Chenxi He

Law School, Anhui University, Hefei, Anhui, 230601, China

Keywords: Generative AI, AI Painting, Subject of Copyright.

Abstract: Nowadays, AI painting technology has rapidly become popular, and the resulting issue of copyright ownership

has increasingly become the focus of attention from all walks of society. This article aims to explore this hot

issue in depth. It takes the ‘legal status of AI’ and ‘the allocation of copyright’ as the starting point for the

research, and then carry out detailed analysis of the dilemma faced by the copyright ownership of AI painting

works. By analyzing the views of different scholars and cases, this article finds that the focus of the

controversy is concentrated on three key points. Firstly, whether artificial intelligence has legal subject

qualification and enjoy independent rights. Secondly, the characteristics and rights attribution model of AI

painting works. Thirdly, the profit allocation mechanism between developers and users. Through research,

the author found that AI cannot be a subject of rights and software developers as rights holders or developers

sharing rights with users is not appropriate. Based on these conclusions, this study argues that the copyright

of AI painting works is most feasible and reasonable to belong to the user of the software, and it can also be

clarified in advance through the user agreement.

1 INTRODUCTION

In recent years, with the continuous development of

generative artificial intelligence technology, AI

painting tools such as DALL-E and Stable Diffusion

have emerged as the times. They can learn from

image data in a large database and generate artistic

images autonomously according to text instructions.

The emergence of AI painting has brought great

impact to society. On the one hand, this technology

has reduced the threshold of artistic creation to some

extent, encouraging people's enthusiasm for creation.

For example, by applying the text-to-image

technology, software users can create artworks

simply through text. However, on the other hand, the

popularity of AI painting poses severe challenges to

the traditional copyright system. The issue of

human only provides initial text instructions and the

AI subjectively creates has sparked a great deal of

controversy and discussion.

In the issue of copyright ownership of AI painting,

the main controversial point lies in whether the

or the software developer. Some scholars believe that

the work of AI always needs human intervention and

can only be regarded as an auxiliary tool, so it can not

be the owner of copyright, but the user and the

developer of the software both participated in the

creative process and thus could both be considered

the copyright owner (Li, 2024). There are also some

scholars have proposed to grant AI a fictitious subject

status, making AI the owner of the copyright of such

paintings (Wu & Chen, 2024).

At present, "human-centered" is firmly supported

in China. In 2023, in the "First Case of AI Text-to-

Image", China recognized that natural persons

painting large models under certain conditions for the

first time, which aims to encourage the use of

artificial intelligence creation (PKULAW.COM,

2024). But the conditions mentioned in it are still to

be clarified, and there are still many legal gaps in

judicial practice.

The attribution issue of AI painting has not yet

been clearly regulated. Based on the different

viewpoints of other scholars, this article will discuss

the legal status of AI and the copyright allocation of

AI paintings, hoping to provide a force and help to

solve the problem of copyright attribution of AI-

generated art.

210

He, C.

DOI: 10.5220/0014359600004859

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

In Proceedings of the 1st International Conference on Politics, Law, and Social Science (ICPLSS 2025), pages 210-214

ISBN: 978-989-758-785-6

2 ANALYSIS OF THE

DIFFERENCES IN THE

CREATION OF AI PAINTING

AND OTHER TYPES OF

AI-GENERATED WORKS

The issue of copyright ownership of AI-generated

paintings shares similarities in core controversies

with that of other types of AI-generated works, such

as texts, music, code, etc. But there are some

differences due to the technical process

(implementation method, creative process)

characteristics and differences in legal practice.

2.1 Technical Process and the

Difference in Human Participation

AI painting usually relies on the prompt words and

parameter adjustments input by the user. Humans

need to repeatedly adjust the prompt words, screen

the results, and make post-edits, which may be

regarded as a higher degree of "creativity." For

example, the creator of the piece Space Opera, Jason

Allen, used AI painting software, spent over 80 hours

and 900 revisions, then polished by using image

editing software to shape the final piece (Li, 2024).

However, for text, music and code generation, the

user input may be shorter and the generated results

would often require less manual adjustment. For

example, an article generated by GPT might only

need polishing, whereas when composing music, AI

might directly output a complete score. Human

participation tends to be more about "selection" than

"creation," making it more difficult to establish

originality in legal terms.

2.2 Differences in Legal Practice

In the "First Case of AI Text-to-Image ", China

recognized that natural persons have copyright over

the images generated using AI painting large models

under certain conditions, aiming at encouraging the

use of AI for creation (PKULAW.COM, 2024). The

long-term presence of "instrumentalism" in the art

may make AI be accepted as a tool more easily. Some

scholars point out from the perspective of

philosophical instrumentalism that AI is the "product

of human technological development" and only exists

as a tool in artistic creation. At this stage, AI can

imitate the style of artists or assist in creation, but it

has not reached the level of replacing artists' creation,

emphasizing that AI is more likely to be accepted as

a tool by people because of its auxiliary nature (Yi,

Unknown). Additionally, artistic creation has

traditionally emphasized the result-oriented

approach, rather than the process-oriented approach.

Traditional painting emphasizes the "audience's

reception" and social aesthetic consensus, historical

technological shifts (such as the advent of

photography) also forced artists to turn to process

innovation; and literary or musical fields have more

strict requirements for "originality," traditionally

emphasizing the author's direct expression, a creative

process with personal uniqueness (Yi, Unknown).

3 EXPLORATION OF THE ISSUE

OF COPYRIGHT OWNERSHIP

IN AI PAINTING

There are three possible rights holders for the

developers, and the software users. The main issues

of controversy focus on whether AI has the legal

personality to be the copyright owner and how the

rights and interests between software developers and

software users are allocated.

3.1 Analysis of the Legal Subject

Qualification of AI as the Author of

3.1.1 AI Cannot Be a Legal Person

Some scholars explain from the perspective of legal

personality, believing that AI essentially exists as a

machine tool and does not have social attributes. It

does not conform to the legal subject attributes of

natural persons, legal persons and other

organizations, and therefore cannot enjoy copyright

(Wu & Chen, 2024). In addition, AI lacks physical

organization, independent property, and risk-bearing

capacity, which do not conform to the characteristics

of a legal person (Wu & Chen, 2024). In other words,

viewed as a tool created by human beings, AI does

not possess the awareness of rights or subjectivity,

and the nature of rights is a legal relationship

regulates human behavior, so AI, as an object, cannot

enjoy independent rights, and therefore AI cannot

become the subject of property rights (Ma & Yang,

2024). Some scholars also explain it from the

perspective of AI assistance and tool attributes,

proposing that the essence of AI is a tool, which relies

on human intervention and lacks consciousness, so it

is impossible to complete creation independently, and

211

therefore it is impossible to become a rights subject

to enjoy copyright (Li, 2024). There are also scholars

who oppose the "man-machine dichotomy" and the

"algorithm-only" view, believing that the former

mistakenly regards AI as a subject, ignoring its

instrumental nature; and the latter unilaterally

emphasizes the determinacy of the algorithm,

ignoring the creative contribution of human beings at

the input end (Guo & Li, 2024). Beyond these, from

the perspective of philosophical instrumentalism, the

subject needs to possess consciousness and practical

ability, while AI is merely the product of human

technological development and its agency completely

depends on humans (Yi, Unknown). From the

perspective of the art field, the core of art creation is

emotional connection, but AI can only generate

aesthetically pleasing images through data analysis,

which is unable to experience or convey human

emotions. And art needs to reflect the spirit of the

times and convey individual emotions, while AI

cannot understand the social and cultural background

or actively adapt to changes in aesthetics (Yi,

Unknown). Therefore, AI is not suitable to be the

subject of artistic creation.

3.1.2 The Disadvantages of AI-Fictitious

Subject

Based on the preliminary exploration of the fictitious

personality of AI in relevant legislation of the

European Union and the United Kingdom and other

countries, some scholars believe that the law endows

fictitious subjects with legal personality, avoiding the

potential risk of infringement in AI painting,

encouraging people to participate in creation and the

convenience of the attribution and protection of

painting works (Wu & Chen, 2024). However, it is

untenable to personify AI as a civil subject, because

AI has no independent property and no capacity for

civil liability (Sui, 2024). Conferring legal

personality on AI is not only contrary to the current

social consensus, but also is prone to ethical and

moral risks. Additionally, it is incompatible with our

civil law legal subjects and is likely to have a

subversive impact on our civil subject related laws as

well (Wu & Chen, 2024).

For the two viewpoints, the author believes that

the former should be supported more. In the author's

view, the core characteristic of a legal person lies in

its ability to independently bear rights and

obligations. AI, as a tool created by human beings,

lacks self-awareness and independent will, and thus

is unable to bear legal responsibility and fulfill legal

obligations. If there is a situation where AI infringes

on someone else's copyright, it is meaningless to hold

AI accountable. Moreover, AI does not have the

ability to exercise power, and there is a high

probability that it will lead to the situation where AI

appears to enjoy rights, but in fact, it is exercised by

others on behalf of it, which means it is still human

beings that exercises power rather than AI. If AI is

given subject status by law to become a legal entity,

some of the infringement risks that may exist in AI

painting can indeed be avoided. But at the same time,

there is also the potential for people to use the legal

personality of AI to shift blame, evade obligations

that should be borne, and damage public interests.

Besides, from the perspective of social ethics, modern

law follows a "human-centered". Endowing AI with

subject status will blur the boundary between "tool

and "subject", causing a series of ethical problems

and impacting the current law. So, AI can only be

used as a tool but cannot be conferred with legal

personality.

3.2 Analysis of Copyright Ownership

of Software Developers

Some scholars have proposed that although AI-

generated artworks may not appear to be directly

created by humans, they are fundamentally based on

algorithm technologies developed by humans,

reflecting the personalized expression of participants.

Therefore, its copyright should belong to the

enterprise or developer to whom the technology

belongs. For example, in the case of Tencent

prosecuting Shanghai Yingxun Technology, the court

ruled that the AI-generated article constituted a legal

entity work, and Tencent held the copyright (Zhou,

2024). Although some scholars support this view,

arguing that while developers do not directly

participate in the generative process, their algorithm

design forms the foundation of AI creation, and as

technology providers, they make decisive

contributions to the underlying architecture of the

generated works (Wang, 2023). Moreover, software

developers have already obtained investment returns

through software copyrights, and attributing the

owner is more conducive to technology promotion

and maintenance (Lu, 2023). However, based on this

perspective, the issue of dispersed rights subjects

remains to be resolved. The development of AI

models involves multiple teams or institutions,

making it difficult to determine the rights holders

(Wang, 2023). Besides, this perspective only partially

considers the personalized involvement of software

developers while neglecting the original creative

ICPLSS 2025 - International Conference on Politics, Law, and Social Science

212

expression of software users when utilizing AI to

produce artistic works.

In addition, there are also some platforms (such as

"Wenxin Yige") that explicitly stipulate in their user

agreements that the intellectual property rights of AI-

generated works belong to the company (Zhou,

2024). The author believes that the point that the

intellectual property rights of AI-generated works can

be clarified through the user agreement can be

supported, since the user agreement is established

through mutual negotiation between the software

developer and the user, the user's utilization of the

software signifies recognition of the intellectual

property ownership. This approach can effectively

circumvent issues related to profit distribution, rights

division, and risk allocation between developers and

users. Therefore, the copyright of AI-generated

artwork can be clearly defined in advance through

user agreements to determine ownership.

3.3 Analysis of Copyright Ownership

of Software Users

Some scholars argue that copyright ownership should

generally follow the principle of "belonging to the

user". Firstly, user contribution is the most direct, as

users facilitate the generation of works by paying

fees, inputting instructions, and providing materials.

Secondly, designers should not reap double benefits.

Having already obtained protection through software

Thirdly, by analogy with the legal person system,

treating developers as authors is inapplicable, as

developers have already profited through technical

licensing, making dual protection unreasonable (Ma

& Yang, 2024). Fourthly, universal ownership is

unfeasible, as it would significantly dampen the

enthusiasm of both developers and users, while the

ambiguity of rights holders would make it difficult to

safeguard rights (Zhou, 2024). However, this

viewpoint is controversial. Supporters argue that the

attribution of property rights needs to balance the

interests of developers, owners, and users, but it is

more reasonable for users to enjoy exclusive property

rights. Because developers profit through service

charges rather than by dividing the intellectual

property rights of user works, and users exclusively

retaining these rights can stimulate public creative

enthusiasm and promote the sustainable development

of AI art (Zhou, 2024). Opponents argue that users'

input of prompts, parameter settings, and adjustment

behaviors in AI painting fall under the category of

"ideation" guidance or literary works, and do not

directly participate in the creation of artistic

expression, thus should not be entitled to copyright

(Zheng & Zhang, 2024). Granting users copyright

may also lead to contradictions in legal logic and

could increase the infringement risks for AI

developers, which would ultimately hinder the

development of artificial intelligence technology

(Zheng & Zhang, 2024). However, the author

believes that compared to other perspectives,

attributing the copyright of AI-generated artwork to

the software user is the most feasible solution. In

China's "First AI Text-to-Image Case", the copyright

of AI-generated artwork was granted to the software

user (PKULAW.COM, 2024). This reflects the value

guidance of the "human-centered principle" upheld

by our country in technological development. It can

effectively achieve the core goal of copyright law to

encourage the creation of works" by granting the

By clarifying the ownership of rights, user motivation

can be stimulated, encouraging more people to

leverage AI tools for creating high-quality works and

reinforcing human dominance in technological

applications (Zhou, 2024). Therefore, at present, the

solution of copyright attribution to AI painting to the

software user not only follows the "human-centered"

view, but also conform to the current judicial practice,

and also achieves the core goal of the Copyright Law,

which has the highest feasibility and rationality.

3.4 Analysis of Joint Ownership of

and Users According to

Contribution

Some scholars believe that the attribution of

of both the developers and the users. Because the

developers design the algorithm and training data,

providing the creative basis for AI; and the users

generate specific works through instructions, the two

together constitute a cooperative by default. And in

accordance with the rules on joint works in the

distributed by both parties. For example, China's

GPen platform attributes copyright to the user (Li,

2024). However, attributing the copyright of AI

painting by solely considering the contributions of

developers and users has evident shortcomings. One

reason is that contributions are difficult to quantify,

making it challenging to balance the distribution of

benefits and division of rights between developers

and users. The second reason is the difficulty in

distinguishing the responsibilities and obligations

213

that should be borne by developers versus users when

facing legal risk liabilities.

3.5 Compromise Solution: Protection

of Neighboring Rights

The proposer argues that the protection of narrow

determining authorship, and suggests that AI-

generated content should be protected through

neighboring rights systems. The proposer argues that

the protection of narrow copyright faces dilemmas in

judging originality and determining authorship, and

suggests that AI-generated content should be

protected through neighboring rights systems.

Therefore, the protection of neighboring rights can

circumvent disputes over legal subject qualification,

balance the interests of all parties, and adapt to the

needs of technological development (Sui, 2024).

Moreover, the ownership of rights should be vested

in the user, with the content primarily governed by

property rights, and a protection period limitation

should be established (Sui, 2024). However, in

reality, the solutions for protecting neighboring rights

also have significant drawbacks. Firstly, since

neighboring rights do not require the generator to be

a natural person nor necessitate originality,

incorporating AI painting into the neighboring rights

system may implicitly acknowledge the protectability

of "non-original generated content," which

contradicts the legislative purpose of copyright law to

"encourage originality". Secondly, the rights subject

of neighboring rights cannot be clearly defined.

Although some scholars have proposed that the

attribution of rights should be granted to users.

However, if neighboring rights are granted to AI

users, developers may argue based on the "dominance

of algorithm design", shifting the dispute from

"copyright ownership" to "allocation of neighboring

rights" without fundamentally resolving the issue.

4 CONCLUSION

The issue of copyright ownership of AI-generated art

is essentially a conflict between technological

innovation and legal lag. This article argues that it is

more reasonable and feasible to attribute copyright to

software users or to clarify the ownership of

the one hand, the users participate in the creative

process through behaviors such as instruction design

and parameter adjustment, and their intellectual input

meets the requirement of ‘originality’ under the

be clarified in advance by user agreement, which can

resolve the disputes of the subject according to the

principle of autonomy of private. However, it is

necessary to pay attention to whether there are

unfavorable terms, and avoid developers from

invalidating the rights and interests of users through

standard terms. So, future legislation should establish

a dual safeguard mechanism–recognizing the

effectiveness of the user agreement while setting

minimum rights retention rules, and clarify the

standards for the allocation of risk between

developers and users in AI tort liability. Only if it is

limited by law while respecting the autonomy of

expression can AI creation develop sustainably.

REFERENCES

from AI painting in the AIGC era. Leg. Syst. Panor. 1-

H. Zheng, W. Zhang, 2024. The original contribution of

users in AI-generated images - Taking the copyright

case judgment of 'AI text-to-image' as an example.

China Publ. 53-57.

L. Ma, D. Yang, 2024. Identity, property rights,

competitiveness: Three fundamental dimensions for

understanding AI painting. Fine Arts 14-15.

generated content - Taking China's first 'AI text-to-

image case' as an example. Publ. Print. 11-18.

P. Guo, Z. Li, 2024. On the qualification of complex AI-

generated works in copyright law - Concurrently

reviewing the 'AI text-to-image copyright case'. Sci.

Technol. Law 77-78.

generated art belong?. Fine Arts 44-45.

S. Yi, Between art and technology: Artificial intelligence

and painting creation from the perspective of

instrumentalism. J. Jilin Univ. Arts 41-46, 178.

The Beijing Internet Court released one of the top ten

dispute between Li and Liu - the "AI-generated text-to-

https://www.pkulaw.com/pfnl/08df102e7c10f206b28e

a71d0a826224c5a46d3dea394abbbdfb.html

(Accessed: 2025-5-18)

W. Li, 2024. Exploring the nature and protection of AI-

generated paintings from the perspective of copyright

law. Commun. Copyright 108-109.

Forum Rule Law 286.

Y. Zhou, 2024. Reflections on the rights of AI painting and

countermeasures. Sci. Technol. Commun. 16(10), 16-

17.

ICPLSS 2025 - International Conference on Politics, Law, and Social Science

214