Chinese artificial intelligence models are rapidly transitioning from mere competitors to core infrastructure for technology companies worldwide. This shift is creating a complex policy paradox for the United States, as American regulators attempt to erect guardrails around domestic frontier models while confronting a robust, open-source ecosystem emerging outside their jurisdictional reach, as Paul Triolo argues in a new Rest of World analysis.
Recent developments highlight the deep integration of Chinese models into global tech products. Mira Murati’s newly launched startup, Thinking Machines, which recently raised $2 billion, revealed that its first foundation model, “Inkling,” was developed using synthetic data generated by DeepSeek-V3 and Moonshot AI’s Kimi K2.5.
Similarly, Apple recently received approval from the Cyberspace Administration of China (CAC) to launch Apple Intelligence in the Chinese market utilizing models from Alibaba and Baidu, both of which have been designated by the U.S. Department of Defense as “Chinese military affiliated companies.”
The adoption of Chinese models by prominent Western companies underscores the growing competitiveness of China’s AI sector. Despite stringent U.S. export controls on advanced semiconductors, Chinese firms have developed highly capable models that rival their American counterparts in performance and efficiency. This success is largely attributed to innovative algorithmic approaches, efficient data utilization, and a strong emphasis on open-source development.
The Rise of Chinese Open-Weight Models
Paul Triolo, partner and global tech policy lead at DGA-Albright Stonebridge Group, noted that Chinese frontier models are no longer just subjects for evaluation; they are foundational tools that global developers are actively building upon. The open-weight strategy adopted by many Chinese AI firms has made their models highly accessible, particularly to developers in the Global South.
Recent data indicates that a significant portion of global AI traffic is now being routed through Chinese systems. U.S. companies now route 46% of AI tokens through Chinese models as OpenAI costs surge, according to data from the OpenRouter platform. The recent release of Zhipu’s GLM-5.2, which has been made openly available, further cements China’s position as a primary provider of accessible, high-performance AI infrastructure.
The appeal of Chinese open-weight models lies not only in their performance but also in their cost-effectiveness. For startups and developers operating with limited budgets, the ability to access powerful AI capabilities without paying exorbitant API fees is a significant advantage. This dynamic is driving widespread adoption of Chinese models across various industries, from healthcare and education to finance and manufacturing.
The U.S. Policy Paradox
The widespread adoption of Chinese models presents a significant challenge for U.S. policymakers. While the administration considers imposing restrictions or bans on Chinese open-weight models to protect American intellectual property, such measures could have far-reaching consequences for developers in regions like Southeast Asia, Africa, and Latin America who rely on these accessible tools.
The situation underscores the difficulty of containing AI proliferation in an interconnected global ecosystem. As Chinese models become the de facto open-source infrastructure for many international developers, U.S. efforts to restrict their use may inadvertently isolate American companies and accelerate the development of parallel technological ecosystems outside of Western influence.
Furthermore, attempting to ban or restrict the use of open-weight models is technically challenging. Once a model’s weights are released publicly, it can be downloaded, modified, and deployed by anyone with sufficient computing resources. This reality makes it nearly impossible to completely prevent the proliferation of Chinese AI technology, forcing U.S. policymakers to grapple with the limitations of traditional export controls in the digital age.
Navigating a Bifurcated Future
The growing reliance on Chinese AI infrastructure highlights the potential for a bifurcated global technology landscape. As the U.S. and China continue to compete for dominance in artificial intelligence, developers and companies around the world may increasingly find themselves forced to choose between competing ecosystems.
For Western tech companies, navigating this complex environment will require careful strategic planning. While using Chinese models may offer short-term cost and performance advantages, it also carries significant geopolitical and regulatory risks. As tensions between Washington and Beijing continue to escalate, companies must be prepared to adapt to a rapidly evolving policy landscape that could fundamentally alter the rules of global technological engagement.