Balancing interests in copyright protection for artificial intelligence outputs - International legal practice and some recommendations for Vietnam [1]

Abstract: The development of creative AI and its penetration into the cultural field is raising important global debates about the purpose and scope of copyright law in ensuring and maintaining a balance of interests between creators and investors and the interests of society to ensure the goal of encouraging creativity and innovation. This article examines international legal practices in addressing the issue of copyright protection for products generated by artificial intelligence, from the perspective of balancing interests. Based on this analysis, it offers several recommendations for Vietnam to adapt its legal framework in response to emerging challenges.

Keywords: Copyright, AI, works, balance of interests

1. Introduction

The core principle of copyright protection is to encourage the creation of new works by giving creators the exclusive right to control exploitation activities, the ability to license or transfer copyright so that they can gain economic benefits. This monopoly is compensation for creative efforts and investments in creating cultural and artistic works, in order to create incentives to encourage and continue to nurture creative activities. On the other hand, copyright law must also ensure that this monopoly does not unduly restrict access to knowledge, culture, hinder creativity, or access and sharing of information by the public. Copyright law, since its inception, has always faced challenges in improving to maintain a reasonable balance between the interests of creators and investors and the interests of the public and users of works to ensure the goal of encouraging creativity, while ensuring that society benefits from access to culture, knowledge and innovation. The rapid development of creative AI (GenAI) and its penetration into innovative activities, including the cultural field, is raising important debates about the purpose and scope of copyright law in ensuring and maintaining a balance of interests. Was traditional copyright law initially established to encourage human creativity still appropriate when AI plays the role of the subjects creating works? If copyright protection is provided for AI output works, who will be granted exclusive rights to ensure the goal of encouraging creativity and innovation? How can Vietnam adjust its legal framework to meet these emerging challenges?

2. Copyright protection for AI-generated works – an approach from the perspective of balancing interests

The rapid development of new-generation AI creativity and the increase of AI-generated works alongside traditional human creative works raise the question of whether to grant copyright protection for AI outputs. The opposition to copyright protection for AI creations is based on the following main arguments: First, protecting AI-generated works will not promote “creative progress” but may even neutralize creative activity by a series of privately owned algorithmic works [2]. With the argument that AI creates works automatically, there is no need to encourage or protect its material or spiritual rights, so the goal of encouraging creativity of copyright law becomes meaningless. The development of AI's creative capabilities may make artists complacent and dependent on technology, leading to the risk that AI-generated works may gradually replace human creations, causing artists to lose their jobs, affecting the goal of encouraging creativity under copyright law [3]. The use of AI to create copyrighted works raises concerns that it will blur the line between human and machine creativity [4], lowering the value of human creativity when it is placed on the same level as machines. Second, allowing AI-generated works to be in the public domain is necessary to encourage and facilitate cultural development. AI technology enables the creation and dissemination of millions of works on a large scale in a short time. Copyright protection for AI-generated works would impose costs on users and expose them to legal risks without the permission of the owner and would also dilute the public domain. This could tilt the balance of interests towards a small group of organizations and large companies that own AI systems, negatively impacting innovation and competition [5]. Determining AI-generated works in the public domain is not about declaring them valueless, but about keeping them out of monopolistic control that could enrich the actors behind the AI ​​- those who invest in, create, own, train or deploy the AI ​​[6].

In contrast, the view in favor of copyright protection for AI-generated works is based on the fundamental theories of copyright protection to protect the rights of creators and encourage innovation. John Locke's “natural rights theory”, with its argument that creators have a natural right to the “fruits of the labor”, explains that the subject who creates intellectual achievements has the right to own, control and gain benefits from exploiting them [7]. “Utilitarian theory”, focusing on the economic aspect of copyright, explains the granting of exclusive rights as “adding fuel to the fire of genius”, adding fuel to motivate creativity and innovation [8]. The side that supports this argument argues that if AI output is not copyrighted, it will belong to “public property”, meaning that all in society will be free to use it [9]. Investors and creators will not be rewarded properly for their creative efforts and investments, thereby reducing creative enthusiasm. Copyright protection for AI output works aims to protect the legitimate rights of entities investing in AI research and development, providing them with legal rights to exploit, use and prevent unauthorized use by other entities, and creating incentives to encourage investment and development of AI.

In reality, we are witnessing AI as the subject of creative works, but it is clear that AI is just a “creating machine” and behind that creative process there is often an important human role. Although the development of creative AI is promising a future scenario of a completely autonomous and automatic creative robot, it is still difficult to deny the connection between AI output and human intervention. The connection of human intervention is reflected in the development of AI software, the collection and selection of AI training data, the construction of functional specifications, monitoring the creative process, and editing AI output works [10]. The Industrial Revolution 4.0 with the emergence of AI has provided humans with more effective “tools” to support creative activities, making creative works increasingly more diverse, of higher quality and commercial value, and creative activities can save more time and effort. From this perspective, copyright protection for creative AI-assisted works still ensures the goal of copyright law to encourage creativity. Traditional copyright law has historically had flexible applications and reforms as technology has developed. AI is the result of technological development, and therefore should be considered within the flexible scope of copyright law historically. Therefore, it is important not to not recognize copyright for AI works, but to have a suitable legal mechanism for works created by AI, to encourage investment in AI [11].

3. Conditions for protection of works for AI outputs – international legal practice

The development of creative AI today shows that AI has been able to create almost all types of literary and artistic works mentioned in Article 2 of the Berne Convention, including literary works (novels, poems, articles), music, works of art, films, photographs, etc. The use of AI in creating works is becoming an increasingly popular and inevitable trend in artistic creative activities. From there, the important question is whether AI outputs meet the conditions of creativity to be protected by copyright? Since its inception, although the approach to copyright and the scope of copyrighted works have been different over time and in different legal regions, copyright laws around the world have agreed on a common standard for a work to be protected: it must have “creativity” or “originality”. The creativity/originality of a copyrighted work is a combination of two factors: (i) the work must be a “creative product” and (ii) it must be a “creative product of the mind” or “creative spirit”, that is, related to the human element. The requirement of “creativity of the mind” has excluded works created without any human intervention or intellectual effort from the scope of copyright protection. However, this requirement does not exclude creations made by humans with the help of machines, provided that the human contribution to the product meets the legal standard of creativity/originality [12]. So, what makes an AI-generated work unique/creative? Most computer-generated works rely heavily on the underlying algorithms and inputs provided by programmers, while computers are considered tools for creating works of art. Copyright protection for AI outputs will depend on whether the work is considered “the result of a machine-generated process” or whether it reflects “the author’s own mental conceptions" [13].

The legal practice in many countries around the world shows that the basis for copyright protection for AI-generated works is based on considering the requirement of creativity/originality of the AI ​​output and the role of humans in the AI ​​creative process. The requirement of “creativity” of copyrighted works has changed through case law in Australia. In 2002, in the case of Desktop Marketing Systems Pty Ltd v Telstra Corporation Ltd (2002) 119 FCR 491, the Federal Court applied the “Sweat of the brow theory” to evaluate the efforts and costs of creating telephone directory entries and concluded that the “collection” of telephone directory entries of Telstra Company met the requirement of original creativity to be protected by copyright [14]. Eight years later, in 2010, the Federal Court of Australia in Telstra Corporation & Anor v Phone Directories Company Pty Limited & Ors (2010) FCA 44 took a different approach to copyright in collections. The Court argued that in order to qualify as an “original creation” for copyright protection, the expression of the collection must involve human creativity or intellectual effort [15]. Similarly, the case of Acohs Pty Ltd v Ucorp Pty Ltd (2012) also held that a work created with the intervention of a computer is not protected by copyright because it was not created by a human [16]. This decision suggests that the owners of AI systems will have a more difficult time establishing copyright in works created through the use of computer programs, especially when they cannot show the creative imprint of specific individuals in the creation of the work.

The European Union (EU)’s law requires creativity/originality “if the author has shown his creativity in the process of creating the work by making creative choices freely” [17]. In the Football Dataco case (Case C-604/10 – Football Dataco), the European Court of Justice (CJEU) flatly rejected the “substantial skill and labor” of the producer of the football match schedule as a relevant factor for originality. According to the Court, the effort and skill of the producer of the football match schedule did not demonstrate any creativity in the selection and arrangement of the data. Similarly, in the Funke Medien case (Case C-469/17 – Funke Medien) the Court considered “the mere intellectual effort and skill in producing the reports (on military situations) are not relevant to the question of creativity” [18]. This means that if a product (such as a collection) is the result of diligent work without the intervention of the author (human) in selecting, arranging, or expressing it, it is not protected by copyright. Although the European Court of Justice (CJEU) has not yet ruled on the issue of computer-generated works, there have been a number of national case laws that have recognized copyright protection for computer-generated works. For example, the Paris District Court (Paris District Court, 5 July 2000, No. 97/24872 in Matt Cooper v. Ogilvy and Mather) held that “in the case of computer-assisted musical composition, if there is human intervention such as the author’s selection, the work is original”. Similarly, the Bordeaux Court of Appeal (Bordeaux Court of Appeal, 31 January 2005, No. 03/05512) stated that “a work generated by a computer system may benefit from copyright protection, provided that it exhibits even a minimal degree of originality intended by its creator” [19].

17 U.S.C. § 102(a) of the United States Copyright Act requires a work to meet certain requirements for protection, including that it must “contain a minimal amount of creativity” [20]. One of the typical precedents often cited in this regard is Feist Publications v Rural Telephone Service Company, Inc. 499 U.S. 340 (1991), in which the Court held that the alphabetical telephone directory, although independently created at Feist's expense and effort, did not meet the minimum creative requirement for copyright protection [21]. In response to the growing demand for copyright registration for AI-generated works, as exemplified by Ms. Kashtanova’s application for copyright registration for her AI-generated comic book “Zarya of the Dawn” [22], the United States Copyright Office (USCO) issued “Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence” [23], which affirms the policy: “Copyright is reserved for human created works only” and USCO will only consider copyright protection for works containing AI-generated content if there is sufficient human creative element, depending on how the AI ​​tool operates and how it is used to create the final work.

Article 2(1) of the Japanese Copyright Act stipulates that a copyrightable work is a work that expresses “human thoughts and feelings”. A work with human input and AI-generated content may qualify for copyright protection in general, based on certain factors: The number and content of the AI ​​user’s instructions and input prompts; The number of attempts to create and modify the output so that the AI ​​user achieves the desired result; The AI ​​user selects the work from many generated works; Subsequent human modifications to the AI-generated work [24].

It can be seen that the laws of the European Union or many countries require that the condition for an AI-generated work to be protected by copyright is that it must have a certain degree of human creative imprint. Compared to the above countries, the Copyright Laws of the UK, Ireland, and New Zealand seem to have a more open approach in establishing separate copyright regulations for “computer-generated works”, even without the creation of a human author [25]. The exception to copyright protection for works created by computers is granted to the person who made the necessary arrangements for the creation of that work.

Through research on international law and practice, it can be seen that current jurisdictions and countries are quite consistent in refusing to grant copyright to works created by AI without any intervention of human creative thinking. Although there are different approaches to the “creative standard”, many countries accept copyright protection for “AI-assisted” works if it can be demonstrated that a human creative role (or creative intervention) was played at any stage in the creation of the work.

4. Copyright Ownership of AI-Generated Works – International Legal Practice

The principle of balancing interests in copyright law mainly considers the interests between two groups of subjects: Copyright holders and Copyright users. Approached from the perspective of copyright holders, copyright protection provides an economic incentive to encourage creativity. Creators will tend to create more works when they know that they will receive a fair reward for their efforts [26]. Although not denying the principle of copyright encouraging creativity by rewarding creators, from the opposite perspective, copyright users argue that expanding the rights of owners will narrow the public domain - where everyone is free to use copyrighted materials. Users argue that the public domain is essential for creativity because it provides creators with a valuable repository of material from which they can develop new works, and access to this repository encourages creativity and innovation. Conversely, for works created by AI, who should be granted copyright to ensure a balance of interests?

Copyright law does not grant copyright to animals or machines, so AI is not an exception to the granting of copyright. In copyright law, the terms “Author’s right” and “Copyright” are not entirely consistent in their approach. Countries following the Continental European system place the author at the center of the copyright system and therefore focus on protecting the rights of creators. Countries following this system use the term “Author’s right” to identify authors who have direct rights to their intellectual creations. In these countries, copyright is only established when the work is created by an individual author. Countries following the Anglo-American system approach copyright from the economic aspect, according to which the intellectual product will be absolutely owned by its creator. The term “Copyright” means the right to own copyright. This right is often reserved for entities that invest in creating works such as publishers, companies that produce and exploit cultural works.

Considering the case of works created by AI, the laws of countries following the Continental European system will require that the creative subject be clearly identified as a human in order for the work to be protected. Meanwhile, countries following the Anglo-American system can grant copyright to individuals or legal entities – the employer of the creative author. Although the creative subject is often in a different position from the copyright owner because the exclusive rights to control, exploit economic benefits and enforce rights belong to the copyright holder. But in terms of interests, creators are often seen as being in the same group as copyright holders because they share the same interest in protecting their rights and being rewarded for their creative efforts and investment of labor and intelligence. Regardless of the approach, copyright law is unified in the common point of protecting the interests of both the creative subject (author) and the investor - the copyright owner.

The European Court of Justice often relies on a three-stage model of creation: conception, execution, and editing to determine who is the author of a work. For the process of creating a work with the support of machines or AI, the creative intervention of the author can occur at different levels and stages: from preparation, execution, and completion of the product. The output of AI is likely to be eligible for copyright protection if it “reflects human creative choices at any stage of the production process” [27].

In countries following the Anglo-American system, the subject can be considered the author if they are the person behind the control, using AI as a tool to create the work according to the idea, the way of expression they want. Section 178 of The UK Copyright, Designs and Patents Act 1988 (CDPA), recognises copyright in “computer-generated works” whereby the author of a work (for copyright purposes) is “the person who makes the arrangements necessary for the creation of the work” [28], which includes the copyright owner of the computer program, the creator of the computer system, or both. For AI outputs, depending on the case, the person who invests in research, development and training of AI, or the user of AI, may act as the person who makes the creative choice so that the AI ​​output qualifies for copyright protection.

For the creation of a work, the conceptualization stage includes the formation of a general idea for a work (theme), building a series of designs from the selection of genre, style, technique, material, method, form of expression [29], etc. These creative choices at the pre-production stage can be an important factor in creating the uniqueness of the final product. However, in the case of output generated with the support of machine learning (ML) algorithms, creative choices will depend on input data and other parameters, so the AI ​​system at this stage “has no role in the creative process other than acting as an external constraint limiting the designer’s creative ability” [30]. For some current AI technologies such as Chat GPT, the user's role is simply to "enter commands" for AI to create a product on demand, or the AI ​​user simply presses a button to create a product (like some natural language generation software, text generation) is not considered as direct intervention to create the work.

The implementation stage includes converting the design or plan from the sketch (rough) version into the final work. With traditional forms of creativity, the role of the human author at this stage is very important, because the author is the one who converts the plot into the words of the novel, converts musical ideas into music with sounds, converts ideas into paintings with vivid lines and colors of artistic nature. From the 19th century onwards, with the development of film technology, recording equipment, and machines play an important role in the creative process. For example, it is impossible to create photos and movies without cameras or camcorders. However, humans are always in complete control of the creation stage [31]. For works created by AI, the output depends heavily on the data provided to train and educate AI. Therefore, the owner and provider of input data for AI plays an important role in the output of AI. However, the development of machine learning algorithms (ML), especially Deep Learning systems, has created a significant change in the “support” role of machines when the AI ​​system is an automatic operating system that the user does not fully control or determine with certainty the output of AI. Therefore, the quality of AI output “is not necessarily proof of human creativity” in the implementation process.

The editing phase involves processing and revising the versions created during the execution phase to create a final finished product before it is widely released or published. In some cases, AI-powered output is a ready-to-use product, for example, an automated translation by DeepL and Google Translate. However, in many cases, AI output still requires human intervention in editing and revising to create a finished work before it is published or commercially exploited. In this case, creativity is expressed in the author's choice of necessary edits to the AI ​​ output.

Thus, in relation to AI-assisted creations, the “potential candidates” for the position of “right holder” can be one of the following: the developer or programmer of the AI ​​ system; the person providing input data to train the AI; or the user of the AI ​​system. Any entity that deserves to be granted copyright rights must also rely on the principle of balancing interests to ensure the goal of encouraging creativity.

The developer or programmer of the AI ​​system is the creator of the AI ​​system, and can play a creative role in the preparation, implementation, and completion stages of the work. They not only play the role of creative selection in the pre-production stage (conceptualization, design, etc.), but can also be the trainer, provide input data for the AI, or decide to edit and select a certain output for the AI. They are the ones who put in the main creative effort and investment, and also the ones who have to bear the costs and risks in the process of developing the AI ​​system. From the perspective of balancing interests, granting copyright to AI system developers or programmers is reasonable to compensate them for their creative efforts and investments and to create incentives to encourage the development of AI technology. However, granting copyright to AI developers still faces some concerns: First, according to copyright law, copyright is closely linked to the process of creating a work. If copyright is granted to AI programmers or producers, it will invisibly remove copyright from the creative process [32]; Second, due to the increasingly "autonomous" and "unpredictable" nature of AI, it is possible that the output of AI will no longer be related to the original requirements of the programmer. What the work created by AI will be like may be beyond the prediction and control of the programmer. In other words, that work is not a creative product that reflects the thoughts, emotions or intelligence of the programmer [33]. Therefore, granting copyright in this situation is no longer appropriate.

To realize the national strategy on artificial intelligence, the French Republic has made positive moves in amending the law to address the core issue of the use of copyrighted content by generative AI technologies. On September 12, 2023, Bill No. 1630, aiming at “protecting artificial intelligence through copyright”, with four important amendments to the Intellectual Property Code (IPC), was submitted to the French National Assembly, proposing to expand Article L321-2 of the IP Code: “The legitimate subjects of a work generated by AI without human intervention shall be the authors or owners of the rights of the works used to create that work”. At the same time, it is necessary to supplement the provisions on the role of collective management organizations in collecting money and distributing rights on AI works” [34]. According to this proposal, if AI creates a work completely automatically without direct human intervention, copyright ownership will belong to the copyright holders of the input data used by AI. Bill 1630 to amend the French IP Code opens a pioneering approach by: (i) recognizing the granting of copyright to AI-generated works even without direct human intervention; (ii) determining the copyright holder of AI-generated works in this case as the author or owner of the work used by AI; (iii) Granting the right to collect money and distribute royalties for AI-generated works to collective management organizations.

AI users often have a rather limited and passive role when they can only make requests or press a button but cannot control the creative process of AI. Typically, for the ChatGPT application, the output content of ChatGPT involves two subjects: ChatGPT users and ChatGPT. Although ChatGPT users do not directly perform the creative process, they are the ones who give "suggestions" by entering information into ChatGPT, thanks to which ChatGPT creates valuable content. If AI users simply "press a button" or make requests (commands) for the AI ​​system to operate, it is not enough to prove their control or creative choice over the AI ​​output product. In these cases, AI users cannot claim copyright. However, AI users can be recognized as authors if they play a role in directly intervening in the creation of the work, as in the following cases: (i) Users play the role of editors, selecting, editing, arranging materials from the AI ​​output to create to produce a final product that bears the user’s unique creative stamp; (ii) The user provides their proprietary data for AI to assist in creating the product. For example, a user writes an article and asks the Chat GPT tool to edit their article.

The creation of an AI can involve multiple individuals at different stages, but co-authorship only arises when individuals work together on a common plan so that the AI ​​output is the result of a “combined creative effort”. Where both the developer of an AI system and the user collaborate to create a work that meets the conditions for copyright protection, they can, in principle, become co-authors.

5. Commentary on certain provisions of Vietnamese law concerning copyright protection for AI-generated works and proposals for amendment

In order to promote research and application of AI to contribute to socio-economic development in Vietnam, the National Strategy on Research, Development and Application of Artificial Intelligence to 2030 sets the goal of "making this new technology an important technology field of Vietnam in the Fourth Industrial Revolution" in order to "contribute significantly to creating a breakthrough in production capacity, improving national competitiveness, promoting sustainable economic growth" [35]. Resolution No. 57-NQ/TW dated December 22, 2024 by the Politburo identifies artificial intelligence as "one of the strategic technologies" to promote digital transformation and digital economic growth in Vietnam. To realize these strategies, it is extremely important to build and perfect the legal framework regulating AI activities, including regulations on copyright protection.

Through international legal practice in handling issues related to AI and copyright, it can be seen that countries around the world tend to find a balance between the need to create incentives to promote innovation in the development and application of AI and protect the interests of related parties. The traditional view that considers the author as the center of the copyright system, only protecting copyright for humans can be changed in a flexible direction, accepting copyright protection for works created with the assistance of AI to ensure the interests of related parties, especially creating incentives for innovation and creativity, bringing benefits to the whole society. Currently, Vietnam's Intellectual Property Law has not had any amendments or guidelines on this issue. In order to have a legal basis for protecting intellectual property rights for works created by AI to encourage creativity, innovation, promote the development and application of AI technology, Vietnam intellectual property law also needs to have appropriate amendments or guidelines.

- Regarding the conditions for protecting works

The provisions in Clause 7, Article 4 of the Vietnam Intellectual Property Law on the conditions for protecting works are quite similar to international law: “A work is a creative product in the field of literature, art, or science expressed in any means or form”. The condition on the creativity of a work is stipulated in Clause 3, Article 14, according to which, a protected work “must be directly created by the author through his/her intellectual labor without copying from the work of another person”. The conditions for creativity in Vietnam law emphasize two factors: (i) the work is the result of the author's intellectual labor, that is, the work must be the result of creative spiritual activities, the result of the process of thinking, exploring, creating, reflecting the thoughts and feelings of the author; (ii) it is an independent creation, not copied from the work of others. This factor requires that the work is created by the author himself, and can be distinguished from the works of others. Compared with the requirements for creativity of works in Vietnam law and the laws of some countries, Vietnam law does not require a "minimum level of creativity" or the uniqueness or creative mark of the author. This shows that Vietnam IP law approaches the condition of original creativity of a work according to the “Sweat of the brow theory”, emphasizing the requirement that the work is the result of the author’s actual creative efforts and investment of effort and finance rather than requiring the work’s uniqueness. This approach will make it easier to recognize copyright protection for AI-generated works.

However, like countries following the continental European system, Vietnam law considers the author as the center of the copyright system, so copyright can only be reserved for human creators. Therefore, works created by AI can only be considered for copyright protection if there is human creative participation in the process of creating the work, or in other words, AI only plays a supporting role for humans in the process of creating the work. The “direct creation” element in the provisions of Clause 3, Article 4 of the current IP Law may lead to difficulties in applying to determine the protection conditions for works created by AI. For normal works, the person performing the creative act is the person who directly expresses his thoughts and feelings in the form of independent creation. For works created by AI, AI is the subject that helps the author express his creative ideas; in other words, AI is a tool that directly creates the physical manifestation of ideas under human control.

In order to have a basis for applying copyright protection to AI-assisted works, the author recommends: First, the word “directly” should be removed in Clause 3, Article 14 of the Law on Intellectual Property, accordingly, this provision should be amended as follows: “Works protected under the provisions of Clauses 1 and 2 of this Article must be created by the author through his/her intellectual labor without copying from the work of others”. The phrase “created by the author” is enough to assess the condition of originality of copyrighted works; Second, issue separate guidelines on copyright protection conditions for works created with the support of AI technology, based on the experience of the United States Copyright Office (USCO), specifically: (i) Works created by AI are only copyright protected if there is human creative intervention, expressed in the selection, arrangement of input materials, or editing and changing AI-generated documents to the extent that it meets the standards of original creativity. For example: Users select and build datasets to train AI to create text or images; or users edit, modify, and add new content to AI-generated works. (ii) For works with AI-generated content, copyright only protects human creative elements, independent and unrelated to the AI-generated content. For example: From images created by AI, users arrange the layout, write content to create a comic book, then users are only protected by copyright for the content they create.

- Regarding the subject of copyright protection

Vietnam IP Law recognizes two subjects of copyright protection: the author - the person with personal (moral) rights to the work and the copyright owner - the individual or organization holding the right to publish and property rights to exploit the work. According to Clause 1, Article 12a of the Vietnam IP Law, "the author is the person who directly creates the work". "Direct creation" is understood as the author who conveys his creative ideas to the physical world in a form that others can recognize. The emphasis on the element of “direct creation” in this definition is to distinguish the author - the person who creates the work from those who indirectly support creative activities such as those who provide information, give advice, and help the author create. Clause 2, Article 12a of the Vietnam IP Law stipulates as follows: “A person who supports, gives advice or provides materials for others to create a work is not recognized as an author”. Normally, when creating a work, the author is the one who chooses the materials and tools to express his/her creative ideas. For works created by humans, the author expresses his/her thoughts, concepts, ideas, and feelings through the way the work is expressed. As for works created by artificial intelligence, programmers will be the ones who choose and provide input data to the computer, and will be the ones who use and operate the computer as a tool to express the work. In this case, the programmer creates a computer brain to replace humans in learning, exploiting input data sources, self-selecting, processing information and self-deciding how to create new works.

The current provisions on authors and co-authors in the Vietnam IP Law are difficult to apply to recognize copyright for subjects such as AI creators, data providers for training, coaching or AI users because they do not directly create works, although through AI tools, they can create works according to their desired ideas and ways of expression, and the works created by AI are expressions of their thoughts, emotions and outlook on life. For works created by AI, AI algorithms or input data often play a decisive role in transforming those materials into creative works. AI only plays the role of a production tool to turn ideas and materials into works under human direction, guidance and control. In this case, the author's rights can be granted to AI creators, data providers for training, coaching or AI users. Copyright laws in some countries define an author quite simply as “the person who creates the work”, without mentioning the element of “directly” or “indirectly” creating the work [36]. To create a legal basis for copyright protection for works involving AI and to conform to international trends, the author proposes: First, the definition of author and co-author in Clause 1, Article 12a of the Vietnam Intellectual Property Law should remove the word “directly”, accordingly, the law can be amended as follows: “An author is the person who creates the work. In the case where two or more people jointly create a work with the intention that their contributions are combined into a complete whole, those people are co-authors”. Second, supplement the regulations guiding the determination of authors and co-authors in cases where works are created with the support of AI: “The person who plays a decisive creative role in the output of AI is recognized as the author/co-author”. The author/co-author can be the creator, programmer of the AI ​​system, creator of algorithms to direct, guide, and control the process of AI creating the work. The author/co-author can also be the person who provides input data to train AI, makes edits to AI works to create an original creative work. In cases where these subjects have creative contributions to the AI ​​creating a joint work, they can be recognized as co-authors. For example, the creator of the AI ​​system can play a creative role in conceptualizing, designing, providing data, editing, and selecting output for the work. AI users can also contribute creatively to the creation of a work when they provide their proprietary data for the AI ​​to create the work, or edit, select, modify, or arrange raw materials from the AI ​​output to create the final product.

The author of an AI-generated work can also be the copyright owner if they are the one who invested financially or provided the facilities to create the work [37]. In cases where the author creates a work as part of an administrative task or under a contract, the employer or the person who hires the creative author is the copyright owner [38]. For works created with the support of AI technology, the owner of the AI ​​system, the owner of the data system used to train the AI, or the investor in the AI ​​can be the copyright owner of the created work.

Until now, in cases where AI creates works relatively independently, without proof of significant human creative intervention, countries around the world do not recognize AI as the author, nor do they have a clear and definitive view on copyright protection for these works. Similar to countries following the continental European school, copyright protection in Vietnam is only established when there is a human author who creates the work. According to this traditional approach, if a work created by AI does not have clear human creative intervention, the author cannot be identified, and copyright ownership is not recognized. Recognizing AI as the author is not consistent with the principles of intellectual property protection. However, if the work created by AI is not protected by copyright, it will become public property that all subjects in society can freely use. Investors and developers of AI are unlikely to reap material benefits to compensate for the efforts and costs they have spent, nor do they have the financial resources to continue investing in and developing AI systems. In fact, when AI creates a work, that work is a “creative product” created by many people and in a chain that makes it difficult to distinguish the creative efforts of each individual, and the final product is the result of the organizational efforts of a legal entity, not the mark of individual individuals. Therefore, the idea of ​​establishing a type of sui generis right to provide limited exclusivity to investors who create AI systems to protect investment activities and encourage research and development of AI technology is worth considering. The French Bill No. 1630, which proposes to supplement Article L321-2 of the IP Code on copyright protection in cases where AI creates a work independently without direct human intervention, is considered bold and pioneering. The law determines that copyright will belong to the copyright holders of the input data used by AI; at the same time, it gives the right to manage the rights to AI-generated works (including the right to collect money and distribute rights from AI-generated works) to copyright organizations or other collective management bodies [39]. The idea of ​​creating a separate protection mechanism for works created by AI without having to prove the author's personal spiritual creativity is worth considering. In this case, there is no need to determine who is the author but only consider granting the property rights under copyright to the investor to create that work. The newly passed Law on Digital Technology Industry [40] in Vietnam specifically stipulates Article IV on “Artificial Intelligence”, in which Article 45 stipulates as follows: “Responsibilities of entities in the activities of developing, providing, and deploying the use of artificial intelligence systems”. In the author’s opinion, in addition to responsibilities, the law also needs to recognize the rights of these entities to ensure sustainable development and application of artificial intelligence. Therefore, it is possible to study and supplement separate rights protection mechanism for entities developing, providing, and deploying the use of artificial intelligence to create works.

- Regarding copyright registration

Currently the Vietnam IP Law does not have specific guidelines on copyright registration for works AI-assisted works. In fact, many countries in the world have added regulations on labeling AI or annotating “work generated by AI” to ensure transparency when using AI. In that trend, Vietnam has also stipulated in Clause 2, Article 44 of the Law on Digital Technology Industry: “Digital technology works in the List of digital technology works created by artificial intelligence must have identification marks for users or machines to recognize”. However, for more transparency and clarity in copyright protection as well as identifying the author and copyright owner of the work, the author recommends: First, in addition to the mandatory annotation of works created by AI or with AI creativity, it is necessary to include an annotation of the author and the original work that is the source of data for creating the AI ​​work (if any). Second, supplementing guidance for the provisions in Point a, Clause 2, Article 50 of the Law on Intellectual Property on copyright registration dossiers: In the copyright registration declaration, if the applicant uses AI in the process of creating the work, the applicant is responsible for providing specific information on the participation of AI and the creative part of AI in the work. If the applicant does not provide or provides incomplete or untruthful information, this is one of the grounds for revoking the Copyright Registration Certificate according to Article 55 of the Vietnam IP Law.

6. Conclusion

The rapid development of AI technology as well as the increasingly popular application of AI in many areas of social life is having a strong impact on many areas of law, including copyright law. Research on international legal practice shows that countries today tend to accept copyright protection for creative works with the participation of AI support, but there are almost no specific regulations to assess the standard of "creativity" for these works as well as determine which subjects to grant copyright to. For works purely created by AI, the issue of whether to grant a "limited monopoly" to the developer or provide data to AI to protect the rights of these subjects is still an open question that needs further research. Research on international trends in copyright protection for works created by AI shows that the core issue that countries around the world are concerned about is determining the balance between the need to create incentives to promote innovation in the development and application of AI and protecting the interests of stakeholders. In that trend, Vietnam IP Law also needs to have revisions and additional guidelines to have a legal basis for copyright protection for works created by AI to encourage creativity, innovation, promote the development and application of AI technology in Vietnam.

Reference material

1. Carys Craig, (2022), The AI-Copyright Challenge: Tech-Neutrality, Authorship, and the Public Interest, Osgoode Hall Law School of York University (2022)

2. Jozefien Vanherpe (2022), “AI and IP – Great Expectation”, (2022) https://www.researchgate.net/publication/356811479_AI_and_IP_-_Great_Expectations

3. Mohd Akhter Ali & M. Kamraju (2023), Osmania University Journal of IPR [OUJIPR] Vol.1 | Issue 1 July 2023, “Impact of Artificial Intelligence on Intellectual Property”, Osmania University Journal of IPR [OUJIPR] Vol.1 | Issue 1 July (2023) https://www.researchgate.net/publication/376751087_Impact_of_Artificial_Intelligence_on_Intellectual_Property_Rights_Challenges_and_Opportunities

4. JOHN LOCKS, KHẢO LUẬN THỨ HAI VỀ CHÍNH QUYỀN (LÊ HUY TUẤN DỊCH), NXB TRI THỨC (, 2013)

5. Andres Guadamuz, (2017) “Artificial intelligence and copyright” (2017), https://www.wipo.int/wipo_magazine/en/2017/05/article_0003.html

6. P. Bernt Hugenholtz, Joa˜o Pedro Quintais, (2021),“Copyright and Artificial Creation: Does EU Copyright Law Protect AI-Assisted Output?” (2021) https://www.researchgate.net/publication/355099987_Copyright_and_Artificial_Creation_Does_EU_Copyright_Law_Protect_AI-Assisted_Output

7. Desktop Marketing Systems Pty Ltd v Telstra Corporation Limited [2002] FCAFC 112 (15 May 2002), https://www.judgments.fedcourt.gov.au/ judgments/Judgments/fca/full/2010/2010fcafc0149,

8. Telstra Corporation Limited v Phone Directories Company Pty Ltd [2010], https://www.judgments.fedcourt.gov.au/judgments/Judgments/fca/full/2010/2010fcafc0149,

9. FederalCourt of Australia - Full Court, (2012), Acohs Pty Ltd v Ucorp Pty Ltd [2012] FCAFC 16 (2 March 2012), http://www6.austlii.edu.au/cgi-bin/viewdoc/au/cases/cth/FCAFC/2012/16.html

10. U.S. Copyright Office Practices (2021), The Compendium: Copyrightable Authorship: What Can Be Registered (bổ sung lần thứ ba), https://www.copyright.gov/comp3/chap300/ch300-copyrightable-authorship.pdf,

11. US Supreme Court (1991), Feist Publications, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340 (1991), https://supreme.justia.com/cases/federal/us/499/340/

12. United States Copyright Office, Zarya of the Dawn (Registration # VAu001480196), 21/12/2023, https://www.copyright.gov/docs/zarya-of-the-dawn.pdf

13. US Copyright Office (2023), ‘Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence’ 37 CFR Part 202 (16 March 2023), https://www.copyright.gov/ai/ai_policy_guidance.pdf

14. Scott Warren & Joseph Grasser (2024), “Japan’s New Draft Guidelines on AI and Copyright: Is It Really OK to Train AI Using Pirated Materials?” https://www.privacyworld.blog/2024/03/japans-new-draft-guidelines-on-ai-and-copyright-is-it-really-ok-to-train-ai-using-pirated-materials/

15. Copyright, Designs and Patents Act 1988 _CDPA https://www.legislation.gov.uk/ukpga/1988/48/section/178,.

16. Ant Horn, "Creators and the copyright balance: Investigating the interests of copyright holders, users and creators" (2004); Alternative Law Journal 112 (2004) https://www.researchgate.net/publication/318532855_Creators_and_the_Copyright_Balance_Investigating_the_Interests_of_Copyright_Holders_Users_and_Creators

17. Olivier and Ines Banazzouz, https://www.mondaq.com/copyright/1415046/french-national-assembly-proposes-new-law-to-secure-copyright-in-ai

18. Decision No. 127/QĐ-TTg of the Prime Minister dated 26 January 2021 on the “National Strategy on Research, Development and Application of Artificial Intelligence until 2030.”

[*] Assoc. Prof. Dr. Vu Thi Hai Yen, Vice Dean, Faculty of Civil Law, Hanoi Law University

[**] Le Vu Minh Duc, MA, Minervas Law Office

[1] This article is conducted within the framework of the university-level scientific research project “Improving intellectual property law under the impact of artificial intelligence in the context of scientific and technological development, innovation, and digital transformation in Vietnam.”

[2] Carys Craig, The AI-Copyright Challenge: Tech-neutrality, Authorship, and the Public Interest, Osgoode Hall Law School of York University, (2022)

[3] Jozefien Vanherpe (2022), AI and IP – Great Expectation (2022), https://www.researchgate.net/publication/ 356811479_ AI_and_IP_-_Great_Expectations,(truy cập lần cuối 22/07/2025)

[4] Mohd Akhter Ali & M. Kamraju, Mohd Akhter Ali & M. Kamraju, Impact of Artificial Intelligence on Intellectual Property, Osmania University Journal of IPR [OUJIPR] Vol.1 | Issue 1 July 2023, https://www.researchgate.net/publication/ 376751087_Impact_of_Artificial_Intelligence_ on_Intellectual_ Property_Rights_Challenges_and_Opportunities

[5] Mohd Akhter Ali & M. Kamraju, ibid, 4

[6] Carys Craig, (2022), ibid, 2

[7] John Locke, Two Treatises of Government, translated by Le Huy Tuan, Tri Thuc Publishing House, 2013.

[8] Jozefien Vanherpe, ibid, 3

[9] Andres Guadamuz “Artificial intelligence and copyright”, https://www.wipo.int/wipo_magazine/en/2017/05/ article_0003.html, (2017), truy cập lần cuối 22/07/2025

[10] P. Bernt Hugenholtz, Joa˜o Pedro Quintais, (2021), Copyright and Artificial Creation: Does EU Copyright Law Protect AI-Assisted Output?, (2021) https://www.researchgate.net/publication/355099987_Copyright_and_ Artificial_Creation_Does_EU_Copyright_Law_Protect_AI-Assisted_Output, (truy cập lần cuối 22/07/2025)

[11] P. Bernt Hugenholtz, Joa˜o Pedro Quintais, ibid, 10

[12] P. Bernt Hugenholtz, Joa˜o Pedro Quintais, ibid, 10

[13] Andres Guadamuz, ibid, 9

[14] Desktop Marketing Systems Pty Ltd v Telstra Corporation Limited [2002] FCAFC 112 (15 May 2002), https://www.judgments.fedcourt.gov.au/judgments/Judgments/fca/full/2010/2010fcafc0149, ( last visited July 22, 2025 )

[15] Telstra Corporation Limited v Phone Directories Company Pty Ltd [2010], https://www.judgments.fedcourt. gov.au/judgments/Judgments/fca/full/2010/2010fcafc0149, ( last visited July 22, 2025 )

[16] FederalCourt of Australia - Full Court, (2012), Acohs Pty Ltd v Ucorp Pty Ltd [2012] FCAFC 16 (2 March 2012), http://www6.austlii.edu.au/cgi-bin/viewdoc/au/cases/cth/FCAFC/2012/16.html, ( last visited July 22, 2025 )

[17] P. Bernt Hugenholtz, Joa˜o Pedro Quintais, ibid, 10

[18] P. Bernt Hugenholtz, Joa˜o Pedro Quintais, ibid, 10

[19] P. Bernt Hugenholtz, Joa˜o Pedro Quintais, ibid,10

[20] U.S. Copyright Office Practices (2021), The Compendium: Copyrightable Authorship: What Can Be Registered ( 3d ed. ), https://www.copyright.gov/comp3/chap300/ch300-copyrightable-authorship.pdf, ( last visited July 22, 2025 )

[21] US Supreme Court (1991), Feist Publications, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340 (1991), https://supreme.justia.com/cases/federal/us/499/340/

[22] United States Copyright Office, Zarya of the Dawn (Registration # VAu001480196), 21/12/2023 https://www.copyright.gov/docs/zarya-of-the-dawn.pdf, ( last visited July 22, 2025 )

[23] US Copyright Office (2023), ‘Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence’ 37 CFR Part 202 (16 March 2023), https://www.copyright.gov/ai/ai_policy_guidance.pdf

[24]Scott Warren & Joseph Grasser (2024), “Japan’s New Draft Guidelines on AI and Copyright: Is It Really OK to Train AI Using Pirated Materials?” (2024) https://www.privacyworld.blog/2024/03/japans-new-draft-guidelines-on-ai-and-copyright-is-it-really-ok-to-train-ai-using-pirated-materials/,( last visited July 22, 2025 )

[25] Copyright, Designs and Patents Act 1988 _CDPA, 9.3. “In the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken; The Copyright and Related Rights Act 2000 of Ireland defines ‘‘computer-generated’’, in relation to a work, as meaning ‘‘that the work is generated by computer in circumstances where the author of the work is not an individual”;

[26] Ant Horn, Creators and the copyright balance: Investigating the interests of copyright holders, users and creators, Alternative Law Journal 112 (2004), https://www.researchgate.net/publication/318532855_Creators_and_ the_Copyright_Balance_Investigating_the_Interests_of_Copyright_Holders_Users_and_Creators

[27] P.Bernt Hugenholtz, Joa˜o Pedro Quintais, ibid, 10

[28] Copyright, Designs and Patents Act 1988 _CDPA, https://www.legislation.gov.uk/ukpga/1988/48/section/178,.

[29] P. Bernt Hugenholtz, Joa˜o Pedro Quintais, ibid, 10

[30] P. Bernt Hugenholtz, Joa˜o Pedro Quintais, ibid, 10

[31] P. Bernt Hugenholtz, Joa˜o Pedro Quintais, ibid, 10

[32] Jozefien Vanherpe (2022), ibid, 3

[33] Jozefien Vanherpe (2022), ibid, 3

[34] Olivier and Ines Banazzouz, https://www.mondaq.com/copyright/1415046/french-national-assembly-proposes-new-law-to-secure-copyright-in-ai, (truy cập lần cuối 22/07/2025)

[35] Decision No. 127/QĐ-TTg of the Prime Minister dated 26 January 2021 on the “National Strategy on Research, Development and Application of Artificial Intelligence until 2030.”

[36] See U.S. Copyright Law, 17 U.S.C. § 201(a); Copyright Act of Japan, art. 2(1)(ii); Copyright, Designs and Patents Act 1988 (UK), § 9(1); Urheberrechtsgesetz (Ger.), § 7; Code de la propriété intellectuelle (Fr.), art. L111-1.

[37] Vietnam, Law on Intellectual Property, No. 50/2005/QH11, as amended, art. 37, 38.

[38] Vietnam, Law on Intellectual Property, No. 50/2005/QH11, as amended, art. 39.

[39] Olivier and Ines Banazzouz, https://www.mondaq.com/copyright/1415046/french-national-assembly-proposes-new-law-to-secure-copyright-in-ai? ( last visited July 22, 2025 )

[40] Law No. 71/2025/QH15 on Digital Technology Industry (adopted by the 15th National Assembly on June 14, 2025).