Editor's note: Lin G. is a CGTN economic commentator. This article represents the author's personal views, and not those of CGTN.

At the latest China-US leaders' meeting, the two sides reached an eight-point consensus, with the seventh point focusing specifically on artificial intelligence (AI) governance and communication. They agreed to establish a dialogue on AI to exchange views on AI-related risks and benefits, with the next round to be held in November, and to set up a bilateral communication channel for AI-related incidents.

This is a positive development for global AI governance. It gives the world's two largest AI powers a channel to communicate and manage differences as the technology develops at an unprecedented speed. But this dialogue is more than an act of goodwill. It is also the result of a hard-won shift in the balance of technological capabilities between China and the United States.

A choice shaped by a new balance of power

In the early stage of global AI competition, the US enjoyed a clear first-mover advantage in foundational algorithms, high-end graphics processing units, core software and top-tier research. Washington's strategy was equally clear: Preserve its technological lead and constrain China's progress through increasingly comprehensive restrictions.

Since 2022, the US has repeatedly tightened export controls on advanced computing chips, semiconductor manufacturing equipment and electronic design automation tools, seeking to limit China's access to cutting-edge computing power and critical industrial software. The underlying assumption was that such measures could hold back China's AI development and preserve America's technological dominance.

Years of restrictions, however, have not produced that outcome. Confronted with growing uncertainty over external supplies, China's AI industry has pursued a different path, placing greater emphasis on architectural innovation and computing efficiency rather than simply adding more high-end chips. DeepSeek-V3, for example, uses a mixture-of-experts architecture and multi-head latent attention, activating only a fraction of its parameters for each task and substantially reducing training and inference costs.

According to the 2026 Stanford AI Index, the performance gap between leading Chinese and US models has largely closed, with the two sides developing distinct strengths in AI competition.

The story goes beyond model development. China has also built a broad industrial ecosystem for AI, supported by extensive ultra-high-voltage power infrastructure, abundant renewable energy and the world's largest manufacturing base. China accounts for 54% of global industrial robot installations, while AI applications in smart factories, power-grid management and other industrial settings have moved from experimentation to large-scale deployment.

In semiconductors, restrictions on extreme ultraviolet lithography and advanced manufacturing processes have pushed Chinese companies to pursue breakthroughs across the broader computing system rather than compete solely at the individual-chip level. Meanwhile, the domestic substitution rate for semiconductor equipment has reportedly doubled over four years, and even some overseas manufacturers have begun testing Chinese-made equipment.

As China's AI capabilities have expanded across models, computing infrastructure and industrial applications, Washington has had to confront a changing reality: Unilateral restrictions have not halted China's technological upgrading, while full-scale industrial decoupling would carry significant economic costs, including for US companies.

The balance of technological capabilities has therefore become harder to ignore. Against this backdrop, the US has shown greater willingness to move away from an exclusively zero-sum approach and toward bilateral AI dialogue and mechanisms for managing technological risks.

China's cooperation vision: A shared future for mankind

China's approach to AI cooperation starts from a practical understanding of the technology's global impact. Artificial intelligence can deliver enormous productivity gains, but it also brings challenges that transcend national borders, from algorithmic bias and misinformation to system safety and rising energy demand. No country can address all of these risks on its own.

For China, that makes international cooperation not an afterthought, but an essential part of how AI should develop. This thinking is reflected in China's long-standing support for more open and accessible AI development.

Research institutions and technology companies in China have released open-source model weights, tools and technical documentation that developers around the world can use, adapt and improve. Compared with many closed and high-cost commercial solutions, open-source models and localized deployment options can lower the barriers to adopting AI, particularly for emerging and developing economies. This enables more countries to apply and adapt AI to their own needs, whether for industrial upgrading, public services or broader digital transformation.

At the 2026 World AI Conference, China called for greater openness, greater awareness of risks, inclusive development and closer international coordination on AI governance. It also proposed helping Global South countries strengthen their AI capabilities and announced plans to establish international AI application cooperation centers with ASEAN, the Arab League, the African Union, the Community of Latin American and Caribbean States, the Shanghai Cooperation Organization and BRICS nations.

The logic is straightforward: AI's benefits should reach beyond a small group of countries and companies. Yet this broader accessibility also has consequences for the global market. As affordable and increasingly capable Chinese AI products and services become available to more governments and businesses, they give users more choices in a market long dominated by US technology companies.

From Washington's perspective, this is no longer simply a question of technological development. It also means greater competition for global markets and a growing Chinese presence in the international AI ecosystem. That changing market landscape is another factor encouraging the US to engage with China through dialogue.

At the same time, China's emphasis on self-reliance should not be confused with technological isolation. Stronger domestic capabilities create the conditions for dialogue on a more equal footing. That is why China's pursuit of core technological self-improvement and its support for international cooperation can move in parallel. The China-US AI dialogue fits into this approach.

Conclusion: A hard-won opening to build on

The new China-US AI dialogue does not erase their differences, but it creates something that was previously missing: A dedicated channel for managing AI-related risks. With regular exchanges and an incident-response mechanism now in place, the two sides have a practical way to reduce the risk of miscalculation as technological competition intensifies.

This matters beyond the bilateral relationship. As AI becomes increasingly embedded in economies and societies, major powers will need mechanisms to manage disagreements without allowing technological competition to spill into broader instability. The China-US dialogue provides a concrete starting point for that effort.