According to the technological knowledge of AI, both humans and AI make contributions to the final output. Considering those contributions, the human programmers or legal entities that not only provide existing works to the AI but also decisively affect how the AI processes those works deserve to be viewed as the author in the sense of the copyright system, while the AI merely functions as a mathematical tool helping the author create AI outputs. The so-called computational creativity reflected in those products is actually a form of human creativity. Therefore, AI output is a human creation rather than an AI creation, which means that the issue of AI authorship does not exist at the current technological level, and there is no theoretical disruption to grant copyright protection to AI-generated products.
However, those findings only tell that AI outputs can be protected like any other human-created works under copyright law; they do not tell whether these outputs should be protected, which is an empirical question that needs to evaluate whether granting protection to those outputs will achieve the goal of promoting the public interest or social welfare. The public interest promoted by the copyright system is reflected in the encouragement of learning: granting protection under the copyright system is expected to not only provide an incentive for the creation and dissemination of AI-generated cultural products but also to reserve space for the subsequent and liability-free use of the products. However, the empirical evidence shows that commercial value and private interest have not been confirmed by AI-related industry, investors, and even stakeholders in AI outputs. Therefore, it is not likely that stakeholders will obtain financial returns by bringing those products into circulation, which means that they lack the essential motivation to accelerate the creation and dissemination of AI outputs, and granting copyright protection will not be useful in providing that motivation. In addition, disincentives do not exist now because no evidence has shown that the existing interests in AI outputs are facing actual or possible damage from the free-riding actions forbidden by copyright law, meaning that it is not necessary to provide the incentive by granting copyright protection. More importantly, considering the close connection between AI technology and the digital universe hypothesis, granting protection to AI outputs will risk shrinking and destroying the space reserved for learning and subsequent use. When not limited to the public interest under the context of the copyright system, it can also be found that granting protection to AI outputs may result in a series of social problems outside the realm of copyright law. Therefore, by weighing the possible benefits and expenses in protecting AI outputs, this research suggests not granting copyright protection at least under the current technological level, although those products are human creation.