From Decentralized Licensing to Centralized Governance: Optimizing the Statutory Licensing Regime for Artificial Intelligence Training Behavior

The unauthorized use of other people's works that are still under copyright protection for training of generative artificial intelligence raises a huge risk of copyright infringement. As an important system to cope with the infringement risk of generative AI training by unauthorized use of other people's works within the protection period, the quandary of searching for the right subject and the quandary of payment of exorbitant remuneration triggered by the characteristics of large models training by using massive works have impeded the expansion of the traditional statutory license system built on the underlying framework of decentralized licensing; and the ambiguity of the value orientation of the statutory license system has led to the construction of the system. The ambiguity of the value orientation of the statutory licensing system leads to the wavering direction of the system construction, and the application of the statutory licensing system to the infringement risk management of the training of large models of artificial intelligence will thus become more difficult. The key to solving the problem lies in the conversion of the analytical idea of cracking the AI infringement risk predicament: under the concept of positive-sum game, starting from the attribute of "industrial law" of the copyright law, and following the market value orientation of efficiency, combining with the demand of AI for the utilization of the trained works, the construction of a "centralized governance" model is based on the "centralized governance" model, and the "centralized management" model is based on the "centralized management" model. Based on the "centralized governance" model, a statutory licensing system for "large-scale work database" is constructed. With the help of this underlying structure, under the "collective calculation standard" of the license fee, the upper limit of the calculation standard of the statutory license fee can be clarified to prevent the emergence of "sky-high license fee". At the same time, collective licensing can also reduce the high cost of seeking licenses from all right holders one by one, so as to effectively deal with the problem of identifying right holders when the statutory licensing system responds to the training behavior of artificial intelligence. In addition, the arrangement of the statutory licensing system for "large-scale work database" can also give full play to the advantages of the volume of works and the governance capacity of "large-scale work database", and improve the ability of copyright holders to obtain royalties by confirming and tracing the legitimacy of the training data of large-scale models. While confirming and tracing the legitimacy of large model training data, the arrangement of the statutory licensing system can also enhance the protection of the ability of copyright owners to obtain royalties, so as to accurately respond to the development needs of the industry and at the same time promote the realization of the goal of win-win situation for copyright owners and the AI industry, and help the long-term development of the copyright ecological industry.