Summary of Key Points
The AI short drama industry has experienced explosive growth, with a market size reaching 22 billion yuan in the first half of 2026. However, the problem of "face theft" has become increasingly severe: the faces of internet celebrities, ordinary people, and even stars have been unauthorizedly used in AI short drama characters. Victims face not only financial losses (such as interrupted business deals) and emotional harm (online bullying) but also technical challenges that are virtually irreversible (their faces are stored in AI models and cannot be completely removed). The process of seeking justice is fraught with difficulties, including the challenges of providing evidence and determining liability. Moreover, some producers within the industry exploit loopholes in the rules, resulting in extremely low costs for infringement. At the same time, some stars choose to actively license their images for revenue, which contrasts sharply with the struggles of ordinary people in protecting their rights.
Detailed Analysis
The Triple Blow to Victims
AI face theft is no trivial matter; it directly threatens the livelihoods of the victims. For example, model Qihai did not receive any business deals for three months after filing a lawsuit, and her MCN agency asked her to remove the related videos because brands were concerned about the controversy. Self-media blogger EnSerser faced even more severe consequences: she was bullied online, businesses refused to cooperate with her, she had to wear a mask in public, and her fans would chase her to confirm her identity. Not only did they lose their income, but they also endured immense psychological pressure—Qihai received messages saying she was "going crazy with the desire for fame," while EnSerser was cursed every two minutes. The cost of seeking justice was also high, with both spending nearly 40,000 and 30,000 yuan respectively, and the legal process could take more than a year.
Technical Consequences: Faces "Stored" by AI, Impossible to Remove
The most disheartening aspect is that once a face is captured by an AI model, it seems to be imprinted permanently. For instance, although the original infringing short drama "Peach Blossom Hairpin" was taken down, 80% of the similar faces could still be seen in new dramas. When Qihai uploaded videos of her own face, the platform identified them as AI-generated because her face had likely been included in the model training data. An AI company expert explained that the models "memorize" facial features, and even if the original content is removed, using keywords (such as "model with a mole on the face") can trigger the generation of similar faces. To completely eliminate such images would require cleaning the underlying data, which is almost impossible.
Two Major Barriers to Seeking Justice
- Proving Ownership: In traditional copyright disputes, it is relatively easy to determine if a face has been copied, but AI-generated faces present different challenges:
- Difficulty in Recognition: An AI-generated face may be 70% similar to the original, but not identical, and defendants might argue that it's just a coincidence. For example, in a case heard by the Beijing Internet Court, the judge had to consider user feedback (e.g., "Isn't this someone we know?") to determine infringement.
- Burden of Proof: Generally, the person claiming infringement must provide evidence, but in AI face theft cases, the victims often lack access to the algorithms and model parameters needed to prove their case. As a result, the court may require the defendant to demonstrate the generation process; if they cannot do so, they are deemed liable.
The "Unwritten Rules" of the AI Short Drama Industry
The AI short drama market has grown rapidly, with expectations for a scale of over 40 billion yuan in 2026. However, there are almost no industry regulations to govern it:
- Low Cost of Infringement: Some producers simply use other people's photos in AI-generated content, betting that the victims will not pursue legal action due to the time and cost involved.
- Ways to Avoid Detection: To avoid detection, they may modify a face to make it only 70% similar or combine features from multiple individuals (e.g., using someone's eyes from one person and mouth from another), making it impossible to identify the source of the image.
- Lax Platform Review: The Hongguo platform removed "Peach Blossom Hairpin" because the producer could not provide compliant evidence, but many other infringing content may go unnoticed.
The Difference in Responses between Stars and Ordinary People
Stars and ordinary people have vastly different options when facing AI face theft:
- Star Authorization: Celebrities like Qi Wei and Wu Qihua actively license their images, earning revenue while maintaining control over their image (for example, through AI-based comics).
- Ordinary People's Struggle: Individuals like Qihai and EnSerser have to fight the battles on their own, as they lack the resources to negotiate licenses or confront platforms and producers.
This disparity reflects the unequal distribution of resources: stars have teams and influence, allowing them to turn face theft into a source of income, while ordinary people are forced to take proactive legal action and may be viewed as troublemakers.
In Conclusion
AI technology was meant to be a tool for creativity, but it has become an aid to infringement. If industry regulations do not keep up, anyone could become a victim of face theft tomorrow. Qihai's statement that "the cost of AI face theft should not be so low" reflects not only her concern but also the hope of all ordinary people.