Fair Use of Training Data in Generative Artificial Intelligence

Generative AI data training poses novel challenges to the doctrine of fair use in copyright law. While the "three-step test" can preliminarily justify the legitimacy of AI data training as fair use, it still faces deep-seated contradictions such as the disconnect between current legal frameworks and technological advancements, an overemphasis on "rights protection",and imbalances in copyright interests. Therefore, it is imperative to reform and innovate the fair use system. First and foremost, a copyright system centered on "fair use" should be established to break free from the constraints of "author centrism." In addition, a specific clause for "AI learning and creation" should be introduced to explicitly recognize data training as fair use. Furthermore, by integrating the four-factor test from U.S. copyright law, a general standard for assessing fair use should be established to balance copyright protection with technological development, thereby advancing the adaptation of copyright law to the demands of the digital era.