Rapid Adoption of Chinese AI Models Poses National Security Risks

By the end of 2025, Chinese AI models surged to 30% of global workloads, with Alibaba’s Qwen family leading at over 700 million downloads. These models, while open-source, are developed under China's National Intelligence Law, raising concerns about potential intelligence collection and supply chain vulnerabilities. The integration of these systems into U.S. infrastructure presents four key threats: supply chain poisoning, intelligence collection, capability uplift for malicious actors, and economic displacement. The difficulty in auditing AI software, due to its complex nature and the anonymity of developers, complicates security measures. Recent research indicates that even a small number of compromised documents can create backdoors in large language models, making detection challenging. As U.S. policymakers grapple with these risks, targeted interventions are necessary to mitigate the threats without resorting to protectionist measures similar to those of China.

Key Points: • Chinese AI models now account for 30% of global workloads, up from 1% in late 2024. • Alibaba's Qwen family is the largest provider of open-source AI systems with over 700 million downloads. • The integration of these models poses risks of supply chain poisoning and intelligence collection.