Determining the Originality of AI-Generated Works: A Prompt-Based Perspective ()
1. Introduction
With the continuous advancement of artificial intelligence technologies, AI-generated content has seen increasingly widespread application in the cultural consumption market, revealing considerable cultural value and commercial potential. This development poses significant challenges to the traditional theoretical framework and operational mechanisms of copyright law. In recent years, the legal status and copyrightability of AI-generated outputs have emerged as prominent issues in global judicial practice and academic discourse. Among these developments, the 2023 U.S. case of Zarya of the Dawn (also known as the “Space Opera” case) attracted broad attention, signaling that copyright disputes over AI-generated works are no longer hypothetical but have become pressing regulatory concerns.
China has also made notable progress in this area. In November 2023, the Beijing Internet Court issued a first-instance judgment in a copyright infringement case involving AI-generated images. For the first time, a court formally recognized an AI-generated illustration as a copyrightable work of fine art, establishing a groundbreaking precedent on a global scale (Beijing Internet Court, 2023).
However, this decision left a critical question unresolved: whether the prompts input by human users constitute a legally meaningful contribution of originality to the final AI-generated expression—one sufficient to justify authorship and copyright ownership. This unresolved issue has triggered deeper theoretical and doctrinal inquiry into the legal relationship between prompts and AI-generated outputs.
Against this background, this article seeks to address the gap between theory and practice. By focusing on prompts as the point of analytical departure, the article explores their functional role in the generative process and their connection to the originality of the resulting content. On this basis, it proposes an originality assessment framework that combines the concepts of authorial personality and a minimal degree of creativity. This dual-standard is intended to provide a clearer, more practicable foundation for recognizing copyright in AI-generated works and to offer constructive guidance for future judicial decisions and normative development.
2. The Challenge of Determining Originality in AI-Generated Works
With the rapid advancement of artificial intelligence technologies, the determination of originality in AI-generated content has extended beyond the traditional doctrinal boundaries of copyright law. This issue now encompasses novel legal uncertainties prompted by the integration of autonomous systems into creative processes. This section begins by examining the pre-existing challenges surrounding originality under conventional copyright doctrine, before turning to the new complications introduced by AI-generated works.
2.1. Diverging Standards for Originality in Traditional Copyright Law
Under Chinese copyright law, originality requires that a work be both independently created and demonstrate a certain degree of creativity. While the criterion of independent creation is rarely disputed in practice, the threshold for creativity has long been a subject of doctrinal divergence (Supreme People’s Court of China, n.d.). Judicial interpretations vary across three primary approaches: 1) a requisite level of creative height; 2) the manifestation of authorial personality; and 3) a minimal degree of creativity.
In practice, courts often rely on judges’ subjective assessments when evaluating originality. For example, Judge Li Zizhu of the former Chaoyang District People’s Court in Beijing analyzed relevant judicial opinions via the IP case database “Zhi Chan Bao” and categorized originality determinations into three types: decisions referencing “creative height” (206 cases), “minimal creativity” (32 cases), and “authorial personality” (60 cases, with an additional 128 cases referencing “a certain degree of personality”). These data points illustrate the lack of a unified approach and the extent to which judicial discretion dominates the assessment of originality (Li, 2024).
The view that interprets “creativity” as requiring a certain level of creative height holds that the “originality” of a work must be based on a measurable degree of creativity. Under this approach, only when a work reaches a specific threshold of creative input can it satisfy the legal requirement of originality. For example, Article 2 of the German Copyright Act stipulates that a work must contain a certain degree of creativity in order to be considered original (Gasser, Trans., 2013). This “certain degree of creativity” is understood as the minimum threshold for copyright protection in Germany (Loewenheim, U. (Author), Zheng, C. (Trans.). 1991). Under German copyright law, the intellectual output resulting from the act of creation must exhibit a certain level of creative or design sophistication (Leutheusser, T. (Author), Zhang, H., & Wu, Y. (Trans.), 2019). For example, in the case of photographic works, German copyright law distinguishes between “photographic works” and “simple photographs”, granting the former full copyright protection and the latter only neighboring rights. This distinction is based on whether the creation process demonstrates a sufficient level of creative input (Wang, 2015).
The second approach focuses on authorial personality, emphasizing that originality arises when a work reflects the author’s individual character, subjective judgment, or personal aesthetic (Beijing Haidian District People’s Court, 2007). French copyright law exemplifies this perspective, treating originality as an inherently subjective standard rooted in the unique intellectual and emotional contribution of the author (Li, 2014).
The third approach—the minimal creativity standard—requires only that a work exhibit a distinguishable difference from existing works, without demanding a particular degree of creativity (Zhongshan Intermediate People’s Court of Guangdong Province, 2014). In Chinese judicial practice, this interpretation has gained traction in several cases. For example, the Zhongshan Intermediate People’s Court in Guangdong held that a literary work could satisfy the originality requirement so long as it involved a minimal level of creative effort, producing perceptible differences from prior expressions Zhongshan Intermediate People’s Court of Guangdong Province (2014).
In summary, the standards for determining originality from the perspective of traditional copyright law have not yet been unified. Whether it is “having a certain degree of creativity,” “reflecting the author’s personality,” or “minimum creativity,” a unified consensus cannot be reached. This inherent divergence will increase the instability of judicial judgments. Therefore, the three controversial standards in the context of traditional copyright law cannot be directly applied to the determination of originality of artificial intelligence products alone.
Taken together, these three standards reveal a fundamental lack of consensus on how to assess creativity within the concept of originality. The absence of a coherent and consistent framework not only introduces uncertainty into judicial outcomes but also weakens the law’s predictability and normative stability.
Accordingly, none of these traditional standards can be directly or exclusively applied to the originality assessment of AI-generated works. The involvement of AI technologies introduces distinct complications that challenge the conceptual underpinnings of originality. These developments demand a reconsideration and recalibration of existing legal standards to account for the unique features of AI-assisted or AI-generated creative processes, ensuring the adaptability and forward-looking nature of copyright doctrine.
2.2. The Distinctive Challenges in Assessing Originality of AI-Generated Works
At present, legal disputes over originality in works often center on two core questions: whether the work was independently created and whether it possesses creativity. In the context of AI-generated works, the first condition—independent creation—is generally uncontroversial. As long as the AI user does not engage in plagiarism and initiates the generative process independently, this requirement is typically satisfied. The real point of contention lies in the second element: the source of creativity in AI-generated outputs.
One school of thought holds that creativity originates from the AI model itself. Under this view, the generative process is driven by pre-trained algorithms and large language or image models that interpret user prompts and generate outputs accordingly. As such, the creativity reflected in the final product is attributed to the model’s internal operations, rather than the human user.
In contrast, another position sees AI as merely a tool—albeit a sophisticated one—controlled by human users. From this perspective, under the current legal framework, AI systems cannot be considered authors or legal subjects. Instead, it is the human user’s individualized choices—such as model selection, prompt construction, and parameter adjustment—that contribute creative input. Therefore, the originality of AI-generated content, if any, must be traced back to the user’s contributions.
The debate is well illustrated by the so-called “AI-generated image” case adjudicated by the Beijing Internet Court. Presiding Judge Zhu Ge opined that if the prompts input by an AI user are sufficiently creative, then the resulting AI-generated work may be deemed to possess a certain degree of originality. However, this reasoning stops short of establishing whether there exists a direct causal link between the user’s individualized prompt design and the final expressive form of the output. In other words, the question remains: to what extent can users control or determine the expressive features of the AI-generated work through prompts?
Those who affirm a strong relationship between prompts and creative expression argue that users can, through intricate prompt engineering and parameter configuration, meaningfully guide the generation process to produce targeted outputs. In this sense, users exercise creative control over the output’s expressive form. Critics of this view, however, argue that despite the complexity of prompts, the generative process remains inherently stochastic. As a result, users lack full control over the output, and the final expression cannot be directly attributed to user input. According to this line of reasoning, even if the AI-generated work meets the objective conditions of a copyrightable work, its originality cannot be ascribed to the user—prompt input and final output are not meaningfully connected in terms of creativity.
In summary, existing Chinese copyright law lacks a unified standard for assessing originality, and the rise of generative AI has added a new layer of complexity. As generative technologies reshape creative practices, determining the legal relationship between prompts and originality in AI-generated works has become an urgent issue in both judicial practice and academic inquiry.
3. Analyzing the Role of Prompts in Determining the Originality of AI-Generated Works
The question of whether user-inputted prompts contribute to the originality of AI-generated works has given rise to divergent viewpoints. From a technical perspective, AI users can influence the presentation of generated content through a variety of mechanisms, including affirmative prompts, exclusionary (or negative) prompts, and the configuration of parameters. These tools offer users a means of directing the model’s output and shaping its expressive features to varying degrees. From the standpoint of judicial practice, recent Chinese case law has taken an affirmative position on the legal significance of prompts. In the first-instance ruling of the “AI-generated image” case, the Beijing Internet Court recognized that prompts with sufficient originality may be directly linked to the originality of the resulting AI-generated work. On this basis, the court held that the user’s creative contribution—manifested through the use of prompts—could be attributed as the source of the work’s originality. In contrast, the approach taken in the United States reflects a more restrictive stance. In the now-notorious Zarya of the Dawn case (also known as the “Space Opera” case), U.S. authorities acknowledged that the claimant exercised control over aspects of the work’s design and composition through prompts. However, the Copyright Office ultimately denied registration for the portions of the work generated exclusively by AI. It concluded that only those elements reflecting human authorship—such as edits made using Adobe Photoshop—could be protected. Purely machine-generated content was deemed ineligible for copyright, regardless of the user’s level of input at the prompting stage. These contrasting positions underscore the lack of international consensus on how to assess originality in AI-assisted creative processes. They also highlight the doctrinal uncertainty surrounding the extent to which prompts can be considered a creative act capable of grounding authorship under copyright law.
3.1. Prompts in the Creative Process of AI-Generated Works
In the context of AI-generated imagery, large-scale models such as Stable Diffusion and Midjourney have become particularly prominent. These models operate primarily through two modalities: “text-to-image” and “image-to-image” generation. Among them, Stable Diffusion allows for extensive prompt manipulation, enabling users to iteratively refine the initial output through successive input. While technically complex, this high level of operability makes it well-suited for users with specific creative intentions or professional demands. Within the Stable Diffusion ecosystem, users can influence output through affirmative prompts, negative prompts (which exclude unwanted elements), and a variety of parameter settings. The platform also supports numerous plugins and scripts—such as LoRA integration, high-resolution restoration, localized redrawing, and custom model training—adding to its flexibility but also raising the technical threshold for effective use. This has even given rise to a specialized professional role: the prompt engineer (Smith, 2023). By contrast, Midjourney offers a more accessible interface and simplifies the user experience. However, this ease of use limits the degree of creative control users can exercise over the output. The difference is analogous to a chef choosing between fresh ingredients and pre-packaged frozen meals: even with the same user, the creative space varies significantly depending on the model selected. Accordingly, model selection itself constitutes an initial act of authorial discretion that may carry legal relevance in determining originality.
The “text-to-image” generation process depends on extensive pretraining. During development, model designers compile massive image datasets, annotate visual features, and use these to train the system to associate specific textual prompts with corresponding visual outputs. This architecture enables the model to automatically render images in response to user instructions, approximating the user’s intended aesthetic.
However, the degree of user input varies dramatically. In one scenario, a user may input minimalistic prompts such as “man,” “basketball game,” or “outdoor stadium,” then accept the model’s random output. In this case, the prompts merely provide abstract concepts or thematic direction, offering no meaningful contribution to the expressive form of the resulting image. The user lacks both creative intention and predictive control over the final result. From a copyright perspective, the user’s contribution to originality in such instances is negligible.
In contrast, users who iteratively refine their prompts—adding affirmative instructions and negative constraints—exercise much greater influence over the final expression. Using Stable Diffusion, for example, a user may specify desired content through detailed prompts while also excluding unwanted elements such as low resolution, missing fingers, or distorted facial features. Prompts may consist of full sentences or sets of keywords, and users can continue modifying them after the initial output to incrementally approach a desired visual outcome. Notably, the sequencing of prompts is crucial: those placed earlier in the input string typically carry greater weight in determining the final image composition.
Thus, the final output of an AI system may vary substantially depending on the model selected, the sophistication of prompt input, the ordering of prompt terms, and the parameters or plugins configured. In such cases, the user’s creative contribution may be significant enough to justify claims of originality. Through successive refinements, the user engages in a process of aesthetic authorship—even if mediated by AI—that arguably supports recognition of legal rights in the resulting work.
3.2. Judicial Approaches to Prompts and Originality Determination
In the “AI-generated image” case adjudicated by the Beijing Internet Court, the plaintiff sought to establish their originality contribution to the disputed work by emphasizing their role in selecting the AI model and crafting the prompts. Specifically, the plaintiff presented the arrangement of affirmative and negative prompts, along with the configuration of generation parameters, as evidence of individual creative input. To reinforce the cultural value of the work, the plaintiff also submitted metrics such as viewership and engagement (e.g., likes) after the image was published on online platforms.
In its first-instance judgment, the court first rejected the possibility of recognizing AI models as legal subjects capable of authorship under current Chinese law. It then turned to examine the human user’s creative process. The court found that the work had been independently created by the plaintiff and demonstrated clear differences from existing works. It further held that the image reflected the plaintiff’s individual aesthetic judgment and personal choices, thus meeting the threshold of originality.
As the first domestic case concerning AI-generated works in China, the central issue was whether the user’s actions—particularly the use of prompts—amounted to a creative contribution sufficient to ground authorship. The court concluded that the plaintiff exercised meaningful control over elements such as visual composition and character rendering through prompt input and parameter design. The user’s iterative refinements and modifications to the output were interpreted as manifestations of aesthetic authorship. Accordingly, the court held that the final image was independently created by the human user and possessed a sufficient degree of originality to merit protection as a work of fine art. The ruling explicitly treated AI as a tool, not a co-author, thereby excluding AI from considerations of originality altogether.
Critics of the Beijing Internet Court’s reasoning have drawn attention to contrasting developments in the United States, most notably the Zarya of the Dawn case—commonly referred to as the “Space Opera” case. In that matter, the human user input over 600 prompts into the Midjourney model to generate the initial version of the disputed work. The user then employed Adobe Photoshop to make secondary edits and used another AI tool, Gigapixel AI, to enhance the image’s resolution and scale. As in the Chinese case, the claimant emphasized their extensive role in shaping the output through prompt engineering, as well as the deliberate exclusion of flaws such as cracks, scars, and artifacts, and the selection of themes, tones, genres, and color schemes.
Despite this, the U.S. Copyright Office refused to register the entire work, permitting registration only for the elements that reflected human authorship—namely, the post-generation edits made using Adobe Photoshop. The Office held that the portions generated solely by AI did not qualify for protection, as machines cannot be recognized as authors under U.S. copyright law. Notably, the Office’s position was not wholly dismissive; it left room for partial recognition of human-authored elements within AI-assisted works. According to the Compendium of U.S. Copyright Office Practices, authorship must be attributed to a natural person or a legal entity, and machine-generated content is categorically excluded. Nonetheless, the Office endorsed a case-by-case approach, emphasizing that human users may be deemed authors if they can demonstrate meaningful creative control over the final expressive outcome.
While the EU Artificial Intelligence Act does not directly address the copyrightability of AI-generated works, its mandatory transparency obligations—such as labeling outputs as “AI-generated”—implicitly negate the core requirement of human authorship (Quintais, 2025). Against this regulatory backdrop, the United Kingdom Intellectual Property Office (UK IPO) explicitly denies copyright protection for purely AI-generated content in its January 2025 consultation document, asserting that copyrightable “works” must embody the “author’s own intellectual creation.” The document states: “Works generated solely by prompts do not meet the originality threshold for copyright protection due to insufficient human control over the final expressive form” (UK IPO, 2025: 4.8). This position categorically rejects the copyright significance of prompt engineering while recognizing protectability only for subsequent creative arrangement or modification of AI outputs by humans. This standard aligns with the U.S. Copyright Office’s “human-editing reservation” principle in Zarya of the Dawn but remains stricter than the Beijing Internet Court’s recognition of prompt creativity in China.
3.3. Theoretical Debates: Do Prompts Establish Originality?
The core of the debate concerning prompts and the originality of AI-generated works lies in whether the originality of the prompt can be directly linked to the originality of the resulting output. In other words, the issue is whether users can reasonably predict and control the final expressive form of the AI-generated content through the use of prompts. Currently, academic opinion is divided into two main camps. One view holds that users cannot effectively control the final output through prompts alone, and therefore cannot claim authorship or copyright in the generated work. The opposing view argues that full predictive control over the generated content is not a prerequisite for authorship. Analogies are drawn to traditional art forms such as ink-splash painting, in which the artist creates by freely scattering ink on the canvas. Although the exact outcome may be unpredictable, this lack of precise control does not negate the artistic value of the work, nor does it preclude its recognition as a copyrightable creation (Zhang, Guo, & Feng, 2024). The final expressive form of an AI-generated work is often the result of the user’s deliberate selection, arrangement, and design of detailed prompts. Accordingly, the originality of the generated output can be attributed to the user, who should thus be recognized as the author of the work. Some scholars have summarized the two opposing positions outlined above as the affirmative and negative interpretations of the “expected control” theory.
3.3.1. The Affirmative View: High Control as a Prerequisite
Proponents of the affirmative interpretation of the expected control theory argue that users cannot effectively control the final expressive form of AI-generated content through prompt input alone. Since AI users are generally unable to determine the ultimate output based solely on their own free will, the resulting work cannot be regarded as content created by a human using AI as a mere tool (Feng, 2023). Some scholars further argue that the “text-to-image” generation process should be divided into two distinct stages: the initial input of instructions and the subsequent refinement of the generated output. At the initial stage, when users provide textual or graphical prompts, they typically cannot foresee the specific results produced by the AI model, and thus cannot be said to make a creative contribution at that point. However, if users continue to input prompts after the AI generates a “first draft,” modifying the output and introducing individualized decisions throughout various stages of refinement, they may indeed contribute original expression to the final result. In such cases, the user could potentially acquire copyright in the AI-generated work by virtue of their creative input into the final expressive form.
3.3.2. The Negative View: Control Need Not Be Absolute
The core argument of the expected control theory holds that, in the context of AI-assisted creation, users exercise only minimal control over the final output through prompts. As such, reliance on prompts alone is insufficient to establish that the user has made an original contribution (Lyu, 2023). By contrast, scholars who reject the expected control theory argue that the degree of control users exert over the final expressive form through prompts cannot be generalized and must instead be assessed on a case-by-case basis. From the very outset—starting with the selection of the AI model—users vary in the extent to which they influence the final output. Factors such as the choice of affirmative or negative prompts, the sequencing of prompt terms, and the configuration of generation parameters all contribute to differences in user control. Accordingly, the determination of whether a user’s input reflects sufficient creative control should be made in light of the specific circumstances of each case.
First, even prior to the era of artificial intelligence, human creators were rarely able to exercise complete control over the outcome of their creative processes. Take the example of ink-splash painting: during the act of creation, the artist may intend to apply pigment to the canvas in a particular way, yet cannot precisely predict where the ink will fall or how the resulting patterns will form. This inherent unpredictability does not negate the artist’s original contribution, nor does it diminish the potential artistic value of the work. In fact, the final expressive form of an ink-splash painting often exceeds the boundaries of the artist’s conscious will. Paradoxically, it is the very exercise of free will by the artist that gives rise to the work’s unpredictability, thereby reinforcing its creative authenticity (Burk, 2020).
Second, predictability and control in the creative process are inherently relative concepts and should not be absolutized. Between complete foresight and total unpredictability, between full control and complete lack of control, lies a broad spectrum of intermediate states. It is within these intermediate states that many AI users make original contributions. If an AI user were able to anticipate and control every single pixel of an AI-generated output, the advantages and purpose of large-scale AI models would be entirely undermined. The strength of such models lies precisely in their capacity to process vast datasets and respond to user instructions with expressive outputs that exceed the limitations of traditional human creativity. If every element of a generated work were fully subject to the user’s control and expectations, the process would be functionally indistinguishable from conventional human authorship—rendering the use of AI redundant and, in fact, discouraging further innovation and cultural production. As illustrated by the ancient Chinese idiom “painting the dragon and dotting the eyes”, even a minimal creative act—though occupying only a small portion of the overall work—can possess disproportionate artistic value. Accordingly, as long as the user makes an original contribution to the final expressive form of the AI-generated work, authorship should be attributed to the user with respect to that contribution, regardless of the degree of overall control. Given that AI systems themselves are not recognized as legal subjects under copyright law, it is the user’s creative input and expressive influence over the final work that justify conferring authorship.
4. A Proposed Standard for Determining the Originality of AI-Generated Works
Within the framework of copyright law, originality is the core criterion for determining whether a creative output qualifies as a protected work. It directly affects whether a given intellectual achievement meets the threshold for copyright protection. The concept of originality requires that a work reflect the author’s personal choices and judgments—decisions that are inherently unique. As the saying goes, “no two leaves are alike”, and originality serves to embody the distinctive thoughts and emotions of individual creators. This uniqueness does not demand a particular level of artistic or creative sophistication; rather, it simply requires that the work be the result of the author’s independent thought and expression (Li, 2019). Therefore, when assessing whether an intellectual creation possesses originality, the focus should be on whether it reflects the author’s independent thought and is presented through a creative mode of expression, rather than on the level or depth of its artistic achievement. This approach is more consistent with the fundamental purpose of copyright law and better serves to protect the legitimate rights and interests of creators.
4.1. A Combined Standard: Authorial Personality + Minimal Degree of Creativity
As previously discussed, judicial practice in China has adopted several approaches to the determination of originality, including the “authorial personality” standard, the “minimal degree of creativity” standard, the “certain level of creative height” standard, as well as hybrid models that combine elements of these three (Li, 2024). Among these, the “certain level of creative height” standard is relatively subjective and difficult to define. Different individuals may hold varying opinions on the degree of creativity in the same work, making this standard unsuitable for assessing the originality of AI-generated content. By contrast, evaluating originality from a binary perspective—whether it exists or not—is more objective, precise, and easier to apply and understand (Li, 2013). Therefore, attention should be directed toward the “authorial personality” + “minimal degree of creativity” standard for assessing the originality of AI-generated works.
Generally speaking, the “authorial personality” standard emphasizes whether a work embodies the author’s thoughts and emotions, as expressed through individualized creative choices. These personal elements often reside in the distinctive features of the work and are not assessed merely by examining the originality of the output in isolation. Accordingly, this standard does not require direct comparison with preexisting works. Of course, in the case of derivative works, such comparison becomes necessary. However, the purpose of such comparison is not to evaluate whether the derivative work surpasses the original in terms of creative height, but rather to determine whether the derivative incorporates new, personalized expression while retaining the essential elements of the original. The core inquiry remains whether the author’s unique imprint is discernible in the derivative work, rather than whether it reaches a particular level of creativity. Whether a work reflects the author’s personality ultimately depends on whether the work allows sufficient space and freedom for creative expression (Lipschitz, (Author), United Nations (Trans.), 2000). If the outcome of a work is not the only or one of a very limited number of possible expressions, and it is not produced by following a predetermined formula or specific standard, then the work can be said to possess creative space. This allows for the manifestation of the author’s personality, thereby satisfying the requirement of creativity (Leutheusser, (Author), Zhang, & Wu, (Trans.), 2019).
The “minimal degree of creativity” standard evaluates originality by comparing the new content with existing works, focusing on whether the new work exhibits novelty in its form of expression. This standard assesses the presence or absence of originality, rather than its degree. It does not impose any quantitative requirement on the level of creativity, but merely asks whether there are identifiable differences between the new work and prior works. This approach was articulated in the landmark 1991 U.S. case Feist Publications, Inc. v. Rural Telephone Service Co., which established the principle that only a minimal degree of creativity is necessary for a work to qualify for copyright protection (Feist Publications, Inc. v. Rural Telephone Service Co., 1991). The court held that, so long as a work is independently created, the presence of a “minimal degree of creativity” is sufficient to satisfy the originality requirement under copyright law. Similarly, in the first-instance judgment of the AI-generated image case in China, the Beijing Internet Court used the expression “identifiable differences from preexisting works” when assessing the originality of AI-generated content (Beijing Internet Court., 2023).
Some scholars argue that the “minimal degree of creativity” standard articulated in the U.S. Feist case also reflects an inherent consistency with the “authorial personality” standard—that is, works that convey the author’s ideas or personality often satisfy the threshold of minimal creativity (Lu, 2014). Applying both the “authorial personality” and the “minimal degree of creativity” standards to the assessment of originality in AI-generated works reflects a dual focus: on the one hand, it emphasizes the human element in the creative process; on the other, it addresses the need to protect the proprietary interests embedded in AI-generated content. This dual framework does not impose an undue burden on the determination of copyright eligibility for AI-generated outputs. Rather, it affirms and recognizes the role of human contribution in the process of human-machine collaboration. At the same time, adopting this combined standard also serves as a normative response to concerns over “AI land grabs” in the intellectual property landscape.
4.2. Justifications for Applying the “Authorial Personality + Minimal Creativity” Standard
Applying a combined standard of authorial personality and a minimal degree of creativity to assess the originality of AI-generated works offers several distinct advantages:
First, it enhances the objectivity of originality determinations by reducing interpretive uncertainty. As the well-known saying goes, “There are a thousand Hamlets in a thousand people’s eyes”—that is, artistic judgment is inherently subjective. If originality were assessed using a vague and elevated standard such as “a certain level of creative height,” it would likely result in inconsistent rulings in similar cases. By contrast, analyzing specific elements such as prompt design, parameter selection, and model configuration allows courts to evaluate whether the user’s personal choices have meaningfully shaped the generated work. This approach, coupled with an assessment of whether the resulting expression is recognizably different from existing works, reduces the subjectivity in judicial assessments and promotes greater consistency and legal predictability.
Second, this dual standard maintains objectivity while allowing for flexibility across different types of creative works. Not all works afford the same degree of creative freedom. For example, in the field of software design, user interface (UI) components such as login screens, confirmation buttons, and exit prompts are often functionally constrained, offering limited room for innovation. If originality were conditioned upon reaching a “high degree of creativity”, many UI designers would be excluded from copyright protection altogether. The combined standard, however, accommodates such functional limitations by recognizing that even minimal creative input—so long as it reflects personal judgment—can justify protection. This enhances fairness and promotes equitable treatment across varied categories of works.
Third, the proposed standard aligns with the inherent nature of human creativity. Creation is fundamentally a human activity, and both the process and the output should reflect human thought. This principle is embedded in the classic distinction between idea and expression in copyright law. The use of prompts and related tools in AI creation connects the user’s intellectual contribution to the final output. Under this view, any new expression—regardless of its artistic depth or refinement—has value simply by virtue of being previously unseen. As such, originality should not be reserved for only those works that meet elevated aesthetic criteria; rather, it should focus on whether the user has produced something novel through independent, expressive input.
Finally, the combined standard advances the core legislative objectives of copyright law. If the involvement of AI were used to categorically deny the user’s creative contribution, this would serve as a disincentive for human creators working with generative models. By adopting a framework that protects works reflecting authorial personality and minimal creativity, the law can simultaneously incentivize creation and promote cultural development. Recognizing more AI-assisted works as copyrightable encourages broader public engagement with emerging technologies, fostering both innovation in the AI sector and diversity in cultural expression.
5. Conclusion
As an objective requirement for copyright protection, originality should be assessed based on whether a work reflects the author’s independent thought and creative expression, rather than the height or depth of its artistic achievement. In the current stage of AI development, where artificial intelligence is best understood as a creative tool, it is both appropriate and legally sound to recognize the contribution of qualifying prompts to the originality of AI-generated works. Applying a dual standard that combines authorial personality and a minimal degree of creativity is more consistent with the normative spirit of China’s copyright law. This approach not only reinforces the law’s commitment to protecting both the personal and proprietary interests of authors, but also fulfills the broader public interest in accessing diverse cultural products. In doing so, it supports the dissemination of creative works and contributes to the flourishing of cultural innovation.