US-China AI Policy Channels: Impact on Chinese AI Adoption

US-China AI Policy Channels Formed

On 25 September, a summit in Washington between Donald Trump and Xi Jinping concluded with an agreement to extend an existing trade truce by two months. Crucially for the technology sector, the leaders also ratified the establishment of a formal US-China dialogue on artificial intelligence, building on preliminary discussions held earlier in New York by their deputies.

This new framework includes an American proposal for a direct communication channel, intended to notify each other when an AI incident escalates to a level of national security concern. As currently implemented, this channel functions as an open line between US Treasury Secretary Scott Bessent and Chinese Vice-Premier He Lifeng, rather than a standing technical body.

This policy development follows a November 2024 agreement between Xi and then-President Joe Biden, which affirmed the principle that humans, not machines, should retain ultimate decision-making authority over nuclear weapon deployment.

Distributed Nature of AI Capability

The establishment of these channels occurs against a backdrop where AI capability and associated risks differ fundamentally from traditional state-controlled threats. Unlike nuclear arsenals, which are typically held by a limited number of sovereign states, located in known facilities, and subject to established chains of command, most AI risk is not similarly contained.

Instead, AI development and deployment are highly distributed, with models and tools accessible across various platforms and developers globally. This decentralised movement of AI capabilities means that potential incidents or widespread adoption patterns may not be easily managed through bilateral state-to-state hotlines designed for conventional security threats.

Understanding this distinction is critical for crafting effective policy responses that account for the pervasive and evolving nature of AI.

Rising Market Adoption of Chinese AI Models

Evidence of this distributed nature and the global reach of AI models is apparent in market data. OpenRouter, an online marketplace that allows developers to access numerous AI models through a single interface, has observed a significant shift in usage patterns. In February, Chinese AI models constituted only a small proportion of the platform's overall traffic.

However, by September, these models had grown to represent a majority of the traffic. Furthermore, in regions OpenRouter designates as the "Global South," Chinese models now process approximately two-thirds of all tokens. This data shows a clear trend of increasing global reliance on Chinese-developed artificial intelligence systems across a range of applications and user bases.

US Companies Integrate Chinese AI Technologies

The adoption of Chinese AI models is not limited to the Global South or smaller developers; it extends to prominent American companies operating globally. For example, Airbnb, the international accommodation platform, currently employs Alibaba’s Qwen for its customer service operations.

Similarly, DoorDash, a leading food delivery service in the United States, uses Moonshot’s Kimi to handle routine tasks within its systems. The integration of these Chinese AI technologies by major US firms has drawn attention from policymakers.

Both Airbnb and DoorDash have received official letters from the US Congress, inquiring about their specific use of these foreign-developed artificial intelligence models. This development shows a growing tension between market-driven technology adoption and national policy concerns regarding the origin and security of critical AI infrastructure.

Implications for Technology and Industry Policy

The formation of formal US-China AI dialogue channels, juxtaposed with the increasing global integration of Chinese AI models by businesses, creates a complex landscape for technology and industry policy.

For decision-makers in Asia and beyond, the challenge involves balancing the operational efficiencies and performance benefits offered by accessible, high-performing AI solutions with emerging concerns over data security, intellectual property, and supply chain provenance.

The recent congressional inquiries directed at Airbnb and DoorDash indicate a heightened scrutiny of AI model origins, suggesting potential future regulatory actions or incentives for companies to re-evaluate their AI supply chains.

Businesses operating in Asia, particularly those integrating AI into their core operations, will need to closely monitor further policy developments from both the US and China, as well as the commercial availability and performance of diverse AI models, to inform their strategic technology choices in 2027.