The artificial intelligence content generation mechanism covers the acquisition of data in the research and development stage and the application of generated content in the subsequent utilization stage.In the former stage,it mainly faces the problem of legal authorization of copyright for the acquired data,while in the latter stage,it is mainly oriented to the problems of judgement of copyright attributes of the generated content,attribution of the rights,and assumption of re-sponsibility for copyright infringement.The existing normative analysis framework for the two stages of the main problems faced by the rules of local mismatch,the root cause is that the existing normative design can no longer meet the needs of in-dustrial security brought about by the development of artificial intelligence,and can not effectively respond to the adjust-ment of the development of artificial intelligence industrial policy.The change of AI content generation mechanism driven by technology directly impacts the existing copyright system's recognition of the underlying logic of work expression and"di-chotomy of Idea and Expression",and at the same time,we also face the dilemma of the discrepancy between the property rules for prior authorization and the needs of the learning model for massive resources,the risk of copyright infringement during the whole phase of machine learning content acquisition,and the impracticality of requiring only copyright compli-ance due to the diversity and complexity of data protection interests.In the face of these problems,instead of designing rules in a patchwork manner,we should try to solve the systematic cognitive problems in a comprehensive manner,and try to design a system that separates authorship and other copyrights on the basis of the solid"dichotomy of Idea and Expres-sion",in order to implement the principle of honesty and trust to ensure the authenticity of the source.The"question of development"between the real world and the evolution of technology can be solved through diversified solutions,such as le-gal purchase and contractual risk-bearing,opening the floodgates of copyright fair use for data acquisition in the pre-training stage of AI and exemption from liability through the safe harbor rule,centralized authorization by collective manage-ment organizations,and the establishment of open authorized data resources.Adjustment of the normative framework and breakthrough in the interpretation of rules should be carried out according to local conditions,so as to find the best perspec-tive to adapt to the development of the industry and the safeguard of normative measures for technological upgrading.
万方数据
维普
