The explosion of Generative AI (GAI) tools has brought the complex question of AI authorship to a head. When a machine creates an image or a complex piece of text, who owns it? While various jurisdictions have adopted nuanced positions, the fundamental legal principle remains the same: human creative contribution is the required doctrinal touchstone for conferring copyright on AI outputs.
A recent judicial decision in China—a country known for demonstrating doctrinal flexibility regarding human agency in shaping AI outputs—has solidified a stricter, evidence-based approach to this issue.
The Stricter Standard: China’s 2025 Denial
In March 2025, the People’s Court of Zhangjiagang City, Jiangsu Province, issued a judgment denying copyright protection to AI-generated images in the case involving Ms. Feng and several defendants.
The plaintiff, Ms. Feng, a designer, published a series of “Fantasy Wings Transparent Art Chair” images in August 2023, generated using the AI illustration tool Midjourney. She initiated legal action claiming copyright infringement when defendants, including Dongshan Company and Ms. Zhu, copied her designs for commercial products.
The key defense argument centered on the lack of human originality associated with the creation of the work. For example:
- Simple Prompts are Insufficient: Inputting prompt words (“prompts”) conveys themes or ideas, not specific creative expression.
- Unpredictable Output: The actual output of the large language model (LLM) is unpredictable and not controlled by human will, meaning the output cannot be attributed to the user as their own work.
- The Machine Determines Substantive Elements: The use of AI models does not fit the traditional “human + tool” creation model, as the model, not the user, determines the substantive elements of the output.
The court sided with the defendants, concluding that Ms. Feng’s images do not constitute copyrightable works. The ruling highlighted a trend toward stricter originality proof.
The core reasons for the denial were evidence-based:
- Lack of Original Records: Ms. Feng failed to provide original records or workflows evidencing individualized, creative input for the disputed images.
- Randomness and Non-Repeatability: Ms. Feng acknowledged the randomness and non-repeatability of the AI generation process, making it impossible to reproduce the exact images.
- Insufficient Originality: In the absence of documented original records and individualized creative choices, the court found insufficient evidence that the images embodied the requisite originality.
This 2025 decision underscores a shift: while Chinese courts allowed authorship claims when human agency was present, they are now performing a fact and technology driven analysis that scrutinizes the process’s inherent randomness and the absence of clear records evidencing creative input.
The Universal Rule: Human Control is Non-Negotiable
This new emphasis on documented creative control in China aligns with the fundamental philosophical approach adopted by other major jurisdictions, all of which reject the idea of autonomous machine authorship.
United States: Protection is available for human-authored expression in or around AI output. Federal courts and the U.S. Copyright Office affirm that outputs created autonomously by machines fall outside copyright protection. Simple prompting alone is not enough; protection requires clear evidence of human control over expressive elements, such as selection, coordination, arrangement, or expressive post-editing.
Japan: Authorship is limited to natural persons who express creative ideas, and AI itself cannot be considered an author. Japan’s Copyright Act includes limitations that allow exploitation for AI training if the purpose is not to enjoy the thoughts or sentiments expressed in the work (the non-enjoyment purpose).
European Union: To receive copyright protection, the work must be the ‘original’ work of the author’s own intellectual creation. Works generated by machines without meaningful human input do not qualify for protection.
Practical Takeaways for Creators and Practitioners
The takeaway from the Chinese ruling and the consistency of international standards is clear: if you intend to claim copyright in AI-assisted output, you must prove that the human was the creative force.
For practitioners and companies developing AI-assisted content, the necessity for auditable provenance logging has never been more critical. To ensure that a human authorship claim is viable, implement the following:
- Maintain Detailed Logs: Keep records evidencing the human role in prompt design, selection criteria, post-editing, and curatorial decisions.
- Emphasize Human Intervention: Design systems that incorporate human selection, editing, and approval steps to solidify claims of human authorship where appropriate.
- Show Perceptible Input: Focus on expressive interventions that make the human inputs perceptible in the output.
The 2025 Chinese ruling serves as a reminder that in the face of autonomous AI generation, the burden of proof rests on the human user to demonstrate the unique creative choices necessary to meet the originality threshold. Pressing a button and claiming the result is no longer enough; the documentation of how the human shaped the machine’s output is now the arbiter of ownership.
Conclusion
The global scrutiny of AI authorship reveals a consistent principle across jurisdictions: copyright protection requires documented human creative control. The 2025 Chinese court decision reinforces this standard by demanding clear evidence of human intervention in the creative process. For anyone working with AI-generated content, maintaining detailed records of human input is essential to establish and defend authorship claims.
Justin Miller is a solo patent attorney. In 2025 he started his own law firm, Distinct Patent Law, after nearly 15 years of practice. His firm is located in Saint Petersburg Florida. He serves clients in Tampa Bay, and because patent law is federal, can file patent applications for clients all over the United States.