2025 is hailed as the "First Year of AI-Manju" across the industry. According to a report by iiMedia Research, the market size of China’s AI-Manju sector reached 18.98 billion yuan in 2025, surging 276.3% year-on-year. Data from Ocean Engine shows that in the first half of 2025, the monthly compound growth rates of AI-Manju’s playback volume and likes hit 92% and 105% respectively. Amid such a booming trend, everyone is eager to secure a share of the market. Yet the rapid advancement of technology is inevitably accompanied by the implementation of regulatory oversight.

On April 1, 2026, pursuant to the special rectification notice issued by the National Radio and Television Administration (NRTA), all existing animated short dramas that have not completed the filing procedures will be forcibly taken down across the entire internet. Moreover, filing is only the first step. The successive emergence of numerous legal risks has compelled practitioners across the board to re-examine corporate compliance.

I. Homogenization and Inferior Quality: Mistaking "Digital Waste" for Blockbusters

Driven by the pursuit of quick monetization, the market is flooded with a large number of low-quality and homogeneous AI-Manju productions. Many practitioners assume that content homogenization is merely a market issue and does not constitute illegality—a fatally flawed perception.

First, homogenization pushes the industry to take "shortcuts". When 90% of manju productions on the market feature fantasy, time-travel and rebirth themes, with identical visual styles and clichéd plots, how can works stand out? To compete for traffic, numerous producers opt for the "low-cost, high-stimulus" approach: distorting classic works, producing vulgar borderline content, and engaging in content plagiarism and copying. The end of these "shortcuts" is removal from platforms, account suspension, compensation claims, and even criminal liability.

Second, homogeneous content is more likely to be targeted by regulators. The special rectification campaign launched by the NRTA in November 2025 specifically targeted "homogeneous + non-compliant" content. The regulatory logic is straightforward: mass-produced low-quality content is often accompanied by large-scale violations. Works that look identical to 99 other productions and make the same mistakes are the easiest targets in regulatory crackdowns.

Third, homogenization deprives enterprises of a "compliance grace period". Works with originality and artistic value may be given opportunities for rectification by platforms and regulators even if they have minor flaws. However, for monotonous "digital waste", platforms have no reason to show leniency—removing such content causes no loss to users.

According to DataEye-ADX industry data, a total of 60,946 manju titles were launched in 2025, of which only 96 achieved over 100 million views, representing a blockbuster rate of merely 0.16%. This means that countless producers invested time, computing power and manpower, only to produce "homogeneous waste" that users would not even finish watching.

II. Unauthorized AI Face-Swapping & Portrait Rights Infringement: Mistaking a "Traffic Hack" for a "Court Summons"

Many practitioners hold misconceptions such as: "My content is AI-generated, not real people—no infringement", "I only used the profile, so no one can tell who it is", or "I did not use the actor’s name, so it is not infringement".

Pursuant to Article 1018 of the Civil Code of the People’s Republic of China, no one may infringe upon another’s portrait right by forging or using information technology means without permission. A portrait here does not need to be identical to the person; infringement is constituted if the general public or people in a specific industry can identify the individual depicted.

In March 2026, the Beijing Internet Court issued a landmark judgment. A renowned Chinese actor discovered that his portrait had been spliced onto characters in a 44-episode short drama (total duration 90 minutes) via AI face-swapping technology, leading the public to believe he had starred in the drama. The actor sued the producer and broadcaster respectively.

During the trial, the producer argued that the infringing image was AI-generated and that it had no intent to infringe. The court ordered the producer to re-demonstrate the face-swapping process, but the producer claimed it was unable to do so due to account and technical issues. Meanwhile, the plaintiff used the generated images and drama clips provided by the defendant to conduct face-swapping, producing an image completely different from that in the drama. The court ruled that the infringing clips were not accidental coincidences caused by AI face-swapping, but the result of the production company using the plaintiff’s portrait and deep synthesis technology.

Compensation for portrait right infringement of renowned actors is determined based on their commercial value, the scope of infringement, and the defendant’s profits, often amounting to hundreds of thousands or even millions of yuan. For A-list celebrities, the compensation sum could bankrupt a company outright.

III. Content Plagiarism & Rewriting: Mistaking "Luck Profits" for "Waiting to Be Slaughtered" by Copyright Holders

Many practitioners believe: "I rewrote online novels with AI—this is not plagiarism", "I only borrowed the plot, not copied the text", "Copyright of AI-generated content belongs to AI, not the original author", or "Even if I infringe, I will have made all the money before receiving a lawyer’s letter". All these are dangerous misconceptions.

First, legal accountability depends on "whether infringement occurred", not the speed of profit-making. AI generates content rapidly—producing a manju in minutes. But the logic of legal accountability is: once infringement is established, the right holder may sue at any time within the statute of limitations, regardless of how quickly profits were made. The faster and more profits earned, the larger the compensation base. Thinking of "making money and running away" actually accumulates a greater compensation liability.

Second, right holders’ rights protection capabilities are now "fully armed". In the past, right holders faced difficulties in detecting infringement, obtaining evidence, and long rights protection cycles—by the time a lawyer’s letter arrived, infringers had already made profits and switched accounts. Today, right holders have established professional AI monitoring teams, using anti-infringement tools for mass screening, with AI conducting automatic comparison, evidence collection and notarization. Evidence is fixed as soon as an infringing work is launched, and no speed can outrun automated evidence collection systems.

Third, platforms and regulators are establishing "blacklists" and "industry-shared databases". The notice issued by the NRTA in February 2025 explicitly requires the full implementation of a blacklist system for non-compliant short dramas and non-compliant short drama mini-programs, with resolute removal of non-compliant content. Thinking of restarting with a new account or company name is futile—ID cards and corporate information will be recorded, and marked entities will be barred from opening new accounts on any platform.

Fourth, mass litigation has become the norm, with compensation awards often reaching tens of millions. Right holders no longer sue one work at a time, but file mass lawsuits once infringing works accumulate to a certain quantity. In 2025, Guangdong Higher People’s Court and Chongqing Higher People’s Court issued final judgments in two copyright infringement cases filed by Tencent against Kuaishou involving De Yun Dou Xiao She and Lost You Forever, awarding compensation of 60 million yuan and 29.1 million yuanrespectively, totaling 89.1 million yuan. Right holders are willing to spend years and millions of yuan in legal fees to pursue infringement compensation, and no speed can outrun the entire judicial process.

In 2025, Ningbo Intermediate People’s Court of Zhejiang Province concluded a short drama copyright infringement case. The defendant unauthorizedly adapted the popular online novel Please Summon the Heroine on the Steed into a short drama. The court found infringement of the adaptation right, production right and right of communication through information networks, and ordered compensation of 250,000 yuanto the right holder. The court held that: "The key to determining infringement is whether the unique core of the original work is extracted and reproduced, not word-for-word comparison."

Meanwhile, producers often overlook criminal risks. The filing standard for the crime of copyright infringement is an illegal income of more than 50,000 yuan. Profits of hundreds of thousands of yuan from mass plagiarism far exceed this threshold. Once criminally filed, the defendant will face public prosecutors instead of civil compensation, with a maximum prison term of ten years.

In 2025, Zhu and Wang from Jiaxing, Zhejiang, developed an overseas short drama streaming APP, directly copying short drama content from overseas platforms, modifying posters and titles, and uploading them for paid viewing by overseas users. By the time of the case, the pirated platform hosted 65 titles totaling over 4,900 episodes, with illegal profits of more than 390,000 yuan. The court convicted the two of the crime of copyright infringement, sentencing them to fixed-term imprisonment of two years with a probation period of two years and six months, and fining each a total of 380,000 yuan.

The speed of AI has indeed generated quick profits in the short term, but the law does not waive accountability for "fast profit-making". Right holders are willing to spend years and millions in legal fees to protect their rights, and no speed can outrun the entire judicial process. More importantly, the plagiarism model traps enterprises in the bottom-tier role of "content porters". When regulation tightens, platforms conduct strict inspections, and right holders file mass lawsuits, such companies will be the first to be eliminated. In contrast, enterprises adhering to originality and compliant operations are accumulating IP, nurturing teams and building brands—their value continues to grow, while plagiarists’ value keeps resetting to zero.

IV. Core Changes in New Regulations: From "Platform Self-Review" to "Dual Oversight"

1. Regulatory Vacuum Under Old Rules

Prior to the launch of the special rectification campaign in November 2025, AI-Manju operated in a regulatory gray area. Unlike live-action short dramas, AI-Manju previously only needed to pass platform content reviews to launch, lacking unified industry standards and a regulatory framework.

2. Three-Tier Hierarchical Review System Under New Rules

Pursuant to the NRTA’s special rectification notice, AI-Manju is subject to a three-tier hierarchical review system based on investment amount and theme type:

· Key Short Dramas: Works with an investment of ≥ 3 million yuan, or those involving sensitive themes such as politics, military affairs and ethnicity, shall be reviewed by the NRTA and issued a License for Online Drama and Film Distribution. (Source: NRTA Special Rectification Notice, cited by Yan’an News Network on March 20, 2026)

· Ordinary Short Dramas: Works with an investment of 1–3 million yuan and ordinary themes shall be reviewed by the provincial-level radio and television authorities in their jurisdiction and issued a filing number.

· Other Short Dramas: Works with an investment of < 1 million yuan and general themes shall complete platform self-review, then be uniformly filed with and issued a filing number by the provincial-level radio and television authorities.

All short dramas must prominently display the license or filing number at the beginning of the video, and unapproved/unfiled works are prohibited from online dissemination.

V. Compliance Recommendations: Compliance Is an Investment, Not a Cost—What You Need to Do

1. Qualification Compliance: Secure the "Entry Ticket"

(1) Apply for the License for Radio and Television Program Production and Operation, ensuring complete materials and smooth procedures to avoid rejection and delays due to minor issues.

(2) Sort out the specific requirements of local radio and television authorities for reviewers and make precise matches.

(3) Design an optimal qualification-holding structure based on business models (self-operation, agency operation, MCN institutions, etc.) to reduce risks.

2. Content Compliance: Steer Clear of "Minefields"

(1) Conduct pre-compliance reviews of scripts and finished works, identifying plots that may cross new regulatory red lines and "borderline content" that actually violates rules.

(2) Establish internal review system templates, including review processes, review records and accountability mechanisms, to standardize corporate operations.

(3) Provide technical implementation plans and legal compliance statement templates for the labeling requirements of AI-generated content.

3. Copyright Compliance: Safeguard the "Lifeblood"

(1) Design a full-process evidence retention plan for AI creation: prompts, parameter settings, iterative revision records, material authorization contracts, etc., to ensure clear and traceable ownership of every work. This plan is critical in AI face-swapping cases—the producer’s failure to reproduce the creation process led the court to find intentional infringement. It also effectively prevents internal staff from cutting corners in violation of rules and exposing the company to risks.

(2) Complete copyright registration or retain copyright evidence (including DCI digital copyright registration and blockchain evidence storage) to provide strong ownership proof in case of infringement.

(3) Review user agreements and copyright clauses of AI tools, selecting tools with clear training data sources and explicit copyright ownership clauses to avoid the trap of "tool provider exemption, producer liability".

4. Contractual Risk Isolation: Cut Off "Joint Liability"

(1) Draft/review cooperation agreements with screenwriters, producers, platforms and AI tool providers, clarifying copyright ownership, infringement liability division and recovery mechanisms.

(2) Include infringement compensation and recovery clauses: if the company is sued due to infringing materials provided by a cooperative party, it shall have the right to recover full compensation from the cooperative party.

(3) Establish an internal corporate intellectual property management system, including employee confidentiality agreements, legal person work confirmation documents and ownership agreements for service works, to prevent core employees from taking IP away when leaving.

5. Dispute Resolution: Extinguish "Fires" Timely

(1) Immediately respond upon receiving a lawyer’s letter or being sued, assess infringement risks, formulate litigation strategies, and strive for out-of-court settlements or reduced compensation. In the Tencent v. Kuaishou case, Kuaishou was ordered to pay nearly 90 million yuan in compensation due to algorithm-driven dissemination—early intervention can avoid such sky-high awards.

(2) Conduct copyright license negotiations to obtain authorization at a reasonable cost and avoid litigation.

(3) In case of administrative or criminal reviews, promptly submit and supplement legal opinions and defense materials during investigations to clarify facts. For confirmed infringement, strive for mitigated or exempted penalties.

Producing 100 manju titles a year with a profit of 10,000 yuan per title yields a total profit of 1 million yuan. A single infringement lawsuit can result in compensation plus legal fees of hundreds of thousands or even millions of yuan, wiping out half or all profits. Concurrent account suspension or criminal liability can cut off future revenue and drive the company to bankruptcy.

On April 1, 2026, AI-Manju officially entered the era of "certified operation". For practitioners, this is both a challenge and an opportunity. In the short term, compliance thresholds will eliminate a batch of low-end creators lacking qualifications and content capabilities, and the industry will undergo a painful adjustment period. In the long run, the implementation of regulation will drive the industry to shift from "traffic-driven" to "quality-driven", freeing up market space for practitioners with genuine content creation capabilities.