technical barriers through tech blockades and access restrictions. Yet facts have proven again that scientific and technological progress never follows a zero-sum logic.

In reality, Chinese AI breakthroughs have delivered tangible benefits to US companies, primarily through cost savings and open-source accessibility.

Data cited by CNBC show that the share of tokens used by US companies on Chinese AI models through OpenRouter — a platform that gives developers access to a wide range of models — has remained above 30 percent each week since February, reaching as high as 46 percent. Chinese open- weight models — which offer competitive capability at roughly one-fifth the cost — have become a practical choice for US businesses.

As AI models integrate ever faster into diverse industries, daily life and social systems, competition is no longer defined solely by raw technical performance, but hinges on ecosystems. Once a large and vibrant ecosystem takes root, it fuels faster, broader technological iterations through real-world feedback and data. Thus, the path to sustainable AI development lies not in hoarding proprietary advantages, but in nurturing open, collaborative ecosystems that accelerate collective progress.

This is precisely where China’s AI development path finds its footing. Guided by the principles of openness and security, it systematically advances the AI Plus initiative across critical domains.

China’s AI policy is designed to ensure that the dividends of AI innovation are shared broadly across society and contribute to the collective interests of different countries.

For AI to truly serve the common good of humanity, its global development must move beyond zero-sum competition and toward inclusive cooperation.

Editorial, China Daily