Zhao Liying sues Hongguo for 810,000 yuan: the first AI short drama case moves from who made it to who let it go live

On the afternoon of 10 November, a hearing opened in the cloud courtroom of the Mentougou District People's Court in Beijing. The plaintiff is Zhao Liying. The defendants are three: Beijing Bimo Liuxiang Technology, which operates the short drama app Hongguo; its wholly owned shareholder Beijing Douyin Information Service; and one individual, Shi Moulun. The claim is a portrait rights dispute, seeking 810,000 yuan in economic loss and other damages.
Nobody in the room was arguing about the number. The dispute settles on a technical detail: was the face in the drama swapped in directly by AI, or were features from several faces blended into a new character? The first is easy to establish. The second has fuzzy edges. As of the hearing date, the point remains unresolved.
More interesting than the detail is the order of the defendants. The plaintiff put the platform first.
Why sue the platform first
The usual approach is to sue the producer. Whoever made the show, whoever used the face: liability is clear and the evidence chain is short. Enforcement is the problem. Most AI short drama producers are small companies with a few hundred thousand yuan in registered capital, and once a show comes down the company can be hard to find.
A platform is different. Hongguo's monthly active users rose from 158 million in December 2024 to 212 million in June 2025, which makes it an entity with money to enforce against. Several lawyers told reporters after the hearing much the same thing: naming the platform as first defendant is a change of tactic, because Hongguo reviews content before it goes live, which gives it a higher duty of care than an ordinary hosting provider and makes the safe harbour clause hard to invoke.
Safe harbour was designed for volume: the platform does not know and cannot know about every item, and taking content down on notice counts as doing its part. Once a platform sets its own revenue split, reviews its own scripts and decides its own recommendation slots, the premise of not knowing collapses. If you reviewed the script and signed off on the cover, then when a face that plainly belongs to a real person appears on that cover, claiming you did not see it is a hard sell.
The face in the ad slot is the most expensive asset in this chain. It decides the click rate, and whether a user swipes away at second three. Swapping in a famous face is the cheapest move available, and the easiest one to catch.

A March ruling already gave one answer
In March this year, the Beijing Internet Court reported a face swapping case involving an AI short drama and the actress Dilraba. The ruling held that both the producer and the streaming platform had to apologise and pay damages, and the first instance decision has taken effect. That ruling already wrote platform liability into the case law. Few people discussed it at the time.
By November, the same kind of dispute had moved from producer plus platform to platform in the first seat, and the difference is where the claim puts its weight. The earlier form made the platform jointly liable. This one makes it primarily liable. If the court accepts that ordering, every similar case after it will copy the structure.
On 11 November another related ruling came out: someone who posted an AI image mocking a person in a group chat was ordered to pay 50,000 yuan. Small case, clear direction.
The size of the industry guarantees more of these lawsuits
By public counts, about 128,000 short dramas launched in the first half of 2026, and AI short dramas made up roughly 95 percent of them. Different sources give different ranges for the same figures, so treat the percentage as an order of magnitude and not as a precise number.
The magnitude makes the point on its own. Thousands of new titles go live every day, each needs a few character images, and any of those images may trace back to a photograph of a real person. Checking every title by hand is not realistic. Expecting the platform to escape liability entirely is not realistic either.
Platforms are patching their processes too. Hongguo added a rights confirmation feature for AI drama characters in its creator centre, sorting virtual characters into three types: licensed from a real person, adapted from existing IP, and original design. Creators upload the character and proof of ownership, and once approved the record is filed as the basis for later distribution and for handling disputes. In July it also issued a notice requiring the main characters of a drama to be clearly distinguishable from one another, to cut down on the same AI face turning up again and again.
These moves show the platform knows where the risk sits. The catch is that rights confirmation is voluntary, while infringement is something nobody can declare on your behalf.
Once a face is in the model, the story gets murky
Even with a licence in hand, the story does not end. Engineering has a problem with no solution yet: once a model has seen a face, there is no way to make it forget. Deleting the original photos, closing the licensing intake, writing an expiry date into the contract, all of that stops new collection. None of it reaches the part already inside the weights.
So the question actually before the court has two layers. The first is whether the face was used. The second is how it got in, who supplied it, and how wide the permission ran. The second layer has no answer in this case yet, and it will set the damages scale for every case that follows.
For teams making AI short dramas, the practical impact of this hearing lands on procurement, not on statute. Buying assets used to be about price. Now the question is whether the licence document maps to a specific asset and a specific episode. A document that does not state its scope is the same as no contract at all.
For platforms, the impact lands on review. Review used to ask whether the content broke the rules. Now it also has to ask where the face came from. That costs money, and it is still cheaper than 810,000 yuan.
When the ruling arrives is unknown. What is certain is that from 10 November, an AI short drama lawsuit is no longer only about which company took whose face. It is about you putting that face on the platform, and how you plan to explain it.
Related articles
Chinese text in video always turns to mush. Dreamina handed the typesetting to code
Dreamina launched Motion Studio on November 10. It no longer generates frames; the model writes code that lays out text, shapes and pacing. Chinese titles stop blurring, but whether the content is right is still on you.
A community spreadsheet of taps that turns off the ads inside Chinese Android builds
A no-root tool that switches off the ads Chinese phone makers ship inside the system UI. Its method is a hand-written tap script per phone model, maintained by volunteers every time a ROM moves. It works, and the maintenance is the point.
GPT Image 2.6 held the token price and raised the bill
GPT Image 2.6 kept every token rate from 2.5 and still costs more per image, because more of the work now happens in output tokens. Eleven days into the ramp, the pricing structure says more than the leaderboard does.
Someone rebuilt IDM from the outside and wrote down what they did not touch
A Rust reimplementation of Internet Download Manager claims 5x the speed and 86 languages. The more interesting document is the paragraph explaining how it was built without copying anything.