Artificial intelligence model training(machine learning)infringement is an important issue that needs to be addressed urgently in AI-related copyright conflicts.Expanding fair use system and enhancing remedy of damages cannot effectively solve the problem of imbalance of social interests,while statutory licensing remains essential for reconciling these interests.Statutory licensing for machine learning shall be limited to copying,commercial data mining must comply with the statutory licensing,and fair use shall only apply to cases where public interest significantly outweighs commercial use.Statutory licensing fees can be set with reference to the method of determining royalties for damages,and its collection and allocation should still be carried out through copyright collective management organizations.It is necessary to improve the information labeling obligations of users,and to improve the statutory licensing information mechanism for AI.In the special case of orphan works for machine learning,the dual approach of liability limitation and statutory licensing could be adopted.AI developers have the duty of care for data filtering and information disclosure,the latter should follow the principles of mandatory disclosure and fairness and reasonableness;AI service providers,based on safe harbor rules,must undertake obligations,among others,the formal review of information disclosure by developers.
machine learningstatutory licensingsocial interestsorphan worksduty of care
万方数据
维普
