A Research on Copyright Infringement and Fair Use of AI DubBing

Xinru Shao

Politics and Law School, Yichun University, Yichun, Jiangxi, 331100, China

Keywords: Fair Use, AI Dubbing, Balanced Spirit, Technology Neutrality.

Abstract: The rapid development of generative AI technology in recent years and the arrival of AI Dubbing has led to

huge changes in the music industry, but the relevant laws are still being explored. In the face of the conflict

between new technologies and traditional copyright holders' rights, people should find a balance between the

two sides in our exploration and protect copy-right rights while promoting technological development. The

article examines the fair use regime to explore the possibility of applying fair use to AI Dubbing. The article

concluded that although AI Dubbing technology cannot be simply applied into the traditional twelve special

circumstances of fair use, in terms of the Principle of Technological Neutrality, it is in line with the legislative

purpose of the Copyright Law of the People's Republic of China to pro-mote technological progress, and is

therefore able to be included in the fair use system.

1 INTRODUCTION

Since April 2023, Artificial Intelligence (AI)

Dubbing has gained a lot of attention on social media

platforms.AI Dubbing has triggered a lot of public

discussion and attention since it entered the country.

AI Sun Yanzi serves as a pertinent case study. It

covers songs that have been uploaded to the NetEase

CloudMusic platform, Bilibili, and other digital

platforms. Following the attainment of over eight

million plays on these platforms, and more than two

million on the latter, an independent compilation of

cover songs was released, accompanied by the

establishment of fan groups. Concurrently, a foreign

song, entitled "Heart on My Sleeve", experienced a

meteoric rise on the online social media platform,

TikTok, amassing a staggering 15 million views from

its release until its removal. The song's record label,

Universal Music Group, has sent a letter to streaming

music platforms such as Spotify and Apple Music,

citing copyright infringement, demanding that AI's

access to Universal Music Group's internal music be

halted, with the aim of preventing the operators of AI

Sound Synthesis from using all of its copyrighted

music for AI model training and treating such

behaviour as copyright infringement (Chen, 2023). In

2023, the Beijing Internet Court heard the first “AI

voice infringement case” in China. In the case, the

plaintiff, as a dubbing artist, discovered that the

defendant had, without lawful authorization,

processed the plaintiff's voice with AI, and named the

voice ‘Magic Xiao Xuan’ on the ‘Magic Voice

Workshop’ app for sale to the public, and the plaintiff

sued the defendant for infringement of her own voice

rights (Zeng and Chen, 2024).

Infringement issues arising from AI covers are no

longer limited to the music industry; They have

permeated all aspects of daily life. Nevertheless, as an

emerging technology, AI covers are still in a legal

blind spot.2023 China Audio-Video Copyright

Association and the WIPO Office in China have

joined forces to organise a seminar on Artificial

Intelligence and Copyright Issues in the Music

Industry. In the case that the relevant laws are not yet

perfect, it becomes more necessary to study the issue

of AI covers for copyright infringement and fair use.

This paper is about AI Dubbing infringement and fair

use of AI Dubbing. Since AI Dubbing is primarily a

training model for generation, it relies on inputs from

prior works, which poses a greater risk of copyright

infringement. To know what rights are infringed by

AI Dubbing and what public benefits the technology

can bring, the nature of the behaviour has to be

clarified by understanding the technological path of

AI Dubbing.

2 THE TECHNOLOGY PATH TO

AI DubBing

The technical principles of AI Dubbing are based on

134

Shao, X.

A Research on Copyright Infringement and Fair Use of AI DubBing.

DOI: 10.5220/0013969100004912

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Innovative Education and Social Development (IESD 2025), pages 134-138

ISBN: 978-989-758-779-5

deep learning and artificial neural network

techniques such as Recurrent Neural Networks

(RNN) or Convolutional Neural Networks (CNN).

Using software such as SO-VITS-SVC, a large

amount of the target singer's audio is fed into it and

converted to a digital signal. The data completed

through the model transformation is processed and

generated to form an acoustic model that is highly

similar to the target singer, and the AI Dubbing can

be completed as long as the new input song data is

processed using the trained acoustic model. It follows

that an AI Dubbing is essentially a producer using an

Artificial Intelligence Model to train an AI Vocal

Model that can be infinitely closer to the target singer.

Artificial Intelligence training use of prior works has

a low density, guided by the purpose of generating

and training models, AI Dubbing is bound to come

into contact with numerous protected prior works,

which will cause infringement of the legitimate rights

and interests of copyright owners of prior works.

3 LEGAL RISKS OF AI DubBing

According to Articles 39 and 44 of the Copyright Law

of the People's Republic of China, the risk of AI

Dubbing infringing on a singer-performer's property

rights can be categorised into two situations. A more

common situation is where AI Dubbing commercially

exploits audio and video of performances that have

been legally recorded and distributed by other

subjects, but the exploitation by AI Dubbing is

unauthorised. The other is AI Dubbing commercially

exploiting audio and video of performances that are

live-streamed, transmitted, recorded, or reproduced

by themselves or by other subjects without the

performer's permission (Zeng and Chen, 2024).

At the 8th Music Industry High-end Forum, the

relevant person in charge of Universal Music Group,

which owns the copyright of Sun Yanzi's songs,

stated his stance on the issue of ‘AI Sun Yanzi’: ‘If

AI companies want to use our works, they have to get

permission first. If AI companies want to use our

works, they have to get permission first and use them

legally.’ The executive also said that Universal Music

Group has asked streaming platforms to ‘remove

songs generated by AI (Shou and Man, 2023).’

Article 42 of the Copyright Law of the People's

Republic of China stipulates that production of a

sound recording, for the production of a sound

recording, exploits a musical work which has been

lawfully recorded as a sound recording by another,

does not need to obtain permission from, but shall, as

provided in regulations, pay remuneration to the

where the copyright owner has declared that such

exploitation is not permitted. However, the work of

‘AI Sun Yanzi’is still circulating on the Internet on a

large scale, contrary to the request of Universal Music

Group, which shows that the problem of infringement

is still occurring on a large scale. According to the

data collection and training behaviours in AI Dubbing

are controlled by the reproduction rights of the

producers of sound and video recordings; Article 42

of the Copyright Law of the People's Republic of

China stipulates that a producer of sound recordings

or video recordings shall have the right to permit

other to reproduce, distribute, lease and disseminate

to the public through information network such sound

recordings or video recordings and shall the right to

receive remuneration for it (Zeng and Chen,

2024).Sound recording, which means any aurally

perceptible fixation of the voice and other sounds of

a performance (Huang and Wang, 2021). The

extraction of the singing voice by AI synthesised

music is in fact a copy of the recording (Jiao and

Liang, 2023). Therefore, AI Dubbing requires the

permission and payment of the producer of the sound

recording. However, concerning the attitude of

Universal Music Group and the current situation of

the wide circulation of AI Dubbing covers mentioned

in the article above, a large number of producers of

AI Dubbing covers have not fulfilled their

obligations, thus infringing on the reproduction right

of the producer of the sound recording.

4 THE APPLICATION OF FAIR

USE TO WORKS USED IN AI

DubBing

Based on the above arguments of copyright and

neighbouring rights infringement by AI Dubbing, this

technology does have an impact on the traditional

music industry. But at the same time, AI Dubbing as

an emerging technology is not all bad but no good, AI

Dubbing works in the community widely circulated

at the same time also hit the target performer's fame,

the target performer's original untried style can be in

the AI Dubbing to show the most, for the music

market to inject more vitality and vitality. People

should make active use of copyright exceptions. The

expansion of the proprietary system of intellectual

property and is a constraint on the behaviour of

A Research on Copyright Infringement and Fair Use of AI DubBing

135

potential subjects to monopolise the intellect. In the

practical application of the society, there is a certain

degree of confrontation between the creators, the

distributors, and the users due to the choices of

confrontation, Copyright Law of the People's

Republic of China needs to balance and harmonise

the various conflicts that may exist, and is thus guided

by the balanced spirit of copyright (Jiao and Liang,

2020). People need to prevent overprotection of prior

works in order to achieve the Copyright Law of the

People's Republic of China's goal of encouraging

innovation led by the spirit of balance. Explore the

rationalisation of AI and seek a balance between

promoting AI and protecting the legitimate rights and

interests of copyright holders.

At present, the domestic judgement standard for

fair use mostly adopts the ‘three-step test’ of the

Berne Convention for the Protection of Literary and

Artistic Works, i.e., it can only be used in special

circumstances, cannot conflict with the normal use of

the interests of the copyright owner.

4.1 AI DubBing Meets the Legislative

Purpose of the Copyright Law of

the People's Republic of China

As for the definition of ‘under special circumstances’,

China adopts the form of enumeration. Article 24 of

the Copyright Law of the People's Republic of China

lists the ‘special circumstances’, and only in

accordance with the twelve enumerated

circumstances can it be judged as fair use. When

comparing AI Dubbing with each of the listed

scenarios, it was found that the training of the

Artificial Intelligence Model was difficult to apply to

the twelve ‘special cases’ specified. Examples

include ‘for personal study, research or enjoyment’ or

‘for use by teaching or research personnel in school

classes or for scientific research’.

Nowadays, most scholars take whether it is for

profit as a specific criterion to judge whether fair use

applies, and AI Dubbing covers are mostly posted on

music platforms or self-media ac-counts, so the

following discussion is based on the classification of

the subject.

4.1.1 Music Platform

Regarding the judgement of infringement of AI

Dubbing songs released by music platforms, there are

different views, and some scholars believe that the

judgement of profit-making pur-posefulness needs to

be based on the specific behaviours of music

platforms and others. Some scholars also believe that

for the music platform that can not be identified as

fair use (Zeng and Chen, 2024). Music platforms, as

commercial organisations with a profit-making

purpose, ultimately aim to capture the market and

gain profit from any of their actions. The identity of

the music platform is such that it cannot be considered

as fair use.

4.1.2 Self-Media Accounts

Regarding the release of AI Dubbing songs by a self-

media account, it needs to be judged by several

aspects, the first aspect of judgement is whether the

self-media account belongs to a pri-vate account or a

team account. The second aspect is that the content of

the self-media account should be judged in terms of

profitability in terms of whether the individual AI

Dubbing audios are placed commercially or whether

they are involved in traffic realisation or reward

money along with other content. For the traffic

realisation of AI Dubbing behaviour of natural per-

sons. If the traffic realisation is only limited to a video

of AI Dubbing, it is too one-sided, the fans and traffic

that skyrocketed due to AI Dubbing will be reflected

in the next video released by the account, which will

directly increase the commercial value of the account,

and the AI Dubbing itself is accompanied by a huge

amount of traffic, and the AI Dubbing video, AI Sun

Yanzhi's cover song has even reached a high of more

than 8 million broadcasts, and the huge broadcasts

have obviously exceeded the scope of ‘personal study

or appreciation’, thus it is difficult to be considered as

fair use. The traffic revenue has obviously exceeded

the scope of ‘personal study, research or

appreciation’, so it is difficult to be recognised as fair

use.

4.2 AI DubBing Incorporates a

Pathway to Fair Use

Some scholars have asserted that excluding AI

technology as a fair use solely on the basis of

‘whether or not it is for profit’ would have an impact

on the development of the technology (Xu, 2024).

People can analyze the ways in which AI technology

can be incorporated into fair use in the spirit of

balance mentioned above. Article 1 of the Copyright

Law of the People's Republic of China stipulates that

it ‘promote the development and flourish of socialist

culture and sciences’, reflecting the legislative

purpose of the Copyright Law of the People's

Republic of China to promote technological progress.

IESD 2025 - International Conference on Innovative Education and Social Development

136

Therefore, AI Dubbing can be analysed from the

perspective of technology to incorporate fair use. One

of the basic principles of Copyright Law of the

People's Republic of China, namely the principle of

‘technology neutrality’, requires that copyright law

adopt a tolerant attitude towards technology to

promote technological progress. Technology

neutrality limits the scope of copyright owners' rights

to some extent, balancing new technologies with the

traditional interests of all parties (Craig, 2014).

Standing on the perspective of technology neutrality,

whether the emergence of a new technology can

become an exception to the Copyright Law of the

People's Republic of China on the grounds of

technology neutrality should be compared with the

risks and benefits it brings, and only if the AI

Dubbing can bring sufficient public benefits to the

society, the AI Dubbing can be included into a fair

use.

The public interest attribute is related to the

underlying nature and universality of the technology.

AI Dubbing is actually an AI Sound Synthesis Model,

with sufficient underlying technology brought about

by the supportive nature of the industry, which can be

applied to all sound-related industrial economy, and

is conducive to the promotion of cultural

development and technological advancement, and

therefore has a certain public interest attribute. In

addition, AI Dubbing to generate Sound Synthesis

Models can also bring industrial benefits.AI

technology and business are evolving in convergence,

and the generative AI Dubbing industry is booming,

with detailed data showing that private funding

pouring into generative AI in 2023 has increased

nearly eight-fold compared to 2022, reaching $25.2

billion.1 The AI Dubbing industry is also growing,

with the private sector increasing their funding by

nearly two-fold.(Artificial intelligence

index,2024)The breadth and underlying support of AI

Sound Synthesis Model in all sound economy

industries is an indication of the potential for huge

industrial benefits.

The universality of AI Dubbing lies in the fact that

the sound model formed by Artificial Intelligence

Model is not only the developer who can gain benefits

from it, but also injects more vitality into the music

market and the sound performance market, and

enhances the vitality of the music market by breaking

the conventional sound presentation. Prior works can

be presented to the public in new and more varied

ways, giving the public a different experience while

at the same time contributing to the enhancement of

the activity of prior works for the benefit of copyright

owners.

4.3 The Use of AI DubBing will Have

No Impact on the Normal use of the

Work

Normal use of a work should not be defined as all of

the copyright owner's commercial rights, and

therefore normal use should be understood as whether

the use of a prior work by AI Dub-bing creates a

competitive marketplace for the copyright owner that

harms the copyright owner's interests in the general

of AI Dubbing does not lie in the simple copying of

prior works, but in the mining of new values from

prior works to form independent acoustic models.

Therefore, even if a large number of prior works need

to be copied and invested in when generating the

training acoustic model, the technology will not

produce works that are exactly the same as the prior

works at the end of its operation. On the contrary, AI

Dubbing is a new interpretation and presentation of

the prior work, and the explosion of AI Dubbing will

also bring traffic to the prior work, and will not harm

the interests of the copyright owner. The purpose of

the model output under this technology is not to

reproduce the prior work, but to create new content

using the model formed by the prior work.

4.4 Copyright Owners' Obligation of

Reasonable Tolerance

Artificial Intelligence Model Dubbing as generative

model training, the use of works has a low density.

Model training works from a wide range of sources

and forms, the value of a single work for model

training is small, which is determined by the technical

path of model training, only a large number of large-

scale use of works of artificial intelligence can form

a complete model. The number of works for which

payment is required to train the model is huge, and in

the era of big data, the rapid flow of information

makes it difficult to trace the origin of the works and

to determine the subject of the allocation of fees, thus

making it more difficult for individual works to

obtain licence fees from the AI (Wang, 2022). It

would be an unaffordable and excessive obligation

for the model trainer to have to licence and pay for

each and every work before it could be used (Wang,

2022). This is a difficult problem for the application

and development of AI Dubbing. Therefore, the fair

use of AI Dubbing is particularly important.

A Research on Copyright Infringement and Fair Use of AI DubBing

137

5 CONCLUSION

The article focuses on the fair use of AI Dubbing by

examining the issue of AI Dubbing as a kind of

generative Artificial Intelligence, whose process of

inputting data to generate Acoustic Models

undoubtedly poses a challenge to the existing

one of the twelve special cases of fair use by

interpreting AI Dubbing. In the absence of relevant

legal provisions, judging whether fair use applies to

AI Dubbing solely on the basis of whether or not it is

for profit will similarly have an impact on the

development of AI technology. From the perspective

of the global environment, various countries are

actively exploring a balance between the

development of AI technology and the reasonable

rights and interests of copyright holders. For example,

the Digital Single Market Copyright Directive

introduced by the European Union allows for limited

data mining, providing a favourable legislative

environment for the generative AI industry. The

also exploring the creation of a new fair use with a

view to balancing technological development and the

interests of copyright owners. From a technology-

neutral perspective and guided by the spirit of

balance, the article studies the feasibility of the fair

use of AI Dubbing by analysing several perspectives:

the public and industrial interests brought by AI

Dubbing, and the fact that the use of AI Dubbing does

not cause any normal use of the prior works, and by

relating AI Dubbing to the legislative purpose of the

promoting technological progress. In the course of the

study, it was found that the fair use enumerated in the

traditional system as well as no longer applicable to

explain AI Dubbing, but in practice generative AI

technology is equally imminent, and should not be

solved under the original framework system to solve

new problems. Under the guidance of the spirit of

balance, the interests and needs of all parties should

be analysed and adapted to new developments on the

basis of the original foundation, so as to provide a

legislative basis for the development of generative

artificial intelligence technology.

REFERENCES

Artificial intelligence index 2024. (2024-04-17). https://ai-

index.stanford.edu/wp-content/up-

loads/2024/04/HAI_2024_AI-Index-Report.pdf.

in AI Dubbing. Shanghai Business, (11)

Craig, C J. 2014. Technological neutrality: (Pre) serving the

propose of copyright law, The Copyright Pentalogy:

How the Supreme Court of Canada Shook the Founda-

tions of Canadian Copy-right (Ottawa: University of

Ottawa Pres, 2013), Os-good Halllaws School Legal

Studies Research Paper, 28.

Huang, W.& Wang, L. Introduction to the Copyright Law

of the People's Republic of China and its Interpretation,

China Democracy and Legal System Publishing House.

2021.

cial Intelligence Synthesized Music and Their Resolu-

tion. Intellectual Property, (11)

Press, 295.

be protected in the AI windfall, Beijing Youth Daily,

A07.

Wang, W. M. 2022. Challenge and Response of Artificial

Intelligence for the Restriction and Exception rules of

Copyrights. Journal of Law Application, (11).

(Fourth Edition), China Renmin University Press.

gence Model from Technology Neutrality Perspective.

Law Review, 42 (04):86-99.

Zeng, M. & Chen, N. 2024. A Study of the Legal Risks of

AI Dubbing. Jiangnan Forum, (01).

Zeng, M. & Chen, N. 2024. Legal Disputes and Risk Re-

sponses Regarding AI Music Covers. Tribune of Study,

(04):125-136.

IESD 2025 - International Conference on Innovative Education and Social Development

138