China's Rapid AI Progress Defies U.S. Export Controls

In 2017, Beijing announced an ambitious plan to lead the world in artificial intelligence by 2030. The plan anticipated significant progress by 2020, but in 2022, OpenAI's ChatGPT caught China off-guard. At the time, China's tech companies were recovering from a government crackdown, and the country's AI chatbots awaited approval for public use. Concerns arose that censorship and new U. S. export controls, which limited China's access to advanced semiconductors essential for large AI models, might hinder China's AI ambitions. Despite these obstacles, China's AI capabilities have advanced rapidly. Notable developments in 2024 include Alibaba and DeepSeek releasing reasoning models comparable to OpenAI's technology, Tencent launching Hunyuan-Large, and DeepSeek's AI model topping an online leaderboard. These achievements have even surprised former Google CEO Eric Schmidt, who underestimated the effectiveness of U. S export controls, originally implemented to limit China's tech growth. AI leadership influences global power dynamics since AI can enhance economic and military capabilities. While AI advancements rely on data, algorithms, and computing power, the latter remains a central concern due to the scarcity and geopolitical sensitivity of advanced semiconductors. The U. S.

dominates the chip manufacturing process, leveraging export controls to hinder China's access to the most advanced chips. However, these controls face challenges as Chinese developers have stockpiled chips, skirted restrictions through smuggling, or circumvented controls by using offshore resources. Although the controls aim to slow China's AI progress, their full impact on China's AI development is yet to be seen, especially as China utilizes less powerful chips and better software. The gap between U. S. and Chinese AI capabilities is narrowing. Yet, the U. S. maintains an edge with more advanced proprietary models. Ongoing efforts to enforce stringent export controls might continue to challenge China's AI scale and deployment. With advancements in compute power expected, U. S. export controls could further hinder China's progress if compute remains crucial. Inside Washington, there is reluctance to negotiate AI terms with China, but experts suggest dialogue could mitigate potential risks from AI capable of significant harm. Given China's advances despite restrictions, discussions around ensuring the security of AI systems on both sides could prove essential.