/>iIn a decision issued[1] November 27, 2023, a Chinese court ruled that AI-generated content can enjoy protection under copyright law. The finding, the first of its kind in China, is in direct conflict with the human authorship requirement under U.S. copyright law and may have far-reaching implications.
Background of Case
The plaintiff used a U.S.-based text-to-image artificial intelligence (“AI”) service to generate an image of a woman that he posted to a lifestyle platform called Xiaohongshu, also known as “Little Red Book”. A China-based blogger then used the AI-generated image in a blog post on Baijiahao, a Chinese blogging platform, without the plaintiff’s permission. The plaintiff sued for copyright infringement in May 2023.
The plaintiff argued that the “intellectual investment” of selecting and arranging the inputs necessary to generate the image at issue is akin to photographers manually adjusting antique cameras to achieve a desired result in a photograph. In that scenario, it is universally understood that the photographer is the author of the resulting copyright-protected photograph. Here, the analogy designates generative AI systems as mere tools for humans to create expressive works rather than the AI systems creating such works.
The Beijing-based court agreed and ordered the defendant to publicly apologize and pay the plaintiff 500 Yuan (USD $70) in damages and 50 Yuan (USD $7) in court fees.
The Beijing Internet Court found that the AI-generated image is an artwork that is subject to copyright protection by holding, “as long as an AI-generated image reflects the original intellectual investment of a human being, it should be considered a work that is protected by copyright laws”.
In arriving at its novel holding, the court emphasized that the plaintiff “made a certain degree of intellectual investment” in selecting and arranging the inputs (i.e., a series of creative prompts and parameters that generative AI users feed into the AI system to facilitate an output based upon such inputs). According to the court, the plaintiff-selected inputs were sufficiently original; thus, the output (i.e., the AI-generated image) met the “intellectual achievements” and originality thresholds necessary to enjoy copyright protection. The tech-friendly court strongly emphasized that this decision aligns with the legislative purpose of China’s copyright law. The court reasoned that extending copyright protection to AI-generated content will incentivize individuals to utilize AI for creative purposes, leading to an increased production of creative works.
However, this ruling has sparked intense controversy within the academia in China. If the defendant appeals to the Beijing Intellectual Property Court, the trial court’s apparent misunderstanding of the idea/expression dichotomy seems ripe for challenge. According to the Copyright Law of PRC,only original expressions of ideas – not ideas themselves – may be afforded copyright protection. As for pictorial works like in this case, originality in expression should be embodied by the selection and arrangement of various key elements of an image such as lines, colors, and shapes. The defendant could argue that the plaintiff’s input prompts (e.g., “outdoor environment,” “Japanese idol,” “highly detailed, symmetrical, attractive face,” “perfect skin,” “dreamy black eyes,” “reddish-brown plaits,” “long legs,” “golden hour,” “vivid colors,” “shy,” “graceful,” “cool pose,” “teen,” and “lust”) is merely an idea rather than expression of such idea because the prompts only describe an output, rather than express the necessary elements of the resulting image created by the AI system.
Comparison with U.S. Copyright Approach
The U.S. Copyright Act protects only “original works of authorship”[2] and courts have long held that only humans qualify as authors.[3] With the explosion of new generative AI tools, the U.S. Copyright Office has reaffirmed its human authorship requirement for copyright protection.[4]
The U.S. Copyright Office has notably denied (or limited the scope of) copyright registrations for works lacking human authorship (e.g., “A Recent Entrance to Paradise,”[5] “Zara of the Dawn”,[6] and “Théâtre D’opéra Spatial”[7]). In siding with the U.S. Copyright Office’s position, U.S. District Courts have explained that “The [Copyright] Act’s ‘authorship’ requirement as presumptively being human rests on centuries of settled understanding.”[8] (emphasis in original).
Due to the human authorship requirement under U.S. copyright law, the scope of copyright in a purely AI-generated work is limited to, at most, the unique arrangement of inputs[9] – but not the output (e.g., photograph, video, song). When a work combines both human-created and AI-created elements, the scope of the copyright is limited to the human-created elements after filtering out the other non-human elements[10] (e.g., human created text accompanying AI-generated images).
Notably, the “Théâtre D’opéra Spatial” case indicates the main split between the Beijing Internet Court and the U.S. Copyright Office concerning the level of human involvement sufficient for copyright protection of an expression. There, the individual used a similar text-to-image AI services and went through an even more complex process in an effort to produce the image (in addition to creating multiple input prompts, he also used photo editing tools to polish the image). In contrast to the more lenient stance taken by the Chinese court where creation, selection, and arrangement of prompts and parameters, among others, suffices, the U.S. Copyright Office, in denying the applicant’s registration request, determined that his sole contribution to the AI services was the input of the text prompt and that he did not generate any traditional elements of authorship in the resulting output (e.g., literary, artistic, or musical expression; elements of selection and arrangement).
Another interesting comparison lies in their contrasting interpretations of AI’s role in the creative process. The U.S. Copyright Office understands that current AI services do not understand “grammar, sentence structure, or words like humans” and therefore are unable to “interpret prompts as specific instructions to create a particular expressive result”. In comparison, the Chinese court believes that generative AI services function like humans who have gained skills and capabilities through learning, and therefore possess the capability of producing, for example, an image based upon the text prompt. In essence, AI must be viewed as a tool for human creation by assuming the human’s role of drawing lines, filling in colors, and presenting the human’s ideas and creativity in a tangible medium.
Impact
While the Chinese court’s reasoning is controversial, the outcome can be viewed, to some extent, as facilitating the enforcement of intellectual property rights and potentially deterring copyright infringement within China. Granting authorship recognition to those harnessing the power of generative AI empowers them to intervene in infringements through the judicial system, such as this case, or via the takedown notice mechanisms of various online platforms. Had the court ruled differently, it could have paved the way for plagiarizers to appropriate AI-generated works without accountability, potentially exacerbating the prevalence of plagiarism.
However, the Beijing Internet Court’s holding regarding the copyrightability of AI-generated images could have lasting precedential value, despite the court cautioning that future disputes should be judged on a case-by-case basis. If upheld, we are sure to see a ripple effect on other segments of the creative industry like music, film, and television.
Undoubtably a tech-friendly decision, the court’s ruling also notably aligns with a key policy of the Chinese government. Indeed, Beijing has long provided financial and political support for companies advancing its domestic AI industry.[11] Prior to the Beijing Internet Court’s ruling, China’s exploding AI market was projected to exceed USD $26 Billion by 2026.[12] In a paradoxical twist, a USD $70 court decision could cause that valuation to skyrocket if other courts follow the Beijing Internet Court’s lead.
FOOTNOTES
[2] (17 U.S.C. § 102(a))
[3] Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 56 (1884)
[4] https://www.copyright.gov/ai/
[5] https://www.copyright.gov/rulings-filings/review-board/docs/a-recent-entrance-to-paradise.pdf
[6] https://www.copyright.gov/docs/zarya-of-the-dawn.pdf
[7] https://www.copyright.gov/rulings-filings/review-board/docs/Theatre-Dopera-Spatial.pdf
[8] https://ecf.dcd.uscourts.gov/cgi-bin/show_public_doc?2022cv1564-24
[9] Feist Publications, Inc., v. Rural Telephone Service Co., 499 U.S. 340 (1991)
[10] https://www.copyright.gov/docs/zarya-of-the-dawn.pdf
[11] https://cset.georgetown.edu/article/in-out-of-china-financial-support-for-ai-development
[12] https://www.idc.com/getdoc.jsp?containerId=prAP50688623
This article was co-authored by Zach Dai.
