The in-depth application of artificial intelligence technology in the fields of literature, art and science has brought our human society a large amount of excellent contents that are indistinguishable from human works in appearance. These artificial intelligence generated contents are not only an important part of social spiritual culture, but also have certain economic value. If such artificial contents are not effectively protected, they will only be reduced to public resources for people to use arbitrarily. In this case, due to lack of incentives, the enthusiasm of the stakeholders behind artificial intelligence to invest will be severely frustrated. Over time, it will not only be detrimental to the healthy development of the artificial intelligence industry, but also affect the development of social and cultural diversity. In recent years, domestic disputes over the copyright protection of artificial intelligence contents have appeared and the domestic judicial practice has encountered serious challenges in how to solve the contradiction between intelligent technology and legal lag in the new era. The problem needs to be clarified and solved urgently. However, given the particularity of the way in which AI-generated contents are created, how to effectively protect them under the current copyright law has become extremely difficult. Whether artificial intelligence generated contents can be works in the sense of copyright law, whether affirming the creation subject status of artificial intelligence generated contents will cause damage to the traditional theory of copyright subjects, how to confirm the ownership of artificial intelligence generated contents, and how to choose the copyright protection mode of them, these are all the issues that have caused great controversy in the theoretical and practical circles and need to be solved urgently. This paper defines the concepts of artificial intelligence and artificial intelligence generated contents, and divide all the contents into expert system generated contents based on preset rules and machines learning generated contents by using neural networks according to two different modes of generation in the technical perspective to discuss separately. On the basis of analyzing the dilemma of domestic judicial practice, this paper focuses on the research on the feasibility of copyright protection of artificial intelligence generated contents from the perspective of its work eligibility, the disputes of its copyright subjects, the comparative analysis of the cost and expected social effects of copyright protection for artificial intelligence generated contents. After demonstrating the feasibility, this paper will further proposes a path for the copyright protection of these contents. More specifically, this paper will set some reasonable limits on the subject, object, specific rights protection contents and protection period of artificial intelligence generated contents which can be regarded as works based on the copyright legal system, and then make arrangements for the ownership of the copyright and infringement liability of the contents. In addition, the paper suggest that those contents which cannot be works should be protected by the neighboring rights. Finally, in order to safeguard the copyright protection of artificial intelligence contents and regulate the cultural market more effectively, this paper will propose to give full play to the leading role of the government and establish a special registration database for artificial intelligence contents.