Who Owns Your Face, Voice, and Identity? AI Performer Libraries and Personality Rights in China

24 Jun 2026

By Hansen Tao

The recent controversy surrounding the launch of an “AI performer library” by iQIYI has drawn widespread attention to a legal issue that is rapidly moving from theory to practice.

What’s an “AI performer library”?

For those not familiar with this technology, an “AI performer library” is essentially a digital database of human performers whose identities can be recreated or simulated using artificial intelligence.

In practical terms, platforms like iQIYI use this concept to describe a system where an actor’s face, voice, expressions, and performance style are captured as data and made available for AI-driven content production.

An AI performer library typically includes:

Visual data: facial features, body movements, gestures
Audio data: voice recordings, tone, speaking patterns
Behavioral patterns: acting style, expressions, timing

These elements are used to train AI models that can generate a digital replica of the performer.

Once a performer is “in” the library, AI tools can:

Generate scenes featuring their likeness without physical filming
Recreate their voice for dialogue
Insert them into new or existing content
Scale their “appearances” across multiple productions simultaneously

This is why companies sometimes claim such systems can multiply an actor’s output without requiring their physical presence.

The case

Shortly after the iQIYI platform announced that over one hundred actors had joined its AI initiative, several well known performers, including Zhang Ruoyun and Yu Hewei, publicly denied granting any such authorization.

What initially appeared to be a dispute over commercial communication has evolved into a broader legal question with global relevance.

As artificial intelligence technologies increasingly enable the replication of faces, voices and performance styles, the fundamental issue is no longer whether such uses are technically possible, but whether they are legally permissible.

From technological capability to legal accountability

AI generated content has reached a level where digital replicas can convincingly reproduce not only a person’s appearance but also their voice, mannerisms and expressive style. In this environment, the concept of identity extends beyond static images and enters the realm of dynamic, data driven representations.

Chinese courts have already begun to adapt. In recent cases involving AI face swapping, judicial reasoning has confirmed that personality rights are not limited to facial likeness alone.

A person may be identifiable through a combination of physical traits, movement, voice and contextual elements. This broader understanding significantly expands the scope of protection and reflects the reality of how AI systems reconstruct human identity.

The dispute involving iQIYI illustrates how easily legal boundaries can become blurred. Even if a platform characterizes participation in an AI database as an expression of interest rather than a formal license, the act of listing a performer together with identifiable attributes may itself constitute use.

Under Chinese law, use is not defined by commercial intent but by whether a protected attribute is reproduced, disclosed or made available to the public.

The legal framework in China

China’s regulatory approach to personality rights provides a relatively clear structure for analyzing such disputes. The Civil Code of the People's Republic of China establishes that individuals have the exclusive right to control the creation, use and disclosure of their likeness.

Importantly, this protection does not depend on whether the use is for profit. Unauthorized use can constitute infringement regardless of the underlying business model.

The same legal framework extends protection to voice. This is particularly significant in the AI context, where voice cloning has become both technically accessible and commercially valuable. By explicitly placing voice within the scope of personality rights, Chinese law closes a gap that exists in some other jurisdictions.

A further layer of regulation is introduced by the Personal Information Protection Law, which classifies biometric data such as facial features and voiceprints as sensitive personal information.

The processing of such data requires separate and specific consent. General or bundled consent mechanisms are unlikely to meet this standard, especially when the intended use involves training AI models or generating synthetic content.

Taken together, these rules suggest that any large scale use of human likeness in AI systems must be built on a foundation of clear, informed and individualized authorization. Ambiguous expressions of intent or implied consent are insufficient.

The limits of “intent” in a legal sense

A central issue raised by the controversy is whether indicating a willingness to explore AI collaborations can justify the inclusion of a person in a publicly accessible database.

From a commercial perspective, platforms may view such listings as preliminary or promotional. From a legal perspective, however, the distinction is less forgiving.

The inclusion of a person’s name, image or other identifiable attributes in a structured database designed for commercial exploitation is likely to be regarded as a form of use. Once that threshold is crossed, the requirement for consent is triggered. The characterization of the listing as non binding or exploratory does not negate this requirement.

Chinese judicial practice increasingly emphasizes substance over form in this area. Courts focus on the effect of the conduct rather than the terminology used to describe it.

If the public can associate an AI generated or AI enabled output with a specific individual, the legal analysis will center on whether that individual has authorized such association.

Emerging risks in the AI ecosystem

The case also highlights structural risks that extend beyond a single platform. One of the most pressing concerns is the lack of standardized contractual provisions addressing AI use.

Many talent agreements still do not clearly define whether and how a performer’s likeness may be used in training datasets or synthetic productions. This creates uncertainty not only for performers but also for studios, agencies and technology providers.

Another challenge lies in the persistence of data once it enters an AI system. Unlike traditional uses of images or recordings, data used for model training may be difficult or impossible to fully remove.

This raises complex questions about withdrawal of consent and the long term control individuals can exercise over their digital representations.

Platform liability is also evolving. Regulatory developments in China indicate a shift toward imposing more explicit obligations on AI service providers, including duties related to data governance, ethical review and risk management.

The idea that platforms can rely solely on technological neutrality as a defense is becoming increasingly untenable.

Implications for international practice

Although the controversy arose in China, its implications are global. Jurisdictions around the world are grappling with similar questions, often with less developed legal frameworks.

China’s approach, which integrates personality rights with data protection principles, offers a relatively comprehensive model for addressing AI generated identity use.

International businesses operating in or with China must understand that compliance cannot be treated as an afterthought.

The use of AI involving human likeness requires careful alignment with both contract law and regulatory requirements. This includes obtaining explicit consent, defining the scope of use with precision and maintaining transparent records of authorization.

A shifting balance between innovation and control

The rapid development of AI technologies is transforming the entertainment industry and beyond. At the same time, it is testing the resilience of legal concepts that were originally designed for a different technological era. The controversy surrounding AI performer libraries underscores a fundamental principle that remains unchanged.

Technological capability does not override personal autonomy. The ability to replicate identity at scale makes consent more important, not less. As courts and regulators continue to refine the rules, the central question will remain how to balance innovation with the protection of individual rights in a way that is both effective and enforceable.

In this evolving landscape, the boundary is not defined by what AI can do, but by what the law allows it to do.

Allasya comments

Technology can replicate likeness, but the law protects personal autonomy.

The core issue in this case is not whether the platform had "intentions," but rather that the commercial use of personality elements without explicit authorization constitutes a legal "use."

Enterprises must remember: consent cannot be implied; authorization must be clear. In the AI era, respect for individual will remains an inviolable legal boundary.