AI-Generated Content Falls Short of Copyright Protection, Study Finds
In a landmark exploration of intellectual property in the digital age, a recent legal analysis has concluded that content generated by artificial intelligence lacks the essential qualities required for copyright protection under current law. The study, authored by Manchun Gong from Qinghai Minzu University, centers on the pivotal 2018 “Feilin” case, where a Beijing-based law firm sued tech giant Baidu for reposting an AI-assisted article without permission. The dispute, which unfolded in the Beijing Internet Court, has since become a cornerstone in the global debate over whether machine-generated works can be considered original creations deserving of legal protection.
At the heart of the case was an article published on the WeChat official account of Beijing Feilin Law Firm. The piece, which included data analysis and automatically generated charts, was later reposted on Baidu’s Baijiahao platform without attribution or consent. Feilin Law Firm claimed infringement of its rights to dissemination, integrity, and authorship, seeking both compensation and a public apology. Baidu, however, countered that the content—particularly the charts—was produced by software using pre-existing data, with no original intellectual input from the firm. The crux of the defense hinged on the argument that the article lacked the requisite originality to qualify as a copyrightable work.
The initial ruling sided with Baidu, asserting that the graphical elements in the article were not the product of creative human input. The court reasoned that the visual outputs were determined solely by the data fed into the software, with no evidence of manual design or aesthetic intervention. The differences between such graphs were attributed to data variation, not creative authorship. Feilin appealed, but the higher court upheld the decision, emphasizing that no new evidence had been presented to demonstrate human-led creative adjustments to color, line, or composition. The final verdict was clear: the AI-generated components did not meet the threshold for protection as graphic works under China’s Copyright Law.
This outcome has sparked a broader academic inquiry into the nature of creativity in the age of automation. Gong’s study, published in Technology and Law, rigorously examines the two foundational pillars of copyright: originality and intellectual achievement. Her analysis reveals a critical disconnect between the appearance of creativity and its legal substance when AI is involved. While AI-generated content may resemble human-authored works in form and function, the process behind its creation fails to satisfy the legal criteria that underpin intellectual property rights.
Originality, as defined in copyright jurisprudence, is not merely about being different—it requires both independent creation and a minimal degree of creativity. In China, the Regulations for the Implementation of the Copyright Law define creation as “intellectual activities that directly produce literary, artistic, and scientific works.” This definition implies a human agent engaged in a deliberate, cognitive process. Gong argues that the current legal framework evaluates originality through a dual lens: subjective and objective standards.
The objective standard assesses the final output—its structure, expression, and novelty relative to existing works. It asks: Does the work, as it stands, exhibit a minimal level of creativity? This standard is neutral to the creator’s identity; it does not matter whether a poem was written by a person or a machine, as long as it meets the threshold of original expression. By this measure, many AI-generated texts, images, and even music could qualify as “works.” They often display coherence, stylistic consistency, and apparent innovation. In fact, some AI systems have produced poetry and prose indistinguishable from human writing in blind tests.
However, the subjective standard shifts focus to the process of creation. It examines the role of the author—the intellectual labor invested in conceiving, shaping, and refining the work. This is where AI-generated content falters. Unlike a human writer who draws on experience, emotion, and intention, an AI operates through algorithmic computation. It does not “think” or “create” in the human sense; it processes data according to pre-programmed rules and statistical models. The inputs—keywords, datasets, training parameters—are configured by developers, but the generative act itself is mechanical, not cognitive.
Gong illustrates this distinction with a compelling analogy. Consider a three-year-old child scribbling a simple poem. Though rudimentary, the act involves intention, motor control, and nascent linguistic awareness—hallmarks of human intelligence. The child is engaged in an intellectual activity, however basic. In contrast, when an AI “writes” a poem, it is executing a sequence of mathematical operations. The poetic style it emulates was derived from analyzing millions of human-written texts, but the AI does not understand the meaning of the words it produces. Its “creativity” is an illusion—a sophisticated pattern-matching exercise.
This distinction is crucial for copyright law, which is fundamentally designed to incentivize human creativity. The legal system rewards individuals and organizations for their intellectual effort, encouraging innovation by granting temporary monopolies over their creations. But when the creative process is outsourced to machines, the traditional link between labor and ownership breaks down. Who, then, should be recognized as the author—the user who typed the prompt, the programmer who built the algorithm, or the AI itself?
Gong’s research dismisses the notion of AI as an autonomous creator. She emphasizes that the intellectual effort in AI systems is embedded in the development phase, not the generation phase. The algorithms, neural networks, and training datasets are themselves products of human ingenuity and are already protected under copyright as software. To extend protection to the outputs of these systems would be, in effect, double-dipping—a form of overreach that could stifle competition and innovation.
Moreover, the reproducibility of AI outputs undermines the very notion of originality. If two users input the same prompt into the same model, they will receive nearly identical results. This level of predictability contrasts sharply with human creativity, where even identical prompts yield diverse interpretations. The lack of meaningful variation suggests that AI-generated works are not independently created in the legal sense but are instead algorithmic derivatives.
The implications of this analysis extend far beyond a single court case. As AI tools become ubiquitous in journalism, design, music, and literature, the question of ownership grows increasingly urgent. News organizations are using AI to draft reports, artists are generating digital art with machine learning models, and musicians are composing songs with algorithmic assistance. Without clear legal guidelines, disputes over authorship and royalties are inevitable.
Some jurisdictions have begun to address these challenges. In the United States, the Copyright Office has consistently denied registration to works created solely by machines, reaffirming that copyright protection extends only to “the fruits of intellectual labor” that “are founded in the creative powers of the human mind.” Similarly, the European Union’s Copyright Directive emphasizes the role of human authors, though it allows for limited exceptions in cases of computer-assisted creation.
China’s approach, as reflected in the Feilin case, aligns with this human-centric model. The courts have drawn a firm line between human-led and machine-driven creation, refusing to extend authorship to non-human entities. This stance reinforces the idea that copyright is not just about the end product but about the process—and the person—behind it.
Yet, the debate is far from settled. Proponents of AI copyright argue that denying protection could discourage investment in AI technologies and limit the dissemination of valuable content. They suggest hybrid models where the user of the AI system is recognized as the author, provided they exercise sufficient creative control over the output. For instance, if a writer uses AI to generate a first draft but extensively edits, restructures, and refines the text, the final work could be seen as a collaborative effort with sufficient human input to warrant protection.
Gong acknowledges this possibility but cautions against blurring the lines too much. She warns that overextending copyright to AI outputs could lead to a flood of low-effort, machine-generated content claiming legal protection, overwhelming the system and diluting the value of truly original works. Instead, she advocates for a clear distinction: AI tools are instruments, like cameras or word processors, and their outputs should be treated as such unless significantly transformed by human creativity.
The study also raises ethical and economic questions. If AI-generated content cannot be copyrighted, how should it be governed? Should it be treated as part of the public domain, freely available for use and adaptation? Or should new forms of protection be developed—perhaps a sui generis right tailored to machine-generated works? These are complex policy issues that require interdisciplinary collaboration among legal scholars, technologists, and policymakers.
One potential solution lies in clarifying the threshold for human involvement. Legal frameworks could establish criteria for when AI-assisted works qualify for copyright—such as requiring a minimum level of human modification, editorial judgment, or conceptual input. This would preserve the integrity of copyright while accommodating the realities of modern content creation.
Another approach is to strengthen protections for the training data and algorithms themselves, ensuring that developers are rewarded for their foundational work without extending undue rights to the outputs. This could involve enhanced trade secret laws, licensing models, or even data ownership frameworks.
Ultimately, Gong’s research underscores a fundamental truth: the law must evolve alongside technology, but not at the expense of its core principles. Copyright exists to promote progress by rewarding human creativity. While AI is a powerful tool, it is not a creator in the legal sense. Recognizing this distinction is essential to maintaining a fair, functional, and forward-looking intellectual property system.
As artificial intelligence continues to reshape the creative landscape, the Feilin case serves as a cautionary tale and a call to action. It reminds us that behind every algorithm is a human hand—and that the law must remain focused on the people who drive innovation, not the machines that execute it.
The study also highlights the importance of legal education and public awareness. As AI tools become more accessible, users must understand the limitations of their creations under copyright law. A blogger using an AI to generate posts, a designer using generative models for logos, or a musician using AI for composition—all need to know whether their work is protectable and under what conditions.
In this context, Gong’s work is not just an academic exercise but a practical guide for navigating the uncharted waters of digital authorship. It provides a clear analytical framework for assessing the copyright status of AI-generated content, grounded in established legal principles yet responsive to technological change.
Looking ahead, the conversation is likely to expand beyond copyright into related areas such as patent law, trademark, and data rights. As AI systems become capable of inventing new products, designing brands, and analyzing consumer behavior, the legal system will face even more complex challenges. The principles established in cases like Feilin may serve as a foundation for broader reforms.
For now, the message is clear: artificial intelligence can mimic creativity, but it cannot replace the human mind. The law, as it stands, reflects this reality. Whether future revisions will alter this balance remains to be seen. But as long as copyright is rooted in the idea of intellectual labor, the author will remain, fundamentally, a person.
Manchun Gong, Qinghai Minzu University, Technology and Law