The rapid growth of generative artificial intelligence has raised fundamental questions about whether prompts used to instruct AI systems can attract copyright protection. Courts worldwide are now beginning to address a variety of questions related to AI and copyright and have started to examine whether AI prompts constitute original works or merely unprotectable ideas or other subject matter. A recent case, Chengdu Cultural Communication Co., Ltd v Information Technology (Shanghai) Co., Ltd. 2025) Hu 0101 Min Chu No. 14775 (November 6, 2025),[i] examined this question in considering whether copyright is capable of subsisting in prompts submitted to an image generative artificial intelligence system (a “GenAI system”).
On the facts of the case, the Chinese court, in a case of first impression, held that copyright did not subsist in six sets of prompts submitted to Midjourney and hence the copying of those prompts could not infringe copyright. However, the court suggested that there could be copyright in a more elaborate and specific set of prompts that meets copyright’s standard for originality.
Introduction
The decision arose from a copyright infringement suit brought by the plaintiff claiming that he had created prompts using the Midjourney GenAI system to generate a series of images. He claimed that copyright subsisted in the prompts. The defendant allegedly copied those prompts as they were made available to the public by Midjourney along with the resulting images generated by the prompts. The defendant used the copied prompts to generate similar images on Midjourney. The plaintiff sought to stop the use of the generated images by claiming that the use of the Midjourney prompts infringed his copyright.
Why This Decision Matters for AI and Copyright Law
With the rapid growth of generative AI, creators, businesses, and lawyers are asking an important question:
👉 Can prompts — the instructions that guide AI systems — attract copyright protection?
As courts worldwide grapple with generative AI, the legal status of prompts will directly affect creators and users of systems such as Midjourney, Stable Diffusion, and other image generation platforms. This decision offers early guidance on how originality, the so called idea-expression dichotomy, and public policy may shape future copyright disputes involving AI-generated content.
As prompt engineering becomes a valuable commercial skill, understanding its legal status is increasingly important.
In this blog post, I explore:
• Whether prompts qualify as “original works” under copyright law
• Whether prompts can qualify as literary works or compilations
• Whether prompts can qualify as copyright expression or are ideas or methods of operation that are not subject to copyright protection
A description of the Midjourney prompts.
Central to the issues in the case was whether copyright subsisted in the prompts submitted to Midjourney by the plaintiff. To help understand the reasons for decision, set out below are the prompts created by the plaintiff, an interpretation explaining what the prompts were telling the AI image generator to do, and a likely depiction of the images the prompts would have generated.[ii]
Prompt 1: Aquamarine Stygiomedusa gigantea Jellyfish
Art Nouveau style illustration of Aquamarines Stygiomedusa gigantea, by Alphonse Maria Mucha. Ancient hand-painted manuscripts, Papyrus, Complex and delicate jellyfish texture, Gorgeous gold inlaid wooden picture frame, Mirror symmetry.
Instruction Breakdown:
• Style: Inspired by Alphonse Mucha and the Art Nouveau movement—characterized by flowing lines, ornamental details, and symmetry.
• Subject: A jellyfish species (Stygiomedusa gigantea), rendered in aquamarine tones.
• Texture & Detail: Emphasis on delicate, complex textures mimicking jellyfish anatomy.
• Background: Designed like an ancient manuscript—aged papyrus, possibly with faded inks and floral or marine motifs.
• Framing: Gold inlaid wooden picture frame for a classical, gallery-like aesthetic.
Image: Aquamarine Stygiomedusa gigantea Jellyfish
Prompt 2 Turquoise Butterfly
Art Nouveau style illustration of Turquoise Butterfly, by Alphonse Maria Mucha, Hand drawn manuscript, Papyrus, Complex and delicate butterfly texture, Biological notes, Central composition, Mirror symmetry.
Instruction Breakdown:
• Style: Alphonse Mucha’s Art Nouveau—delicate curves, decorative lines.
• Subject: A turquoise-colored butterfly, centered in the frame.
• Texture: Detailed, fragile wing textures with natural patterns.
• Scientific Tone: Includes biological notes as part of a natural history manuscript feel.
• Background: Papyrus or aged paper with hand-drawn text and sketches.
• Layout: Central composition and mirror symmetry—the butterfly appears like a specimen in a naturalist’s collection
Image 2 Turquoise Butterfly
Prompt 3: Lapis Lazuli Butterfly
Art Nouveau style illustration of lapis lazuli Butterfly, by Alphonse Maria Mucha, Hand drawn manuscript, Papyrus, Complex and delicate butterfly texture, Biological notes, Mirror symmetry.
Instruction Breakdown:
• Style: Again, inspired by Mucha’s Art Nouveau.
• Subject: A butterfly rendered in lapis lazuli tones (deep blues, possibly flecked with gold).
• Texture & Detail: High detail in wing patterns, gem-like coloring.
• Background: Papyrus and manuscript-style background, complete with scientific notes.
• Layout: Mirror symmetry for a harmonious and naturalist presentation..
Image 3: Lapis Lazuli Butterfly
Prompt 4: Amethyst Gem Tree
Art Nouveau style illustration of Amethyst gem tree, by Alphonse Maria Mucha, Hand drawn manuscript, Papyrus, Notes for organisms, Encyclopedia of plants, interpretative statement, Central composition, Mirror symmetry.
Instruction Breakdown:
• Style: Art Nouveau botanical illustration with mystical or mineral-like elements.
• Subject: A tree with branches or leaves resembling amethyst crystals—a fusion of nature and gemstone.
• Scientific Tone: Includes notes for organisms and encyclopedia-style explanations—framing it as a scientific or fictional botanical entry.
• Layout: Mirror symmetry and central composition suggest the tree is the main focus, shown symmetrically as if part of a scientific poster.
• Background: Aged papyrus with manuscript annotations and interpretative statements, simulating scholarly analysis.
Image 4: Amethyst Gem Tree
Prompt 5: Emerald Ganoderma Mushroom Scene
Art Nouveau style illustration of the emerald ganoderma scene, by Alphonse Maria Mucha, Hand drawn manuscript, Papyrus, biolysical analysis, A detailed explanation of plants, Encyclopedia of Plants, Intricate, elaborate, Rich and detail texture, Center of the composition.
Instruction Breakdown:
• Style: Richly textured Art Nouveau illustration with decorative botanical framing.
• Subject: The Ganoderma mushroom (a medicinal fungus) rendered in emerald tones—possibly glowing or gem-like.
• Scientific Content: Includes biophysical analysis and encyclopedic details—framing it as a plant study.
• Texture: Intricate and elaborate textures—emphasizing the surface detail of the fungi and background.
• Layout: Centered subject, likely symmetrical or framed within floral motifs.
• Background: Papyrus manuscript with hand-drawn biological diagrams and notes.
Image 5: Emerald Ganoderma Mushroom Scene
Prompt 6: Black Opal Brocade Carp
Art Nouveau style illustration of Brocade carp in Black Opal, by Alphonse Maria Mucha, Hand drawn manuscript, Papyrus, biological analysis, Detailed description of biological signs, Encyclopedia of Marine life, Fine black koi texture details, Center of the composition.
Instruction Breakdown:
• Style: A stylized illustration of a fish in the Art Nouveau aesthetic.
• Subject: A brocade carp (a type of ornamental koi) rendered in shimmering black opal—likely combining deep blacks with iridescent color highlights.
• Scientific Detail: The background includes biological analysis, marine encyclopedia entries, and anatomical observations.
• Texture: Fine black koi texture details are highlighted—focusing on scale detail and subtle reflections.
• Layout: Centered composition—the carp should be the focus, likely curved elegantly in Mucha’s signature pose.
.
Image 6: Black Opal Brocade Carp
Can Copyright Subsist in AI Prompts? Arguments made by the Plaintiff
The plaintiff argued that the prompt words in question constituted a literary work. It made four arguments in support of copyright subsisting in the prompts.
First, the prompts were the proper subject matter for copyright. They were similar to other forms of expression such as film scripts and stage designs. As summarized by the court:
…the prompt words involved in this case belong to the fields of literature, art, and science, and possess a certain form of expression. Prompt words, especially those used for text-to-image generation, are essentially expressive creative texts, similar in nature to film scripts and stage design plans. A film script provides textual guidance for the overall story structure, shot breakdown, and atmosphere creation during filming, which directors and actors use for their specific performances. Similarly, text-to-image prompts provide the Midjourney platform with “semantic guidance” and “style clues” as a textual basis, which the Midjourney platform then uses to render specific images. Taking prompt word five as an example, it provides the Midjourney platform with the overall structure, main elements, artistic style, materials and details, scientific context, and main composition of the image through textual semantics, arranged in a specific and orderly manner.
Second, the prompts were sufficiently original under Chinese copyright law as the selection and arrangement of the prompts demonstrated sufficient intellectual effort.
Third, the prompts constituted original expression. As summarized by the court:
…the prompts involved in the case go beyond purely functional language and possess characteristics of original expression.
1. The function, usage scenarios, and characteristics of prompts determine that optimal prompts should be composed of words or phrases, and their originality is reflected in the arrangement and combination of phrases in the public domain.
(1) The function and essence of prompts. In the working principle of large language models (LLMs), text is first broken down by a tokenizer into the smallest linguistic units—tokens. A token can be a single character or a complete word. Subsequently, these tokens are converted into vectors (embeddings), which are the digital form that the model can understand and process. Vectors are the only recognizable format in neural networks and Transformer architectures. Therefore, from a technical perspective, the processing order that the model easily understands is “vector → token → word → sentence,” while the order in which humans easily understand language is the opposite: “sentence → word → smaller linguistic units.” Therefore, in actual interaction, words become the most efficient form of communication between humans and the model. To facilitate the model’s understanding, prompts are usually primarily composed of words and phrases, and the selection, combination, and arrangement of words and phrases within the prompts (that is, the author’s experience, aesthetics, and knowledge) are precisely the expression of the prompt author’s originality.
(2) The author’s original expression in this case. Regarding the expression of creative style. The creative style expression in the prompts focuses on “Art Nouveau” and “by Alphonse Maria Mucha.” The plaintiff chose Art Nouveau, a style widely used in building materials, and Alphonse Maria Mucha, who was skilled in drawing figures (mostly beautiful young women) and advertising decorations, as the style for expressing the “plants and animals” theme. ② Regarding the choice of creative theme. The creative theme follows a unique expression method of “gemstone + plant” names, such as “Aquamarines Stygiomedusa gigantea” and “Amethyst gem tree.” ③ Regarding the choice of materials and carriers. The plaintiff chose the creative method of “Hand-drawn manuscript,” and the medium involved was “Papyrus.” Among the common art mediums in the West, the plaintiff chose papyrus as the medium, matching it with fictional plants and the lines and brushstrokes of 19th-century painter Alphonse Mucha, creating a mixed style that is ancient, fantastical, and yet realistic. ④ Regarding the form of the painting, the forms involved include “biological analysis,” “A detailed explanation of plants,” and “Encyclopaedia of Plants,” which is a style of “natural illustration/natural history illustration.” The prompt words chosen by the plaintiff improved upon existing natural history illustrations, constructing a fictional, ancient style of plant notes, thereby creating an original art form.
2. Prompt words, as a means of expression, possess independence and copyright value. The prompt words themselves are the result of human language expression, similar to the conceptual drafts or outlines of literary works. Their independent creative value should not be denied due to the involvement of AI.
3. The determination of originality should focus on the method of expression and personalized choices, not the length or factual content. As long as the content reflects the author’s ideas and choices, even the shortest length can express rich content; therefore, the length of the prompt words does not affect the existence of originality. In addition, although the prompt words involve some public domain or factual information, copyright law protects not the facts themselves, but the way in which the facts are expressed.
Fourth, the prompts resulted from sufficient expression to be protected by copyright. As summarized by the court:
…the prompt in this case does not involve limited expression. Although the prompt in question contains words from the public domain, its overall structure constitutes an original expression in natural language, possessing sufficient expressive space and creative freedom. It is not a form necessarily adopted to express a specific idea or function.
1. The prompt is not a set of instructions; its expression does not belong to limited expressions based on specific formats, norms, etc.
2. The prompt is also not computer software code. The essence of the prompt is the broad expression of natural language, unrestricted by programming languages or technical functions.
3. The prompt in question is a creative prompt, not a search-based prompt. It reflects the plaintiff’s clear intention and creative choices in generating original content, belonging to the category of natural language text with a high degree of expression, rather than being limited to existing facts or commonly known expressions.
Can Copyright Subsist in AI Prompts? Arguments made by the Defendant.
The defendant made 5 arguments as to why copyright did not subsist in the prompts.
First, it was argued that the prompts were not works including literary works. As summarized by the court:
…literary works are expressed in written form, and their core lies in conveying the author’s thoughts or artistic beauty through a system of written symbols. From the perspective of the standards for literary works, the phrases involved in this case are professional terms or proper nouns in the public domain, lacking personalized expression. The order of the prompt words does not have specific meaning, and they do not constitute independent scripts or short articles with basic aesthetic significance, and therefore cannot be monopolized by the plaintiff.
Second, writing prompts is not a creative act that can satisfy copyright’s originality requirement. As summarized by the court:
…language, as a carrier of thought transmission, can serve as both a concrete expression and a tangible manifestation of thought. On the one hand, in traditional creative processes, such as painting and photography, people complete the transformation from thought to expression at the moment their creative purpose is achieved. However, AI tools divide the creative process into two stages; prompt words are only one part of the creative process, not the creative result. On the other hand, people use concrete expressions as carriers to transmit thoughts and communicate through creation, but writing prompt words is merely a human-computer interaction used to instruct AI to complete a painting. This behavior does not have social attributes and cannot serve the purpose of communication and interaction between people. Moreover, the key to constituting a creative act is the qualitative change from thought to expression, and writing prompt words is not a creative act. An increase in the quantity of non-creative activities cannot qualitatively transform the result into something original.
Third, the act of selection of prompts is not a creative act. As summarized by the court:
As for the selection of words in the prompt words, it is essentially a process of selection, arrangement, and combination. This selection belongs to the ability of appreciation, not creative ability, and the act of selection is not a creative act. Furthermore, entering prompt words is not an act of publication. For example, searching for a question on Baidu, entering search keywords is only to obtain search results.
Fourth, the prompts constitute methods which are not protected by copyright.
Fifth, the prompts should be considered as instructions that merely convey ideas, and ideas are not protected by copyright. As summarized by the court:
Prompt words are considered ideas and are not protected by copyright law. According to the idea-expression dichotomy, prompt words are the concretization of an idea, not a specific expression. In the process of AI image generation, the role of prompt words is to instruct the AI to generate content according to the user’s ideas. The U.S. Copyright Office also believes that prompt words are essentially instructions that convey ideas not protected by copyright, and therefore are not protected by copyright law.
Are Generative AI Prompts Protected by Copyright Law? The Court’s Ruling
The court found that the prompts were not protected under the Chinese copyright law because they did not meet the law’s requirements for originality and because, in any event, they were unprotectable ideas.
Originality Requirements for Copyright in AI Prompts
The court recognized that for a work to be “original” it must originate from the author’s independent creation and reflect the author’s individual choices and creative expression. Applying these and other principles related to the requirements for originality, the court expressed the opinion that some series of prompts could be protected by copyright while others might not be depending on the degree of creative intellectual input. The more specific and semantically clear the user’s prompt is, the more defined the theme and style of the generated image will be and the more likely the prompts will be protected by copyright. According to the court:
In traditional creative work, originality requires that the author’s intellectual input in terms of choices, arrangements, and design reaches a level that reflects individuality. From this perspective, when evaluating the originality of prompts in AI-assisted creation, the focus should be on whether the user demonstrates substantial intellectual creativity in the design and adjustment of the prompts, and whether this reflects personalized expression. Generally, the more specific and detailed the prompt, the more fully it reflects personalized choices and aesthetic preferences, and thus the more likely it is to meet the requirements of originality. However, originality does not depend on the length or complexity of the prompt, but rather on whether it reflects the result of creative expression. It is worth emphasizing that human input must be “creative, not merely laborious.” Even if a significant amount of time is spent repeatedly modifying prompts, if it only involves rearranging generic prompts or simply combining common stylistic words, it is usually difficult to consider it as original expression. Conversely, if the prompt is the result of purposeful conceptualization and semantic design, and can present a unique aesthetic orientation or creative expression, it can be considered as creative intellectual input.
The Midjourney platform involved in this case is an image generation tool based on generative artificial intelligence. It utilizes deep learning algorithms such as diffusion models, learning from large-scale image-text paired data to achieve automatic conversion from language prompts to image generation. In the Midjourney model, the prompt structure is divided into image prompts, text prompts, and suffix parameters, resulting in three scenarios: text prompts, text prompts + suffix parameters, and image prompts + text prompts + suffix parameters. Text prompts are usually the main controlling element affecting the generated results. Text prompts can be a single word, or a complete phrase or sentence. The more specific and semantically clear the user’s prompt is, the more defined the theme and style of the generated image will be. Short or vague prompts will cause the Midjourney platform to automatically fill in details using its default style. Due to the specific nature of the model in this case, the prompts are limited to words or phrases.
The court then examined the prompts in issue and held they lacked originality. While not using terminology from Canadian or U.S. copyright law, the court found the prompts to be unprotectable “stock devices” or “scenes a fair” – being standard or conventional artistic styles. It also ruled they lacked both structural progression and a narrative sequence, and a logical coherence when combined, to constitute a sufficient degree of original expression. According to the court:
The six sets of prompts involved in the case adopted the basic structure of artistic style, main elements, materials and details, scientific context, and main composition. Among them, the artistic style included “Art Nouveau style” and “by Alphonse Maria Mucha,” the main elements included “Aquamarines Stygiomedusa gigantea” (Giant Aquamarine Stygian Jellyfish), “Turquoise Butterfly,” “Lapislazuli Butterfly,” “Amethyst gem tree,” “the emerald ganoderma scene,” and “Brocadecarp in Black Opal,” the material was “Papyrus,” the details included “Ancient hand-painted manuscripts,” “Intricate,” “elaborate,” “Rich and detailed texture,” and “Gorgeous gold inlaid wooden picture frame,” the scientific context included “Biological notes,” “Notes for organisms,” “Encyclopedia of plants,” “interpretative statement,” “biophysical analysis,” and “A detailed explanation of plants,” and the main composition included “Mirror symmetry” and “Central composition.”
Based on the criteria for determining copyrightable works, the six sets of prompts in question are essentially instructions or descriptions entered by the user into the AI system to guide the AI in generating specific images. Formally, although they contain various elements, these elements are simply listed without grammatical or logical connections, failing to form a language expression with an inherent structure; the keyword groups are in a disordered combination, lacking both structural progression and a narrative sequence. For example, phrases like “Ancient hand-painted manuscripts” and “Complex and delicate jellyfish texture,” while each containing certain semantic information, still lack logical coherence when combined. From the perspective of originality of expression, these prompts lack the author’s personalized characteristics; the artistic styles, material details, or main compositions used are all conventional expressions in the field, failing to reflect the author’s unique aesthetic perspective or artistic judgment. The plaintiff argues that their use of a new artistic style applicable to building materials and the style of a painter skilled in drawing figures as the expression style for “animals and plants” themes, as well as the naming method of “gemstones + animals and plants” for the creative theme, demonstrates their originality. In fact, in the field of traditional painting, it is not uncommon for artists to transfer familiar styles to different subjects, and combining gemstones with animals and plants based on visual (color, form) or cultural symbolic associations is also a common creative technique, making it difficult to constitute original expression.
Furthermore, although the law does not set a word count threshold for literary works, a work still needs to possess completeness of expression to meet the requirements of originality and a certain depth of expression. The prompts in question, such as “Hand drawn manuscript” and “Biological notes,” are merely combinations of a few phrases, lacking a complete linguistic structure and depth of expression, and cannot fully express complete content through sufficient text like formal works. At the same time, these phrases only simply inform the AI of the image style and scene, without providing a deeper interpretation or assigning special meaning to the style and scene, nor do they demonstrate the necessary artistic beauty. In summary, although the prompt in question contains some descriptive content, its expression is insufficient to demonstrate creative intellectual input, and therefore it should not be considered an original literary work.
AI Prompts as Ideas Versus Expression
The court also considered the prompts to be mere ideas for the creation of artistic works rather than expression. However, it also stated that if prompts went beyond pure functional instructions to “unique language choices and artistic expression” the prompts could be considered a protected work with expressive attributes. According to the court:
Ideas are abstract and not protected by law, such as thoughts, concepts, theories, designs, creative ideas, etc., while expressions are concrete and can be protected by law. That is, the law only protects the expression of a work, not the idea itself. In the context of AI-generated content, prompts serve as both a medium for users to convey their creative intentions to the AI system and a bridge connecting human thought with AI output. In most cases, the function of prompts is to convey creative requirements, such as “write an article about snow” or “create a painting of snow.” These types of prompts are instructions regarding the creative theme and style, and are usually considered to fall under the category of “ideas.” However, if a user writes text with literary originality, such as “Suddenly, like a spring breeze overnight, thousands of pear trees blossomed,” and asks the AI to further create based on this, because this prompt goes beyond pure functional instructions and reflects unique language choices and artistic expression, in this case, the prompt may be considered a protected work with expressive attributes.
In this case, the core of the prompts in question is a list and description of visual elements, artistic style, and presentation form. This content is more akin to abstract creative ideas, specifically the idea of “what kind of painting I want the AI to generate.” This idea, by its nature, belongs to the realm of thought. Only the specific ways in which lines and colors are used to depict the form, details, and stylistic characteristics of the “Turquoise Butterfly,” or the emotional atmosphere the painting conveys, could constitute “expression.” The prompts in question, remaining at the level of abstract creative ideas rather than concrete expression, should not be considered works protected by copyright law.
Public policy favors not protecting short sets of prompts
The interpretation of copyright laws in many countries is often informed by the policies or purposes of the copyright law. In giving reasons for decision, the Chinese court opined that protecting the prompts in issue was not in the public interest as it would adversely affect the balance in copyright that encourages creativity. According to the court:
The purpose of copyright law is to encourage creativity, thereby promoting the development of culture and science. If short instructions, keyword combinations, and similar types of prompts are recognized as works and protected by copyright law, it could lead to two potential risks. On the one hand, the free use of language will be restricted, potentially leading to the excessive privatization of linguistic resources. Once similar prompts are recognized as protected works, others using the same or similar word combinations to guide AI content generation may face the risk of infringement, and creativity will be suppressed. On the other hand, the AI innovation ecosystem will be constrained. Prompts are important guiding factors for AI-generated content; rich and diverse prompts can stimulate AI’s creativity and promote AI exploration in a wider range of fields. If excessive rights boundaries are placed on the use of prompts, the instructions and data available to AI will be drastically reduced, hindering technological innovation and the development of multi-field applications.
In summary, based on a systematic interpretation of the boundaries of copyright law and a consideration of the balance of public interest, the prompt in question reflects a certain creative intent, but it does not demonstrate the author’s individualized intellectual contribution at the level of expression. Therefore, it should not be considered an original work.
Comments
The Chengdu Cultural Communication case is a case of first impression that, once again, attempts to apply principles of copyright law to novel issues, in this case, whether copyright can subsist in prompts submitted to an image generation GenAI system. The reasons for decision raise a number of questions, resolves some of them, and also leaves open whether courts outside of China would reach similar results. By way of example only:
The originality requirement prompts to a GenAI system
The decision summarizes the requirements for originality under the Chinese copyright law, but the laws vary somewhat around the world. The court described the standard as requiring “substantial intellectual creativity” and whether it reflects “personalized expression”. This formulation of the requirements for originality might well be recognized under the copyright laws of the European Union. However, this formulation does not necessarily align with the originality requirements in other countries.
For example, “originality” under U.S. law, as reflected in the leading U.S. Supreme Court decision in Feist, requires s minimal degree of creativity but does not emphasize any requirement for personalized expression.
Under the leading Supreme Court of Canada decision in CCH v Law Society, for a work to be “original” it must be more than a mere copy of another work. At the same time, it need not be creative, in the sense of being novel or unique. What is required to attract copyright protection is an exercise of skill and judgment. “Skill, refers to “the use of one’s knowledge, developed aptitude or practised ability in producing the work.” Judgment, refers to “the use of one’s capacity for discernment or ability to form an opinion or evaluation by comparing different possible options in producing the work.” This exercise of skill and judgment will necessarily involve intellectual effort. The exercise of skill and judgment required to produce the work must not be so trivial that it could be characterized as a purely mechanical exercise.
As the Chinese decision stated, the analysis as to whether there has been a sufficient degree of intellectual effort is a question of fact. While this is likely also to be the case under other copyright laws, if the tests for originality differ, the analysis of the facts might also be different.
The requirement that a literary work have a linguistic structure
The decision applies a requirement that literary works have a “linguistic structure and depth of expression” or what the court also seemed to refer to as “a language expression with an inherent structure”, or a “structural progression and a narrative sequence”. Under the classic formulation of literary work under Anglo-Canadian law,[iii] a literary work is the expression of thought in print or writing. Whether laws outside of China will require the additional requirements set out in the Chinese decision will need analysis on a country or country (or jurisdiction by jurisdiction) basis.
As well, it is unclear whether the court fully appreciated the scope of the instructions given to Midjourney with respect to style, subject, scientific content, tone, or detail, framing, texture, layout, and backgrounds. Each of the words in the prompts would have conveyed more information which Midjourney was capable of interpreting. The likely interpretations for each words in the prompts, reviewed in this context, might well have been a progression of or sequence of words in print or in writing. Also, the court did not consider whether the requirements it formulated had to be true of compilations These frequently do not meet the requirements the court applied for other types of literary works. The sets of prompts may or may not have been protectable under copyright for other reasons, but may have met the requirement of being a literary work.
Are prompts to a GenAI system methods of operation?
Under international agreements related to copyright, such as Article 2 of the WIPO Copyright Treaty and Article 9(2) of the TRIPs Agreement, copyright protection extends to expressions and not to ideas, procedures, methods of operation or mathematical concepts as such. In Lotus Development Corp. v. Borland Intern., Inc., 49 F. 3d 807 (1st.Cir.1995) a U.S. appeals court denied protection for Lotus’ menu command hierarchy because it provided the means by which users controlled and operated Lotus 1-2-3. Similarly, in the European Union computer program artifacts such as programming languages, syntax, commands and combinations of commands are not protectable under copyright. See, SAS Institute Inc. v World Programming Ltd, Case C-406/10. A court outside of China might focus on these limits to copyright protection in reviewing whether copyright is capable of subsisting in a series of prompts that are used to instruct the operation of the Midjourney GenAI system.
Are Prompts to a GenAI system a protectable compilation under copyright law?
The decision acknowledges that copyright can subsist in original compilations. The court’s focus, however, was on whether copyright could subsist in each of the six sets of prompts. It did not consider whether there was any original compilation in the collection of the sets of prompts.
Would copyright subsist in images generated by prompts to a GenAI system?
The decision did not canvass what would have happened if copyright was found to subsist in the plaintiff’s prompts and the defendant had sought to create images similar to those generated by the prompts of the plaintiff by using independently created prompts.
It is entirely possible that a person may have a copyright in prompts submitted to a GenAI system but not in the outputs generated by the GenAI system. The U.S. Copyright Office takes the position that given the state of the art GenAI users do not have sufficient control over AI generated outputs to be the author of the outputs. While this categorical view may not be accurate, it does highlight that while the question of copyright in prompts and outputs may be related, copyright in the former does not guarantee copyright in the latter. See, Understanding the Copyrightability of AI: Insights from the U.S. Copyright Office, https://barrysookman.com/2025/02/03/understanding-the-copyrightability-of-ai-insights-from-the-u-s-copyright-office/.
FAQ and Summary of the Chinese Court Ruling
Are AI prompts protected by copyright law?
Courts are beginning to examine whether AI prompts can qualify as protected literary works. In this Chinese decision, the court held that short and descriptive prompts lacked sufficient originality and were more akin to unprotectable ideas.
Can copyright subsist in detailed generative AI prompts?
The court suggested that more elaborate and specific prompts reflecting creative intellectual input and original expression could potentially qualify for copyright protection.
Are prompts considered ideas or expression in copyright law?
The court found that prompts in issue conveying ideas rather than protected expression. However, according to the decision, creative language choices may in some circumstances rise to the level of expression.
Does copyright in prompts affect ownership of AI-generated images?
The decision did not address whether copyright could subsist in AI-generated outputs, leaving open important questions about authorship and originality in AI generated outputs.
Related AI and Copyright Law Analysis
Understanding the Copyrightability of AI: Insights from the U.S. Copyright Office, https://barrysookman.com/2025/02/03/understanding-the-copyrightability-of-ai-insights-from-the-u-s-copyright-office/
Copyright Infringement and AI: Insights from Getty v Stability AI, https://barrysookman.com/2025/11/05/copyright-infringement-and-ai-insights-from-stability-ai-v-getty/
AI Visual Effects and Copyright Infringement, https://barrysookman.com/2025/06/30/ai-visual-effects-and-copyright-infringement/
AI Copyright and Human Authorship: Thaler v Perlmutter, https://barrysookman.com/2025/03/23/ai-copyright-and-human-authorship-thaler-v-perlmutter/
Copyright AI: Understanding Substantive Legal Developments, https://barrysookman.com/2025/01/18/copyright-ai-understanding-substantive-legal-developments/
Endnotes:
[i] This case summary is based on a Google translate of the case found at the link above. The translation is not an official translation and has not been verified.
[ii] The prompts are taken from the Google translation of the decision. The explanations of the prompts were generated using ChatGPT, and the simulated images were generated using DALL-E and not Midjourney.
[iii] See, University of London Press Ltd. v. University Tutorial Press Ltd., [1916] 2 Ch. 601 (Ch.D.).
