China’s AI Drama Boom Spurs Trade in Digital Likeness, Raising Legal Risks

iQIYI’s proposed digital actor database highlights a fast-growing market for facial data as platforms industrialize AI content production

When iQIYI Inc. proposed building a database of digital actors last month, Chief Executive Gong Yu suggested traditional filming could one day become “intangible cultural heritage.”

The remark quickly drew backlash, prompting the company to clarify that actors listed had only expressed interest in AI-related projects—not granted blanket authorization.

But the episode exposed a broader shift already underway. China’s AI-generated short drama industry is rapidly building a market for human likeness, turning faces into low-cost, tradable digital assets.

Multiple production companies said they routinely purchase image rights from models, aspiring actors and ordinary individuals for as little as 200 yuan ($28) per project. The practice is reshaping how content is produced and priced, replacing high upfront talent costs with scalable digital licensing.

Platforms are formalizing the process. ByteDance’s Jimeng, for example, has introduced authorization systems that convert facial likeness into tradable assets, supported by QR-code verification and traceable records.

Regulatory Pressure Consolidates the Market

Authorities are moving to curb unauthorized use of digital likeness.

On April 2, the China Radio and Television Social Organizations Federation’s Actor Committee issued guidance calling for stricter identity verification. Short-video platform Hongguo said it removed 1,718 dramas in the first quarter of 2026, after reviewing about 15,000 titles and sanctioning hundreds for violations.

The crackdown has not eliminated the practice but is instead consolidating it among compliant players.

Libre, founder of an AI comic-drama production company who asked to be identified by a pseudonym, said her firm now relies entirely on licensed images. “After compliance requirements tightened, we shifted fully to purchased authorizations,” she said, describing a process that typically requires only a handful of photos and identity verification.

Social media platforms including Xiaohongshu now host open calls for facial data. One listing reviewed offered three-year contracts at about 200 yuan per drama, requiring only front, profile and ID photos—no physical audition.

Technology Scales Production Faster Than Rules Evolve

AI-generated dramas typically follow two production models, according to industry participants.

One relies on text prompts trained on vast image datasets, often producing faces resembling real actors. The other uses uploaded photos directly, improving efficiency but increasing legal risk.

Production timelines have shrunk sharply. Teams of three to five people can now complete 90- to 120-minute dramas within weeks. The speed is compounding compliance risks, as creators often reuse or modify existing facial features rather than design original characters.

Recent court rulings suggest regulators are taking a stricter stance. In March, the Beijing Internet Court found that an AI face-swap drama involving actress Dilraba Dilmurat constituted infringement based on recognizability standards. Within weeks, multiple actors issued legal notices over similar uses.

Non-celebrities are increasingly affected as well. A hanfu content creator discovered her image had been used to depict a villain in an AI drama, prompting its removal after complaints.

Legal experts expect disputes to rise. “Cases will continue increasing until clearer rules are established,” said Sun Qimin, a partner at Beijing Xingye Law Firm.

Legal Gaps and Low Enforcement Incentives

Legal recourse remains slow and costly.

While platforms can remove or alter content quickly after complaints, court proceedings often take six to 18 months. By then, public attention has typically faded.

Ordinary individuals face additional hurdles. Unlike celebrities, they often struggle to prove recognizability when images are modified, and lack clear benchmarks for damages. Compensation is usually limited to proven losses.

A fragmented legal framework adds complexity. A single image may trigger claims under portrait rights, data protection, unfair competition and defamation law, yet no unified statute governs AI-generated likeness.

Technical workarounds persist. Some creators convert real faces into stylized images before reprocessing them into new human-like outputs, bypassing platform restrictions on direct uploads. Training data presents another challenge, as companies often scrape publicly available film and television content while keeping algorithms opaque.

Pricing Reflects Early Market Stratification

The emerging market shows clear price tiers.

Ordinary individuals typically receive 100 to 500 yuan per authorization, while professional models command several thousand yuan. Celebrity pricing remains largely undisclosed.

Even so, some industry participants expect downward pressure on prices.

“If the cost is too high, we simply substitute,” said one director. Minor adjustments to facial features can reduce legal exposure while avoiding payment obligations.

Others see a different trajectory. Some producers are shifting toward fully synthetic virtual characters, which offer greater control and fewer legal risks.

“What’s being replaced is not people, but highly standardized content production,” Libre said. “Distinctive creators will still stand out.”

Platform Power Outpaces Accountability

Platform incentives may ultimately shape how the market evolves.

Historically, internet platforms operated as neutral intermediaries with limited liability. But AI-driven content systems now generate, distribute and promote material, blurring the line between platform and creator.

“Responsibility and control are increasingly misaligned,” Sun said, noting that platforms face limited pressure to go beyond minimum compliance requirements.

The confusion extends to consent itself. Several actors publicly denied authorizing inclusion in iQIYI’s proposed database, highlighting uncertainty even among industry professionals.

Meanwhile, intermediaries continue sourcing facial data from a widening pool, including students and stay-at-home parents—groups often less sensitive to long-term rights implications.

An Expanding Market With Unclear Boundaries

As facial data becomes more liquid, fundamental questions remain unresolved.

Images licensed for one-time use may become embedded in training datasets. Low-cost contracts can grant broad or unclear rights over future usage. And the transformation of faces into digital commodities is outpacing legal definitions.

For now, the economics are clear: scalable production, low input costs and strong platform demand.

The legal framework is not.

As China’s AI content industry scales, the commercialization of human likeness is accelerating faster than regulation can keep pace.

The question is no longer whether faces will be traded, but under what terms—and at whose expense.