CHINESE STARTUP DEEKSEEK AI UPENDED THE CONVENTIONAL WISDOM ABOUT AI INNOVATION. When it released its R1 language model and V3 general-purpose large language model (LLM) in January 2025, which demonstrated unprecedented reasoning capabilities, the company sent tremors through markets and challenged assumptions about American technological superiority. Beyond debates about DeepSeek’s computation costs, the company’s breakthroughs speak to critical shifts in the ongoing global competition for AI talent. In our paper, “A Deep Peek into DeepSeek AI’s Talent and Implications for US Innovation,” we detail the educational backgrounds, career paths, and international mobility of more than 200 DeepSeek researchers. Nearly all of these researchers were educated or trained in China, more than half never left China for schooling or work, and of the nearly quarter that did gain some experience in the United States, most returned to China. 

Policymakers should recognize these talent patterns as a serious challenge to U.S. technological leadership that export controls and Policy Implications of DeepSeek AI’s Talent Base computing investments alone cannot fully address. The success of DeepSeek should act as an earlywarning signal that human capital—not just hardware or algorithms—plays a crucial role in geopolitics and that America’s talent advantage is diminishing.