In copyright law, the human control standard represents typically an important yet implicit element. In the context of AI technology, this control requirement has emerged as a “drawn sword”, serving as a fundamental premise in the examination and determination of the copyrightability of AI-generated content (AIGC), and playing a significant role, as evidenced by recent judicial cases in China. In scenarios involving random factors in creation, including AI-assisted creation, the control requirement assumes heightened importance. It is necessary to analyze the extent and effects of random intervention on the assessment of human control, and in combination with the specific context of AI random intervention in creative processes, to further analyze its impact on the determination of control. “Limited Control” Standard should be established as a necessary requirement for judicial practice. The examination of this requirement encompasses the possibility of human’s intellectual input, elements of original human expression, embodiment of creator's personal will, and related factors. Accordingly, “Limited control” standard necessitates the human intellectual input to lead to the crucial expression in AIGC, and such expression can be perceived. Based on the justification of the limited control standard, this paper analyzes its key aspects, including: the size of the creative space constrained by AI technology framework, whether human intellectual input can and whether it leads to the formation of key expressions in AIGC, and whether such key expressions can be perceived or recognized by humans. Furthermore, this paper analyzes the application of “limited control” standard across specific scenarios: ① In cases of weak control, AIGC is likely to be protected by copyright law, despite varying levels of control strength, with protection typically limited to elements reflecting human contribution, resulting in two outcomes: first, weak control in human-machine symbiosis can demonstrate weak human control capability at any link, such as input end, process end, or output end of AI generation, thus meeting the requirement of human control; second, under weak control, the generation process of AIGC fails to reflect human intellectual input, making it difficult or impossible to prove that the human control has been implemented; ② In the absence of control, it is difficult to grantcopyright, but under specific conditions,alternative protection paths, such as the Anti-Unfair Competition Law, may be sought; ③ In strong control scenarios, AIGC likely qualifies for copyright protection. Under strong control, human authors either guide the AI generation process or merely use the AIgenerated products as constituent elements during their creation. Having the author lead and control the formation of key expression elements in the work can meet the requirement for controlling elements.