An AI short drama called Menghui Honglou (Dreaming Back to the Red Chamber) stands accused of fusing the faces of 17 actors from the 1987 television classic Dream of the Red Chamber into an altered plot without consent, according to two Tencent News entertainment reports. [1][2] The topic reached China's hot-search lists on November 12. [1] The named cast runs from Zhang Li and Deng Jie to Gao Hongliang and Li Ting, and includes Chen Xiaoxu, the Lin Daiyu actress who died in 2007; screenshots users shared show an episode titled "The Grand Princess Rages; Baochai Slapped in the Street; Will the Xue Family Still Dare to Be Arrogant". [1] Under the identifiability standard Chinese courts apply to synthetic likenesses, a synthetic face the public can still recognize may still be deemed portrait use, AI-generated or not; no ruling has applied that standard to this drama yet. [4]

The timing is what matters for tool users. The account behind the drama had posted 368 Red Chamber works, 21 of them AI episodes; its top AI episode drew 115,000 likes against the 100-to-200 range its clip posts usually received, and it works with MCN Qichuang Media. [1] Between November 1 and November 10, the regulator opened a two-month governance window that names "face theft" and "voice theft" in AI micro-dramas, [8] the short-drama app Hongguo announced a voluntary rights-record capability for AI virtual characters in its creator center on November 9, [7] and a portrait claim against Hongguo's operator and its parent was heard at Beijing Mentougou District People's Court on November 10. [3] For anyone assembling AI short drama with human likenesses, the next step is documentation. Have consent evidence for every face before submitting work, and file it through the voluntary rights-record channel where the platform offers one.

The identifiability test decides portrait cases

The answer starts in the Civil Code. Article 1019 bans harming portrait rights through forgery using information technology, and bars making, using or publishing a person's likeness without consent, unless the law provides otherwise. [6] The Supreme People's Court published an opinion on adjudicating AI disputes at a September 7 press conference. [10] As TMTPost quotes the opinion, even where a synthetic image differs from the person, it can still be deemed to use that person's portrait when the general public or a specific group can identify them. [4]

Beijing Internet Court rulings run the same way. In a March case Xinhua documented, the producer that used deep synthesis to generate a highly similar likeness had to apologize in writing and pay damages, and the streaming platform that skipped a reasonable review shared liability. [5] TMTPost's account of another ruling adds that identifiability does not require an exact match, that a producer unable to reproduce its AI-generation records bears the consequences, and that a platform holding a valid license for the work still shares liability when it skips review. [4]

Lawyers quoted in the reports put the test this way. If an AI-built figure can be recognized as the corresponding person and is used commercially without permission, it is infringement, and an AI virtual-persona label is no defense. [1][2] Portrait interests survive the person, and close relatives can pursue the claim, the point one report makes for Chen Xiaoxu. [1] The pattern predates this case. Earlier in 2026, AI face swaps hit Yang Zi in March and Yi Yangqianxi in April, and Dilraba's suit against a face-swap company ended in a judgment ordering a public apology and damages. [1]

None of this has been tested against Menghui Honglou in a ruling. One report, published on November 13, says suits have been filed by Zhang Li, Ouyang Fenqiang and Deng Jie, with a takedown across platforms among the demands. [1] An earlier report from a second outlet, published the afternoon of November 12, says the production's identity and reach are still being verified. [2] Neither account carries a court filing.

The dated moves on distribution, September 1 to November 10

The legal test was settled enough for the distribution side to act before any ruling. Order No. 16 of the National Radio and Television Administration, the Measures for the Development of Micro Short Dramas, took effect September 1. Micro-dramas are serialized episodes under 20 minutes; they sort into three investment classes with filing and licensing duties; the third class needs the distributing platform's pre-broadcast review and a program number; works without a license or a number cannot be broadcast. [9] Article 46 of the order allows fines up to ¥100,000, about US$14,000, for producing other than the filed content, broadcasting other than the licensed content, or transferring a work to another reviewer (dollar figures use ¥1 ≈ US$0.14, approximate). [9] On November 1 the regulator opened a two-month special governance that runs to the end of December. Its content-side tasks name "face theft" and "voice theft" in AI micro-dramas, and an NRTA official's year-to-date tally is 75,000 violating micro-dramas taken down and more than 1,500 accounts handled, counts that cover all violating micro-dramas nationwide rather than AI output alone. [8]

Then the platform itself moved. Hongguo's creator center announced on November 9 an AI virtual-character creation-and-rights-record capability; creators holding real-person consent, upstream IP licenses or copyright-registration certificates can submit materials for review, and approved records serve the content's use and rights handling as reference, with the announcement framing the capability as optional support. [7] Scale is why the platform checks matter. QuestMobile's September figures put Hongguo at 430 million deduplicated users, up 82% year over year, past ByteDance's Doubao assistant and second inside ByteDance only to Douyin. [4]

The operator chain reached a courtroom on November 10. Actress Zhao Liying's claim at Beijing Mentougou District People's Court names Beijing Bimo Liuxiang Technology, the Hongguo operator founded in July 2021 with ¥1 million in registered capital and wholly owned by Beijing Douyin Information Service, together with the parent company and an individual surnamed Shi. [3] The claim asks the court to stop production and sale of the works, secure a public apology, and pay ¥810,000 for economic losses and rights-protection costs over alleged face-swap and face-merge uses of her likeness and photo covers used for traffic. [3] The case is undecided. [4] Hongguo had already run a material-governance round in April that took down 3,522 low-quality AI-comic dramas. [4]

What a face costs, and what it can cost you

The review capacity answers to money as much as to regulator pressure. Xinhua's May investigation into the face trade quoted gray-market rates of ¥100 to ¥200 per person per year, about US$14 to US$28, with some producers paying per use, from several hundred yuan (about US$42 if anchored at ¥300) to over ¥1,000, about US$140. [5] Demand for faces climbed while the market flooded. DataEye counted AI simulated-human dramas above 60% of the top-100 AI-comic chart in March, from about 7% at the start of last year, and more than 100,000 AI drama and AI-comic titles launched in the first quarter, 55% of them simulated-human. [5] A national industry report puts 490,000 micro-dramas online in the first three quarters of 2026 (TMTPost's arithmetic works that out to roughly 1,800 new titles a day), in a market of 850 million users with monthly actives above 600 million. [4] Some 221,900 new AI dramas arrived in the first half, and 1,055 of them, about 0.47%, passed 100 million plays. [4] In a market where almost nothing breaks out, a recognizable face is the cheapest shortcut to recognition, and courts reviewing the cases describe the pipeline: batch-crawled public photos and video stills, extracted facial features, fine adjustments that weaken identifiability. [5]

The exposure numbers are bigger. The Mentougou claim asks ¥810,000, about US$113,000, in a single actress's case, an unresolved claim rather than an award. [3] Xinhua documented one platform's 72-hour compliance window closing with the producer unable to show source-compliance evidence, and the work came down across the board. [5] The asymmetry shows in arithmetic, offered here as a labeled hypothetical only. Licensing 17 faces at the top of the quoted annual range would cost ¥3,400, about US$476, a year, against the ¥810,000 in one unresolved claim. [5][3] The comparison combines Xinhua's published gray-market quotes with the claim's stated amount; it is not an account of Menghui Honglou's costs, which remain unverified.

What the rights record covers, and what it leaves open

The rights record answers the filing question, within limits its announcement states. October's governance notice had already listed "many people, one face" and "cross-drama face collisions" among problems created by batch-applying the same face models. [7] The record it issues aids the retention of rights information and does not represent an absolute guarantee against content or rights risks; published content still has to comply with platform rules and the law. [7] Third-class distribution still runs through pre-broadcast review and a program number regardless of any filing. [9]

The unsettled part matters for anyone reading the November headlines. The two reports, published November 12 and November 13, do not record a takedown of Menghui Honglou. [1][2] The 17-actor suits rest on one outlet's account, a second puts the scope at "at least 17" actors across the 1987 classic and other works with the production still under verification, and the Zhao claim is undecided. [1][2][4] Whichever way the lawsuits end, the changes around distribution hold. The regulator's two-month window runs through the end of December, [8] Order No. 16's review and program numbering decide what can be broadcast, [9] the rights-record channel gives Hongguo creators a voluntary way to document face sources, [7] and the one operator-level claim now documented, Zhao Liying's, was heard on November 10. [3] The burden has moved upstream to the people assembling the videos. The face source is the input a platform can ask to see, and the rights record is the filing that says where it came from.