AI Anime Short Dramas Obtain the First Data Intellectual Property Certificate: How the Creative Process Becomes Evidence for Rights Protection

In early October, a company in Guizhou that makes anime short dramas received a certificate that did not look much like an award. The Data Intellectual Property Registration Certificate, issued by the Guizhou Provincial Intellectual Property Office, registered the creative process data of the drama “The Regent Prince's Beloved Has Been Reborn,” from project initiation to final cut. The finished video itself did not appear in this registration. According to Cover News, this is the first case in China of registering the creative process of an AI anime short drama as intellectual property.

Placed against this year's industry backdrop, its significance becomes clearer. Wang Xiaoliang, director of the Network Audiovisual Program Management Department of the National Radio and Television Administration, said at a September press conference that in the first eight months of this year, 430,000 micro-dramas went online, 13 times the total for all of last year, and more than 90% of them were AI dramas. Output changed by an order of magnitude within half a year, and disputes followed by an order of magnitude: this year alone, 68,000 non-compliant micro-dramas were taken down across the internet, and more than 1,200 violating accounts were handled, over 90% of which involved AI dramas.

A Problem Drowned Out by Output

Let's start with the problem to be solved. AI has pushed the threshold for content production very low, so the marginal cost of “content laundering” has also approached zero. Extract the core logic, character relationships, and story framework of a hit, throw them to a model to rewrite, change the dialogue, alter a few names, and another drama can be produced. The industry calls this “changing the skin but not the bones.”

Public reports already contain a number of precedents. A company in Hangzhou, without authorization, extracted the story outline, core plot, and character settings from the novel “Our Girls' Dormitory Survives the Apocalypse” to produce an AI short drama; it later publicly apologized and removed the work from all platforms. In the adaptation case involving “Peach Blossoms on Horseback, Please Take Up the Long Tassel,” the Ningbo Intermediate People's Court found that the short drama “General! Madam Has Petitioned for Divorce by Imperial Decree” had been adapted and disseminated online without authorization, infringing the original work's rights of adaptation, production, and information network dissemination. In addition, AI face-swapping and voice cloning have become common sources of disputes. This year, Dilraba, Yu Shuxin, and other artists spoke out one after another, accusing AI anime short dramas of using their facial features or voices for commercial profit without permission.

The trouble lies in proof. Creators usually have only the finished work on hand; screenshots cannot prove how much intellectual labor they invested, nor can they explain which frame they designed and which frame the model generated at random. A report jointly released by the Chinese Academy of Press and Publication and the Development Research Center of the National Radio and Television Administration shows that in 2025, 70% of China's micro-dramas originate from online fiction. The boundary between original work and adaptation is already blurry, and once litigation begins, whoever produces a credible process record first gains the upper hand.

Registering the “Process”

Guizhou Tianyue Technology's approach is to install collection software on the computers used for creation, recording the entire creative process in data form. After registration, if an infringement dispute arises, the dataset can be retrieved directly as evidence. Company head Liu Qiao's description is that the rights-confirmation process is “very simple, one could say imperceptible.”

The key word here is “process.” Traditional rights confirmation looks at results, placing the finished work and the original side by side to compare similarity. Process evidence goes in a different direction: it records how you made the thing step by step, including prompts, parameter adjustments, iterative selection, and traces of manual modification. When finished-work similarity comparison gets stuck in the gray zone of “does it look alike,” a complete creative timeline can provide another kind of fact.

This path is not being opened for the first time. In 2023, the Beijing Internet Court heard China's first “AI text-to-image” case. The plaintiff used an AI tool to generate a character image, “Spring Breeze Brings Tenderness,” and the defendant used it without authorization and removed the watermark. The court found that the plaintiff, through prompt design, parameter adjustment, and iterative selection, had demonstrated original intellectual investment and enjoyed copyright; the defendant was ordered to pay compensation and apologize. The judgment likewise rested on the process, not merely the image.

Does Reskinning Count as Plagiarism?

Regarding the boundary of infringement, a relatively consistent judgment has formed in the industry. Wang Qian, a professor at the Law School of East China University of Political Science and Law, divided copyright protection into two layers in an interpretation: word-for-word plagiarism is infringement, and beyond that there is another layer—using word combinations and phrasing to express an original plot design. “Even if one rewrites another person's novel from beginning to end by replacing synonyms, with not a single sentence identical in word combination or phrasing, if the two express the same original plot design and use the substantial expression in the original literary work, it can still constitute infringement.”

For face-swapping and voice cloning, the criterion is “identifiability.” Experts point out that if an AI character in a short drama blends the facial features of many celebrities, or mixes many voices into a new voice, once it reaches an identifiable degree, it is suspected of infringement. This line does not require one-to-one copying; it only requires that the audience can identify who it is.

New Rules Push Responsibility onto Platforms

On September 1, the “Measures for the Administration of Micro-Drama Development” officially took effect. This is China's first specialized regulation governing the micro-drama industry. It brings AI-generated micro-dramas into standardized supervision for the first time and makes clear that AI technology cannot be used as a “shield” for infringement.

Half a month after implementation, real-world tests have already emerged. Reporters browsing various platforms found that most AI micro-dramas had been labeled “AI-generated content,” but some still carried no label at all. He Yuan, executive director of the Data Law Research Center at Shanghai Jiao Tong University, believes that the massive entry of AI into micro-drama generation involves the public's right to know, while AI hallucinations—such as shattered glass remaining intact—can easily mislead viewers; at the same time, the division of legal responsibility among content producers, platform broadcasters, and technology providers remains unclear, and watermarks will play a large role in that.

One of the most substantive changes in the regulation is that the role of broadcast platforms shifts from “safe harbor” to active gatekeeping. Platforms must strictly uphold bottom lines such as portrait and voice authorization and strengthen full-process management. This also explains a phenomenon: platforms are becoming increasingly cautious when purchasing AI dramas, locking in short-term exclusivity only for top innovative works, while most others receive only distribution authorization. The hit rate for AI dramas is around 0.5%; with short life cycles and strong replicability, permanent copyright has limited value. What is truly scarce is live-action content.

Rights Confirmation Has Changed Its Coordinate System

Connecting these developments, a direction becomes visible: in the AI era, rights confirmation is shifting from “looking at the finished product” to “looking at the process,” and from “whether there is a signature” to “whether you can prove it was made by you.”

For creators, this means that preserving creative records in advance is no longer optional but a step as important as writing the script. For platforms, labeling and authorization review will be written into daily workflows. Looking further ahead, if mechanisms such as data intellectual property registration can be rolled out in more places, the starting point for dispute resolution will move earlier, and many disputes that previously could only rely on screenshots and self-proof will be replaced by a retrievable dataset.

Technology lowers the threshold for production, so the rules must clarify the threshold for proof. Only when these two things happen together can the industry be considered stable.