CKN Knowledge Session - Cutting Edge: Governing the ‘Ungovernable’ Generative AI in China
In this talk, Dr He shares insights from his current research on how Chinese large language models (LLMs) like DeepSeek and Wenxin Yiyan handle politically sensitive prompts. By collecting and analyzing prompts that the Chinese LLM DeepSeek refuses to respond to—and comparing how similar prompts are treated across multiple Chinese models—this study offers a rare empirical glimpse into the boundaries of content moderation in generative AI.