º¬Ðß²ÝÉçÇø

Widespread adoption of generative artificial intelligence (AI) in the film, television and music industries in recent years has not only triggered a structural shift in content production but also posed a substantive challenge to the sector¡¯s underlying business logic and existing rights framework.

Yet rapid technological iteration is often accompanied by a lag in industry norms and legal regulation.

From the dispute sparked by OpenAI over Scarlett Johansson¡¯s voice to Hollywood strikes prompted by Sora, a series of incidents has exposed legal blind spots and practical pain points as old rules give way to new ones.

While handling recent commercial projects in areas including artist management as AI enters legal practice, the authors have found that market participants are facing increasingly complex infringement disputes.

The boundaries of the rights involved remain unclear, while traditional licensing contracts are ill-equipped to address the new issues raised by the technology.

For entertainment companies seeking financing or preparing for an IPO, intellectual property and data compliance flaws tied to AI-generated content have already become central risk points in regulatory and investment-bank due diligence.

Against this backdrop, this article ¨C the first in a series ¨C explores core legal issues at stake and moves beyond a single-focus infringement analysis to examine the challenges AI poses to commercial exploitation of personality identifiers, assessment of originality in AI-generated output, and compliance in model training.

Digitisation of personality identifiers and limits of commercial exploitation. Under traditional entertainment law, portrait rights and voice rights have generally been defined as personality rights closely bound up with the individual.

In the age of AIGC, however, an artist¡¯s voice, face and even bodily characteristics can be broken down by technical means into ¡°data assets¡± for machine learning, giving them substantial commercial value.

In judicial practice, Beijing Internet Court ¨C in what was described as the country¡¯s first case involving AI-generated voice infringement ¨C made clear that a person¡¯s voice, as a form of personality identifier, is personal and exclusive in nature; and that using AI to synthesise another person¡¯s voice without authorisation constitutes infringement.

Today, an artist¡¯s voice and likeness are no longer merely personality identifiers; they are core commercial assets. Where a third party uses deepfake technology without permission to create a highly recognisable ¡°digital clone¡± for commercial purposes, that conduct may not only infringe voice and portrait rights protected under the Civil Code, but may also amount to a ¡°confusing act¡± under the Anti-Unfair Competition Law.

In substance, it involves the improper appropriation of the artist¡¯s traffic and commercial opportunities.

How to commercially price, clear and license the ¡°AI training rights¡± for voices and likenesses, alongside the ¡°usage rights¡± for generated output, has become an urgent challenge for the industry.

Authorship and originality in AI-assisted creation. As AI tools become deeply embedded in film, television and music production, disputes over the legal characterisation of AIGC ¨C namely, whether it qualifies as work protected by copyright law, along with ambiguity surrounding ownership of rights ¨C have become a central trigger for multiple infringement disputes.

Current domestic judicial practice, as seen for example in the first copyright case over AI-generated text-to-image works, has established a preliminary standard centred on ¡°original expression¡±. Where a human, in the process of generating content with AI, has engaged in complex prompt design, parameter adjustment and aesthetic selection demonstrating substantial intellectual input, the resulting output may be capable of clearing the threshold for copyright protection.

By contrast, where a creator merely inputs simple instructions into an AI system, the output is unlikely to satisfy the originality requirement because of the absence of deep human intellectual contribution. Instead, it will be treated as material in the public domain.

That means AI-generated content produced with significant investment by film and television companies may nevertheless fail to attract exclusive rights, creating a material commercial risk.

In addition, the China Audio-Video Copyright Association issued new rules in January 2026, making clear that audio-visual programmes generated by AI will not be registered unless and until the law provides otherwise.

Whether in judicial practice or in the protection of works, this poses a challenge for identifying subsequent creative works.

Copyright risk and liability allocation at the model-training stage. Beyond copyright disputes surrounding generated output itself, another hidden danger for film, television and music producers lies in the legality of underlying training data used by AI tools.

Aftershocks of the Hollywood strikes underline deep concerns of creators that AI companies may scrape copyrighted material without authorisation for transformative training.

Copyright infringement suits brought by The New York Times, prominent authors and music publishers against leading AI companies underscore the point.

In the Ultraman case, Guangzhou Internet Court held that the defendant AI company infringed the claimant¡¯s rights of reproduction and adaptation in the Ultraman works, making it an early precedent on front-end liability for generative AI platforms.

If an AI tool used by an enterprise has been trained on content scraped without lawful authorisation, any script or score it generates may be found substantially similar to the original work.

For content producers, this represents a serious commercial hazard. Once a finished production containing such material is released to the public, the producer may not only face infringement suits and substantial damages claims from the original rights holder, but may even be forced to withdraw the project because defects in title undermine its core intellectual property rights, destroying its ability to generate commercial returns.

Takeaway

AI is reshaping the content production model of the film, television and music industries, but while doing so it is also generating complex and multi-layered rights conflicts and compliance challenges, while exposing limits of the current legal framework in practical application.

For entertainment companies, accurately identifying and clarifying these core legal issues is not only essential to prevent infringement disputes and avoid takedowns, but it is also a necessary precondition for ensuring the smooth progress of financing and listing plans.

On the basis of these infringement risks, the next article in this series will focus on specific compliance priorities and practical guidance for entertainment companies applying AIGC.

Jiang Shen is a partner and Xu Wen is an associate at Jingtian & Gongcheng