Abstract: AI-generated content may be identical or substantially similar to protected works of others,raising issues of infringement determination and liability. When AI-generated content is substantially similar to another's work but exhibits originality in form,it does not qualify as a work under copyright law. However,the definition of derivative rights in the Copyright Law requires the creation of a derivative work,leading to a paradox in determining infringement of such rights. Based on the principle of separability between infringement standards and rights-generation standards,this paradox can be resolved by interpreting the term “work” in the definition of derivative rights to include “other content with the appearance of a work”. Furthermore,when AI generates content identical or substantially similar to protected works,it is not the result of acts regulated by exclusive rights carried out by developers,operators,or users,creating another paradox:the harm exists without an identifiable infringer. This can be addressed by classifying AI as a “product”,designating developers and operators as “producers”,and applying product liability rules under a reasonable interpretation of “product defects”.