Abstract: The Interim Measures for the Management of Generative Artificial Intelligence Services and Basic Requirements for the Security of Generative Artificial Intelligence Services(TC260-003) establish legal frameworks for using prior work data in generative AI. Generative AI behavior, trained on massive prior work data, integrates prior creative expression with user input, creating copyright infringement risks. Considering the need to reduce authorization costs for generative AI providers, balance multi-stakeholder interests, and control algorithmic works' market substitutability, fair use demonstrates superior advantages over statutory licensing, emerging as the preferred legal pathway for prior work data utilization in generative AI. Due to restrictions imposed by the "three-step test" in article 24(1) of the Copyright Law, which obstructs commercial applications and excludes for-profit entities from fair use under its first twelve provisions, the article suggests breaking through the Copyright Law Implementation Regulations to permit fair use by for-profit generative AI providers; Adopting "standard-essential" licensing for commercial generative AI applications; Redefining the "three-step test" through dual perspectives of interest-balancing and technology-promotion, machine learning's impact on individual works, and algorithmic works' substitutability for human creations.