The emergence of the innovative AI model DeepSeek, a low-cost and high-efficiency model, is having a significant impact on the global AI industry. By utilizing technologies such as the Mixture of Experts, FP8, and Knowledge Distillation, DeepSeek has successfully reduced hardware costs and development time while demonstrating performance comparable to that of leading models. Following DeepSeek’s success, interest in knowledge distillation for developing cost-efficient AI models has surged. Knowledge distillation is a technique that transfers knowledge from a large-scale “teacher” model to a smaller “student” model, contributing to model lightweighting and faster inference speeds. However, this process has intensified intellectual property disputes. Leading companies like OpenAI and Google have accused DeepSeek of intellectual property theft, claiming it illegally trained on their models, and have raised the possibility of legal action. Previously, IP issues in AI development primarily focused on copyright infringement during the training process, but with fast-followers using knowledge distillation on models from leading groups, the issue is now expanding to data misappropriation. As knowledge distillation becomes a crucial tool for latecomers to secure a competitive edge, it also necessitates a review of whether such actions constitute IP infringement. This paper examines the allegations of DeepSeek’s data misappropriation from an intellectual property law perspective, given the intensifying conflict among global AI developers surrounding knowledge distillation spurred by the DeepSeek case.