1. Introduction

Today, in the wave of digital technology, AI has been applied to various aspects of society and has become one of the hot topics of discussion. The continuous iterative upgrading of large models such as ChatGPT and DeepSeek has transformed AI from a simple auxiliary tool into a quasi-autonomous content-generation system. A large amount of text, images, audio, and other content generated by AI has flooded into the cyberspace, reshaping the content ecological pattern. However, the proliferation of AI-generated content—through its dissemination, use, and derivative development—has created significant challenges for the digital economy, particularly concerning copyright ownership, infringement determination, and benefit distribution. If the legal vacuum is allowed to continue, it will be difficult to protect the rights and interests of creators, the innovation ecosystem will be easily unbalanced, and even an industry trust crisis may be triggered. In 2024, China's first "infringement of AI-generated images" case was judged [1]. In this case, the plaintiff safeguarded their rights because their AI-generated painting was used commercially without permission. Regarding the controversial points of this case, such as "whether AI-generated content has the originality of works" and "how to determine the infringement judgment standard", the court finally determined based on the provisions of the Copyright Law of the People's Republic of China (hereinafter referred to as the "Copyright Law") that the images generated by the plaintiff using AI software belonged to the works of fine art and had a certain form of expression, and should be protected by the Copyright Law. However, the US Copyright Office has repeatedly rejected applications for registering AI-generated works based on the underlying logic of "human creative will as the core". At the same time, theoretical research also shows a trend of multi-dimensional exploration. For example, focusing on "whether AI-generated content constitutes a work", some scholars have proposed the subjective standard of originality with the "author as the center" as the core [2], and some scholars have proposed the objective standard of originality with the "work as the center" as the core [3]. Based on this, this paper is based on the existing judicial practice and theoretical basis, takes the eligibility for copyright protection of AI-generated content as the core issue, sorts out existing theoretical viewpoints, and combines the judicial practice experience related to AI-generated content to deeply explore the copyright protection mechanism of AI-generated works, hoping to provide intellectual reference for the legal regulation of new objects in the digital economy era.

2. Theoretical disputes over the eligibility for copyright protection of AI-generated works

2.1. The affirmative theory represented by meeting the constitutive requirements of works and the legislative purpose

This view is mainly based on the constitutive requirements of works in the Copyright Law and the legislative purpose. Taking the "AI text-to-image" case as an example, some scholars believe from an appearance perspective that the content in question belongs to a work of fine art and has a certain form of expression. From the perspective of the generation process, they argue that the content in question has a recognizable difference from prior works, which reflects the plaintiff's choices and arrangements and meets the "originality" requirement. In addition, the results generated by AI need to be based on the prompt words designed by people, and the results generated by different people are not the same. This difference can reflect the original intellectual input of humans [4]. Some scholars also propose that AI-generated works are not only output by using input instructions through established algorithms and programs, but can also create through active learning without pre-set algorithms and rules [5]. It is worth noting that the court also supports this view in judicial practice. In the "AI text-to-image" case, the court determined the work "Spring Breeze Brings Tenderness" was copyrightable primarily because its appearance was indistinguishable from ordinary photographs or paintings, belonging to a work of fine art and having a certain form of expression. On the other hand, the plaintiff made a significant intellectual contribution, from conceiving the idea to making the final selection of the image. In short, the court believed that the image in question reflected the plaintiff's intellectual input and met the "intellectual creation" requirement. At the same time, the act of creating using this model also reflected the plaintiff's original expression and met the "originality requirement". In addition, some scholars propose based on the legislative purpose of the Copyright Law that AI, as a tool to extend human creative capabilities, its generated content is essentially an externalized carrier of human creativity. Therefore, determining that AI-generated content constitutes a work is in line with the core goal of the Copyright Law to encourage creation [6].

2.2. The negative theory represented by not meeting the requirements for protected subject matter and authorship under copyright law

In the case of Beijing Filin Law Firm v. Beijing Baidu Netcom Science and Technology Co., Ltd. [7], the court held that the graphics in the article in question were automatically generated by AI software, did not meet the originality requirements of graphic works, and did not constitute graphic works. The plaintiff's claim to enjoy copyright over them was not established. The court also specifically pointed out that during the process of the AI software automatically generating content, the behaviors of the software developers and users were not creative acts in the legal sense, and the relevant content did not convey their original expressions. Therefore, this content could not qualify as a work. Subsequently, some scholars proposed that AI-generated content does not meet the provisions of the Copyright Law on the protection of subject matter. For example, some scholars believe based on Article 11 of the Copyright Law that the current Copyright Law generally limits "authors" to natural persons or legal persons. Natural persons, as the only subjects with real intellectual activities and subjective wills, their creative acts directly reflect human creativity and are regarded as the most basic type of authors. However, AI, as an inanimate machine, does not have an independent legal personality and creative will. From the perspective of the originality of works, AI-generated content is the result of the application of algorithms, rules, and templates. Its generation process is neither a creative act nor can it reflect the unique personality of the creator. Therefore, it does not meet the requirements of the originality of works [8]. It is worth affirming that this view has also been recognized by the judiciary.

2.3. The compromise theory with core of the eligibility for copyright protection

It is worth noting that some scholars have proposed a compromise theory to reconcile the contradictions between the "affirmative theory" and the "negative theory", that is, neither completely denying the eligibility for copyright protection of AI-generated works nor directly giving them the same status as works created by natural persons. The reasons are as follows: On the one hand, some scholars propose that for AI-generated works that do not directly originate from humans and do not require humans to define rules in advance, although it is difficult to constitute works in the sense of the current Copyright Law in the interpretive theory, when the industry develops to a certain extent, such as when AI can create independently, it is necessary to amend the relevant provisions of the Copyright Law to provide institutional protection for such works. China's current generative AI technology innovation is still in its initial stage, and the formulation of laws and regulations should leave a certain space for scientific and technological innovation [9]. On the other hand, some scholars also propose that China's current generative AI technology innovation is still in its initial stage, and the formulation of laws and regulations should leave a certain space for scientific and technological innovation. For the requirements of data legality during the product research and development stage, an open and inclusive regulatory principle should be adopted according to the data type and basic ethical requirements [10].

3. Feasibility analysis of the eligibility for copyright protection of AI-generated works

3.1. The realistic demand for the development of AI-generated content

The industrialization process of AI-generated content has formed a cross-field penetration scale effect. Driven by both technological iteration and business practice, there is an urgent demand for the adaptability of the copyright system. From the perspective of industrial data, the global AIGC market size is expected to exceed $120 billion in 2025, with a compound annual growth rate of up to 42% [11]. By the end of 2024, the number of AI enterprises in China reached 13,069, ranking second in the world. In 2023, the scale of AI publications increased by 316% compared with 2016, accounting for more than 33% globally, but saw a 14.8% year-on-year decline in 2024. Among them, the number of AI invention patent applications exceeded 140,000, a 36-fold increase compared with 2012 [12]. At the same time, China's AI painting industry is expected to experience explosive growth, and its market size is expected to climb from 207 billion yuan in 2025 to over 500 billion yuan in 2030, with a compound annual growth rate of over 35% [13]. It can be seen that market development has provided a huge realistic demand for the research on the copyright issues of AI-generated content.

3.2. The development of technology requires timely provision of systems

At present, China's AI industry has ranked among the "top tier", gradually achieving a dual leap in technological breakthroughs and infrastructure layout. In 2025, the scale of China's AIGC market will exceed 800 billion yuan, accounting for 35% of the global share [14]. At present, the theoretical circle has basically reached a consensus on the industrial status of AI, that is, China's AI industry has entered a "scaling and innovation phase of large-scale application and parallel technological originality". The industrial ecosystem covers the entire chain of "model training - application development - scenario implementation", with a technology self-sufficiency rate of over 70%, reducing dependence on foreign core algorithms [15]. Despite its significant industrial status, there is still a "structural shortage" in the regulation of AI-generated content by the copyright system, which has led to multiple dilemmas in judicial practice. For example, there are prominent institutional problems such as a logical break in right ownership [16] and chaotic distribution of infringement liabilities. Facing the contradiction between "high industrial breakthrough" and "institutional gap", the theoretical circle has formed a strong consensus that "the system must adapt to technology". Therefore, in the face of the strong industrial development driven by technological development, the timely provision of the system is particularly important.

3.3. The revision of the copyright law leaves room for AI-generated content to be incorporated into law

Prior to its 2020 revision, Article 3 of the PRC Copyright Law utilized an exhaustive "closed list model," which enumerated only eight specific categories of works and lacked a catch-all provision [17]. This legislative model can ensure the clarity of rules but is difficult to adapt to the rhythm of technological iteration. For example, when emerging forms of expression such as AI-generated content, virtual idol performances, and algorithm-generated music appear, due to the lack of regulations on such emerging forms of expression in the Copyright Law, it is impossible to effectively play the protection function. However, in 2020, when the Copyright Law of the People's Republic of China was revised, Item 9, "other intellectual creations that meet the characteristics of works", was added to Article 3, changing the protected object of copyright from a "closed list" to an "open catch-all" provision. As the Legal Work Committee of the Standing Committee of the National People's Congress clearly explained, when legislators revised the law, they had foreseen that technological development would give birth to new types of intellectual creations, and reserved space for future system adaptation through the catch-all provision [18]. The underlying logic of this change lies in "returning to the essence of works", that is, the core of copyright law protection is "original expression", not "creation method" or "work type" [19]. As long as an expression meets "originality" and "reproducibility", whether it is independently created by humans or a new type of expression born from future technologies, it can be included in the protection scope through the catch-all provision. In fact, judicial practice has already started to explore the standardized application. For example, in the "AI-generated painting infringement case" [20], the Shanghai Intellectual Property Court determined that "the user's design of AI instructions, theme settings, style, and color preferences all reflect the author's personalized choices, making the generated paintings original", and relevant rights should be protected in combination with the catch-all provision. As scholar Li Yufeng pointed out, this judgment essentially incorporates AI-generated works into the existing regulatory system by interpreting the connotation of "intellectual creation". That is, as long as there is human "substantive control" over the creative direction and form of expression during the creation process, even if the final expression depends on the operation of the algorithm, it can still be determined to meet the applicable conditions of the catch-all provision [21].

4. Theoretical reasons for the eligibility for copyright protection of AI-generated works

4.1. AI-generated works meet the requirements of work originality

The Copyright Law of the People's Republic of China and the Implementing Regulations of the Copyright Law of the People's Republic of China do not clearly define "originality", but through the interpretive inertia of judicial practice, a judgment standard of "minimal degree of creativity + personalized expression" has been formed [17]. Among them, minimal degree of creativity does not require the work to be "unprecedented", only that it reflects the creator's active selection and rejection of expression elements. As the Supreme People's Court explained, "even if the creative idea originates from existing materials, as long as the expression has a personalized arrangement, it meets the requirements of originality" [22]. Personalized expression emphasizes the subjective initiative of people in the creative process. That is, regardless of whether the tool is a paintbrush or AI, as long as the creator has "substantive control" over the direction of expression, it meets the requirements. In fact, the "blank" in the legislative definition of "originality" is an active choice for technological adaptation. The controversy over the originality of AI-generated works is essentially an "iteration of creative tools", not a "breakthrough of the originality standard". As long as the user has "substantive control" over the expression, it is completely consistent with the underlying logic of the legislation on originality [23].

Based on the existing legislative provisions, currently, the theoretical circle also has different understandings of originality. For example, some scholars have proposed the "control-selection theory", that is, the essence of originality is the creator's "substantive control" and "personalized choice" of the form of expression. As long as there is an active decision-making process on the direction of expression and the combination of elements during the creation process, originality can be determined without worrying about the creative tool [24]. Some scholars have also proposed the "labor + creation dualism", that is, originality requires both "intellectual labor input" and "minimal degree of creativity breakthrough". From the process dimension, "intellectual labor input" emphasizes the subjective initiative of creation. In an AI-generation scenario, users exercise substantive control over the creative direction, details, and results through instruction design, parameter adjustment, and result screening. This process is analogous to the intellectual activities of conceiving, modifying, and finalizing in traditional human creation. From the result dimension, "minimal degree of creativity breakthrough" requires the expression to be different from existing works. In fact, due to the uniqueness of user instructions, AI-generated works can form recognizable differences in form or content, meeting the "minimum creativity" threshold [25].

The justification of the originality of AI-generated works within the framework of the Copyright Law needs to focus on the substantial connection between the personalized characteristics of the form of expression and the human intellectual input in the creative process. From the formal level, the originality of AI-generated content is reflected in the creative selection and combination of expression methods, and its core lies in forming a unique expression different from existing works through element arrangement and style integration. For example, when a user instructs AI to generate an image of "Cyberpunk-style Dunhuang Flying Apsaras", the image output by the algorithm may combine the industrial texture of mechanical gears with the elegant dynamics of the flying apsaras' ribbons. This non-mechanical combination of cross-dimensional elements constitutes a new expression of "the coexistence of tradition and the future" in visual composition and symbolic metaphor, which has legal equivalence to the personalized expression of human creators through the characteristics of the medium [26]. In judicial practice, the visual presentation of such works, due to the user's active selection of style and elements, forms a differential expression with aesthetic value, that is, it meets the requirement of "minimum creativity" in the Copyright Law of the People's Republic of China [20].

In the dimension of the creative process, the originality of AI-generated works essentially stems from human substantive control and intellectual intervention in the generation process. The prompt words input by users are not simple information listings but contain refined designs of themes, styles, and details. The semantic depth and creative complexity of these prompt words directly determine the uniqueness of the output content. This process is similar to the "concretization of ideas" in traditional creation. For example, transforming the abstract idea of "stratification of the right to breathe in a dystopian society" into specific instructions including character settings and narrative perspectives requires the mobilization of logical thinking and expression skills [27]. Furthermore, the parameter adjustment and result screening links constitute a progressive optimization of the expression details. Users adjust parameters such as color saturation and narrative rhythm or select the version that best meets the creative intention from multiple generated results. This is essentially a personalized selection and rejection of expression elements, which is homogeneous with the intellectual labor of human creators in modifying drafts and adjusting details [17]. Such intervention behaviors reflect the user's "active decision-making on the form of expression" and constitute the core source of originality [28].

4.2. AI-generated works conform to a certain form of expression

Article 3 of the Copyright Law of the People's Republic of China regards "being able to be expressed in a certain form" as a necessary condition for a work to be protected. Article 2 of the Implementing Regulations of the Copyright Law of the People's Republic of China further clarifies that this requirement includes "perceptibility" and "reproducibility". Perceptibility means that the expression can be perceived by human senses; reproducibility means that the expression can be fixed and reused. The 2020 revision of the law added the catch-all provision of "other intellectual creations that meet the characteristics of works" in Article 3(9), which recognized the new forms of expression spawned by digital technology. The Legal Affairs Commission of the Standing Committee of the National People's Congress emphasized in the Interpretation of the Copyright Law that "a certain form" encompasses "modern technological means such as digital storage and online dissemination," providing legislative grounding for digital content—including AI-generated texts, images, audio, and multimedia files.

Currently, the scholarly community has developed two core theoretical interpretations regarding AI-generated content's "conformity to statutorily recognized forms of expression", providing doctrinal support for its formal legitimacy.First is the "technology neutrality" theory, championed by Professor Wang Qian, which fundamentally deconstructs the traditional "carrier determinism" mindset. He argues that the core of "reproducibility" under the Copyright Law of the People's Republic of China does not hinge on the form of the carrier but rather on "whether the expression can be fixed in an objective form and repeatedly reproduced." In the digital age, this requirement has been revolutionized through binary technology. AI-generated content exists as structured data composed of 0/1 code, where "reproduction" manifests as data reading, writing, and format conversion. Such technical operations hold legal equivalence to traditional acts like "copying paintings by hand" or "transcribing written text." While the former relies on algorithmic fixation and dissemination, and the latter depends on manual labor, both fulfill the fundamental function of "enabling a work to exist independently of its creator" [17]. Second is the "digital fixation primacy theory", represented by scholar Li Ming, which reconstructs the legal presumption of "fixation" from a technical implementation perspective. He argues that AI-generated content inherently forms a "structured digital file" upon output, a process equivalent to the "fixation" requirement under copyright law. While traditional works require active fixation by the author, the very act of AI output constitutes "digital fixation." Once a user triggers the "generate" command, the algorithm inevitably produces a standardized data file readable by devices—no additional fixation step is needed [29].

The core of the Copyright Law's requirement for a work to "be able to be expressed in a certain form" lies in "perceptibility" and "reproducibility", which are not only the physical existence form requirements of a work but also the carrier basis for originality to be externalized. Therefore, empowered by digital technology, AI-generated works not only fully meet the above requirements but also realize the visualization and verifiability of originality through the technical coupling of "form expression" and "creation process" [18].

Firstly, in terms of perceptibility, the perceptibility of AI-generated text, images, videos, and other content is essentially "the human-perceivable form under human control". From the perspective of technical implementation, the binary data output by AI is decoded by terminal devices and converted into a symbol system recognizable by human senses. For example, text content is presented on the screen in character encoding, and image content forms a visual image with a pixel matrix. These digital forms of expression are completely consistent with the physical carriers of traditional works in terms of perception logic. The former interacts through digital signals, and the latter is presented through physical symbols, both of which meet the core requirement of "being perceivable by human senses". In judicial practice, the involved AI paintings form a unique visual image of "the coexistence of traditional costumes and mechanical structures" due to the user's control over "style instructions" and "detail parameters". When this image is presented through a screen or printed matter, its thematic innovation and the uniqueness of element combination can be directly perceived, constituting an "original expression" in the sense of the Copyright Law [20]. Secondly, in terms of reproducibility, the reproducibility of AI-generated works stems from the standardized fixation and lossless transmission of "expression forms" by digital technology. From the perspective of technical characteristics, the output content is stored in a file format, which is essentially a structured arrangement of binary codes. It can be reused infinitely through operations such as "copy-paste", "upload-download", and "format conversion", and the replication process does not change the core form of expression. For example, AI-generated novel texts can be displayed on different terminal devices without changing the narrative structure, and AI paintings can be converted into vector graphics for advertising designs of different sizes without losing the details of the picture. This “Substantially Similar Reproduction” feature enables the original expression of the work to be completely retained. In judicial practice, the court has clearly recognized that the reproducibility of AI-generated works meets legal requirements. For example, in the first case in China where AI text-generated images do not constitute works and do not infringe copyright, "Feilin Law Firm v. Baidu" [30], the defendant downloaded and uploaded the full text of the AI-generated novel to a pirated platform. The court held that this act constituted an infringement of the "right of reproduction" of the work, on the grounds that "the downloading and uploading operations are essentially the storage and dissemination of digital content, without changing its text arrangement and narrative logic, and belong to the reproduction of original expressions". The more far-reaching significance of this case is that digital replication technology enables the "originality" of AI-generated works to break through the carrier limitations of traditional works. Even if the generated content is temporarily stored in the platform draft box or cloud cache, its fixed nature in digital form has met the requirements of "being able to be expressed in a certain form" in the Implementing Regulations of the Copyright Law of the People's Republic of China, having the same formal legitimacy as "drafts" and "manuscripts" created by humans.

5. Conclusion

At present, there are still judicial differences and theoretical controversies in the determination of the eligibility for copyright protection of AI-generated works. Institutional issues such as indeterminate copyright ownership and unclear allocation of infringement liability are prominent, which are in sharp contradiction with the actual demand for the large-scale development of the AI industry. Based on this, this paper holds that the eligibility for copyright protection of AI-generated works should be based on the core requirements of the Copyright Law of "originality + reproducibility". That is, when humans have substantive control over the generation process, and the output content reflects their personalized expression and recognizable differences, it can be protected through the catch-all provision of "other intellectual creations that meet the characteristics of works" in Article 3(9) of the Copyright Law of the People's Republic of China revised in 2020. This determination is not only in line with the open and inclusive spirit of the legislative amendment but also can balance technological innovation and rights protection. However, in the future, judicial organs should further clarify and refine the judicial judgment standards of "substantive control", clarify the rules of right ownership under different generation modes, and pay attention to the institutional challenges that may be brought by the improvement of AI's independent creation ability. It is worth noting that due to the research perspective, this paper does not conduct in-depth research on the differentiated regulatory paths of content generated by weak AI and strong AI. However, based on the current reality of AI development, weak AI is still the mainstream form in the market, and it is necessary to discuss weak AI at present and strong AI in the future. This paper hopes that more practical workers and scholars will further pay attention to the differences between AI-generated content in the era of strong AI in the future and that generated by weak AI at present, as well as the differences in institutional protection, so as to provide theoretical support for China's Copyright Law to meet the needs of technological development in a timely manner.