Copyright Policy Options for Generative Artificial Intelligence
从经济学角度分析不同版权制度对生成式AI内容质量和训练的影响，指出由于交易成本高，事后机制如合理使用比传统版权制度更能提升社会福利。
New generative artificial intelligence (AI) models have created new challenges for copyright policy as such models may be trained on data that include copy-protected content. This paper examines this issue from an economic perspective and analyzes how different copyright regimes for generative AI will impact the quality of content generated and AI training. Because of transaction costs (for example, because of the large amount of content being used to train generative AI models), it is not possible for copyright holders and AI providers to engage in negotiations. The result is a characterization of the factors that would favor full copyright and no copyright protections, balancing the level of potential harm to original content providers and the importance of content for AI training quality. However, it is demonstrated that an ex post mechanism like fair use can lead to higher expected social welfare than traditional rights regimes.