This is a bilingual snapshot page saved by the user at 2024-7-8 18:02 for, provided with bilingual support by Immersive Translate. Learn how to save?

Research on the copyrightability and ownership of AI-generated works

Wang Yanlin*

Abstract:In recent years, artificial intelligence has been constantly updated and iterated, bringing convenience to human beings and creating many problems and challenges. After the emergence of ChatGPT, it has set off a huge wave in the field of artificial intelligence technology around the world, and its generated content (AIGC) has also been used by the public, but there is a question that needs to be answered urgently is the copyright issue related to artificial intelligence generated content (AIGC), and the current problems mostly lie in: whether artificial intelligence can be used as a civil subject, whether artificial intelligence-generated content is a work under the copyright law, and the ownership of rights of artificial intelligence generated objects (AIGC).

Keywords: AI-generated, copyrightability, ownership of rights



The entry of artificial intelligence into the field of literary "creation" has become an irreversible trend, which has had an impact on the current legal system, especially in the field of copyright law. China's copyright law is constructed with natural person authors as the core, and protects their interests by giving natural person authors monopoly rights, thereby promoting the prosperity of the cultural field. Nowadays, the neural networks of artificial intelligence have reached the level of mimicking human thinking, and artificial intelligence is gradually moving away from human dependence, and in this case, artificial intelligence products have raised a series of problems. In terms of the subject matter of copyright, the work is the object of copyright protection, so does the AI-generated object have the constituent elements of the work? In terms of copyright subjects, who enjoys the relevant rights and interests of AI-generated products? If left unchecked, it will become an orphan work, which will undoubtedly discourage the enthusiasm of relevant subjects and is not conducive to the prosperity and stability of the cultural market. The United Kingdom, the United States, the European Union and other countries have either clarified the relationship between artificial intelligence and copyright through laws or included in the scope of discussion of the bill, but China is still blank in terms of legal policies on the above issues. In this article, we will discuss and study related issues.

1. Discussion on the subject qualification of artificial intelligence

Since the advent of artificial intelligence, the debate over its legal qualifications has been extremely intense. Since then, there has been a lot of discussion about the qualifications of AI copyright subjects, and it is difficult to directly conclude this issue in the field of intellectual property rights and even in the entire judicial field. In view of the issue of the legal personality of artificial intelligence copyright, the mainstream view of the current legal theory community is that the personality in civil law has always shown an expansive trend of change and evolution since the continuation of ancient Roman law, and it is the trend of the development of civil law personality theory to break through the limitations of anthropocentric thinking and recognize the legal personality of artificial intelligence.

Of course, there are different views in the academic community on whether AI copyright has the status of subject qualification.

(1) Negative theory of the qualifications of artificial intelligence subjects

From a philosophical point of view, the essence of man is the sum total of all social relations. Artificial intelligence is not a "human" but a tool of human beings, and of course it cannot be used as the subject of copyright, so it certainly does not have civil qualifications. Artificial intelligence is not a "person", but a thing. Weak artificial intelligence is a program that deals with specific problems in a certain field, but fundamentally speaking, it is essentially different from natural persons, not "people", but "things". Although strong artificial intelligence has reached the level of human intelligence, capable of thinking, planning, solving problems, abstract thinking, understanding complex ideas, rapid learning and learning from experience, but human characteristics such as 'human virtues', 'ability to nourish and feel', and 'free and conscious labor' are not possessed, for example, the little boy David in the movie "Artificial Intelligence" was sent to the repair shop because he ate, so they are not "people", but "things". No matter how advanced AI is at its advanced stage, its nature as a human tool will not change 1 .

From the point of view of legislative purposes, the higher the quality and attention of the work, the greater the demand for others to use the work, the more the author will benefit from it, and his reputation will grow day by day. Under such an incentive mechanism, more people will be willing to devote themselves to creative activities, thus promoting the development and prosperity of socialist culture and science. Obviously, animals cannot be motivated by copyright law, only people can understand copyright law and be incentivized, so only human creations can be protected by copyright law as works 2 .

Because artificial intelligence does not have the instinctive desire of human beings to interact with the outside world to maximize its own interests, artificial intelligence does not have the ability to express meaning and effect, and cannot participate in civil activities in its own name, so it does not have the ability to be independent and does not enjoy civil capacity. At the current stage, it can only be a thing, and cannot be a subject of copyright.

(2) The theory of pseudo-personality

The "pseudo-personality theory" believes that artificial intelligence has the ability to think and is out of the traditional category of "things", but the essential tool attributes of artificial intelligence have not changed. Therefore, artificial intelligence can be given certain rights to become a fictitious civil subject. For this kind of artificial intelligence creation, it is not the result of direct creation by humans, and the creative subject in the de facto sense is artificial intelligence, so it is difficult to directly apply the traditional natural person author theory to solve the problem. If the result of such a creation is to be protected by copyright, one preferred option is to take the approach of "pseudo-producer", i.e., to legally create an author and rights holder for the AI work 3 . The provisions of the existing law on civil subjects do not limit civil subjects to natural persons, and with the development of society, it is possible for inanimate subjects to become civil subjects, and legal persons becoming civil subjects is the best example 4 .

At present, artificial intelligence is highly autonomous and has a tendency to constantly break away from human control, and it seems that artificial intelligence with "spirit and consciousness" should be given the status of legal subject has become a practical problem. When "spirit and consciousness" are no longer exclusive to human beings, it seems that the "insurmountable gap between subject and object" will be shaken, and it is also an irreversible trend

To sum up, there is no doubt that artificial intelligence cannot be a "human", and it is impossible to have the same legal rights and obligations as human beings, but the emergence of civil subjects is historical, and not all natural persons in a slave society can be civil subjects, only slave owners are civil subjects, and slaves can only be property. The "legal person" did not exist at the beginning, and it was only through legal fiction that it became a civil subject and then engaged in civil activities. The rights of slaves were constantly fought for through revolution, and the qualification of "legal persons" to obtain subjects was given by human beings with the development of social economy. Therefore, whether artificial intelligence can be used as a civil subject still needs to wait for the further development of the times, if artificial intelligence can really develop to think independently and fight for its own rights; Or in order to promote the needs of science and technology, culture, and economic development, when the time comes to give artificial intelligence subject qualifications, the scope of civil subjects will inevitably be further expanded.

2. Whether AI-generated content is a work under the Copyright Law

In the Film case, the court held that AI-generated works were not works under copyright law, but in the Baidu case, the court recognized the nature of works that used Dreamwriter to generate articles. Both courts have no dispute that AI-generated artifacts belong to the field of literature, art and science and are reproducible, but there are disputes about whether the requirements for intellectual achievements are met and whether the difference in the creative process affects the recognition of the work. As for the intellectual achievement element, the defendant in the film case argued that the article generated solely by the combined action of data, algorithms, and templates was not intellectual, but the Beijing Internet Court did not give a positive answer to this question in its judgment. In the Tencent case, the Nanshan District Court of Shenzhen held that the plaintiff had paid substantial labor and that the articles in question contained intellectual attributes. Both courts affirmed whether "person" is a necessary element, but the angle of judgment of "person" is different. The Beijing Internet Court denied the qualifications of the articles involved in the case on the grounds that there was no human participation in the birth stage of the creation, and did not consider the work of the human beings in setting the data, creating the template, and opening the artificial intelligence software before the generation, and the labor involved in the selection of the creation after the creation. On the contrary, the Nanshan District Court of Shenzhen held that both the selection of the original material and the final modification process were contributed by human beings, so that the creation of the product could be considered to be created by humans. Of course, more people now believe that with the advancement of technology, many new forms of creation will emerge, and that the subjective standard of "anthropocentric" is no longer in line with the trend of the times, and that objective standards should be constructed, and non-human creations can also be called the object of copyright protection. It can be seen that the determination of the copyrightability of AI-generated works is a very controversial issue.

First, the judgment of the work should be centered on the human reader rather than the human author, that is, whether the work belongs to the work has nothing to do with whether the work comes from humans or artificial intelligence, and "the reader is the interpreter of the work". Second, the judgment of originality should be changed from a subjective criterion to an objective criterion, and the product is no different from a traditional work in appearance, and it is a work protected by copyright.

The negative argument mainly argues the reasons for non-copyrightability from the following three perspectives: first, from the perspective of the creative subject, it is believed that human beings are still a necessary element of creation, and artificial intelligence still belongs to the category of computers in the final analysis, so its products do not belong to works under the Copyright Law; Second, from the perspective of the creative process of the product, it is believed that the birth of the product is only the result of the use of established algorithms and pre-rules, and the final product is only a single or limited number, without the embodiment of individual intelligence, and does not have originality. Third, from the perspective of the thoughts and emotions of AI-generated objects, it is believed that AI-generated objects do not contain personality factors such as concepts and emotions, and cannot achieve the goal of stimulating creation through the granting of copyright. Comparing the two completely opposite theories, the affirmative and negative theories, the focus of the dispute between the two sides is the inconsistent understanding of the concept of "work". The negative theory insists that human participation is needed in the work, and the work should contain human thoughts and emotions; Affirmation focuses on judging the work from an objective point of view, and downplays subjective factors such as personality in the work. Therefore, clarifying the concept of "work" has become the key to judging whether an AI-generated work is copyrightable.

Article 3 of the Copyright Law stipulates that "the works referred to in this Law refer to intellectual achievements in the fields of literature, art and science that are original and can be expressed in a certain form". Therefore, in order to constitute a work protected by China's Copyright Law, three conditions need to be met: first, it must belong to a work in the fields of literature, art and science, which is a scope requirement; the second is to reflect the author's unique intellectual judgment and choice, which is a requirement of originality; Third, it must be able to be expressed in a certain form 5 . At this stage, AI generated objects are limited to the fields of writing, art, and music, which of course belongs to the "fields of literature, art, and science". "Can be expressed in a certain form" is also called "reproducibility", which actually means that the product can be reproduced, regardless of whether it is actually copied or not. At this stage, a series of products such as AIVA's music album and Xiaoice's "Sunshine Lost the Glass Window" need to be expressed by certain carriers, such as hard disks, paper, networks, etc., at this time, they have the possibility of being contacted and copied. Therefore, whether the rest is original or not is the key to whether the AI-generated works are copyrightable.

The most important thing in judging whether an AI "creation" is a work protected by the Copyright Law is to judge the "originality" in the object of the Copyright Law. In fact, in general, we don't have high requirements for "originality". When the court hears a case, as long as the external form of expression of the work has objectively identifiable differences, and can reflect the author's choice and arrangement to a certain extent, it can be regarded as an original work 6 .

The objective criterion for judging originality is to abandon the thought, personality, personality and other factors emphasized by the early "Romantic view of works", and instead only seek objective criteria, that is, "originality" and "creation". "Independence" requires that the work is completed by the creator mentally independently. Under the objective criterion, "creation" should be based on the "minimum inventive step" established by the United States, and the form of expression should be distinguished from the works of others, and the "minimum inventive step" 7 should be satisfied by excluding the customary expression in the public domain and the plagiarism of existing works or works beyond the term of protection.

Under the objective criterion of originality, artificial intelligence relies on artificial neural network systems to perform deep learning on big data, so that the product can be distinguished from other works, and readers can feel that this is new content, which meets the requirements of copyright law for the originality of works.

3. Ownership of rights in artificial intelligence generated objects (AIGC).

The issue of the ownership of rights generated by artificial intelligence is an important issue in the study of "artificial intelligence and copyright", and the ownership of rights is the division of interests, and it is also for better protection and rules for AIGC. China's Copyright Law stipulates that natural persons, legal persons and unincorporated organizations may be copyright owners. Whether it is a legal person or an unincorporated organization, it is a collection of natural persons and embodies the will of natural persons. Artificial intelligence technology itself is not a qualified subject of rights, it does not have its own independent property, and cannot be held liable externally. In judging the ownership of the copyright of AI-generated works, it should be based on human behavior.

There are three main views on this issue in the domestic academic circles: belonging to artificial intelligence; attribution to a natural person; into the public domain. The prevailing view is that it is attributed to natural persons, primarily those who have contributed to the labor of AI or AI-generated objects, including designers, users, and owners.

The first view is that it belongs to artificial intelligence. As an emotionless machine, artificial intelligence does not need to pursue the "spiritual interest" of signature; Investors tend to pursue economic interests and do not care too much about the authorship of AI "creations"; However, based on the right to know, the public has the right to know who is the real creator of the work, and out of the principle of good faith, the copyright owner of the AI "creation" should also be required to bear the "obligation of authorship" and require it to sign the name of the real creator of the AI on the AI "creation". If the AI does not have a specific name, it should also be stated in the author column that the work was created by the AI 8 . The second view is attribution to natural persons. This view does not recognize that AI products belong to the AI itself, but rather to the creator, user, or investor of the AI. Clarifying the ownership of copyright among various subjects is a prerequisite for promoting the orderly use of generative AI works. Copyright ownership should be based on the three basic principles of substantive contribution, investment incentives and balance of interests, based on the important influence of each subject in the process of generating copyrighted works, and pay full attention to the possibility 9 of establishing cooperative works 。 The third view is attribution to the public domain. It is important to face up to and attach importance to the "intellectual public domain", which provides a frame of reference for coexistence with copyright, not only can copyright owners participate in it through consensus agreements, but also that certain information, although it has the appearance of a work, does not necessarily become the object of copyright, and perhaps remaining in the "intellectual public domain" will be more conducive to the development of human culture, art and science in the information age, and there is no need for copyright protection to stimulate creative enthusiasm 10 。 However, Article 1 of China's Copyright Law also clearly stipulates that it should "promote the development and prosperity of socialist culture and science". Therefore, if the rights are attributed to the public, the incentives for the originality of AI-generated products will be reduced, and the problem of ownership of AI-generated products cannot be solved in the long run.

With regard to the second view of attribution to natural persons. In the process of generating artificial intelligence products, many subjects have made some contributions in different forms, which has led to endless debates in the academic community about the ownership of copyright.

Scholars who advocate that AI designers become copyright subjects believe that the process of the birth of the created object is inseparable from the designer's substantive participation in AI itself. Because no matter how intelligent an AI is, it is inseparable from the programmer's programming. The developer of artificial intelligence is the key to the production of artificial intelligence, the developer uses algorithms to write the system software of artificial intelligence, this kind of system software creates works through algorithm programs, and the form of artificial intelligence output works depends to a certain extent on the setting 11 of algorithm parameters 。 Without the designer's act of inventing and creating artificial intelligence, the product is a source of water and cannot be talked about. The designer's investment and dedication are more creative and substantive, so it is fully reasonable to give the copyright to the designer. There are some problems with the above view. Although the designer has designed the operating system of artificial intelligence in advance, the contribution of the designer is limited to this. Without the blessing of raw materials and big data, AI cannot generate any works. The data and materials that are the basis of artificial intelligence are constantly changing, and the creative expression of deep learning tends to be independent, and it is impossible for designers to preset all the works that can be generated by artificial intelligence, otherwise the original materials and new data will lose their meaning to artificial intelligence. The designer does not make a substantial and direct contribution to the creation of the final work, and cannot enjoy the copyright of the resulting work. In addition, the doctrine hides the following insurmountable drawbacks. First, there is the problem of double profit for AI designers. Artificial intelligence itself still belongs to the category of computer software, and the designer is the subject of the copyright rights of artificial intelligence itself; If the designer becomes the subject of the rights of artificial intelligence products, then the designer will obtain double benefits due to one of his intellectual labor, and there is a suspicion of monopoly. Secondly, the designer lacks creative awareness of the final birth of the product. Artificial intelligence is the final generator of the work, and the connection between the designer and the generated object has been severed by the creation of artificial intelligence. In other words, the AI designer did not put in the direct and substantive work on the generation of the final work. In addition, due to the asymmetry of information, it is difficult for anyone other than the user to know the existence of the work due to the use of artificial intelligence to generate the work, and it is not necessary to grant the copyright of the created work to the designer. Thirdly, the purpose of copyright to encourage creation cannot be realized. If the designer becomes the subject of rights and enjoys exclusive rights, then the user needs the designer's authorization when disseminating and reproducing the generated objects, which is undoubtedly absurd, and the owner and user of artificial intelligence at this time are most likely not to purchase and use artificial intelligence, and the whole society will not be able to enjoy the dividends of works brought by artificial intelligence, which is contrary to the copyright purpose of "encouraging the production and dissemination of works". Finally, the generated is not a derivative of artificial intelligence. For example, in the case of a camera, although the designer has designed the technology for imaging photographs, it is absurd for the designer to claim copyright in the work taken with the camera.

Scholars who advocate the copyright of AI owners believe that most of the AI owners are AI investors, and considering that the birth of AI products often requires huge monetary costs, and in order for the industry to develop in the long run, these investments need to obtain corresponding returns, so the copyright should be attributed to the owner based on the consideration of "protecting investment interests and promoting industrial development", and the AI owner is the right holder who enjoys the intellectual property rights of AI technology. Although AI owners may not have a significant contribution to the AI-generated results, based on the Nash equilibrium theory, this approach is the most effective way to motivate the market and promote the prospects of the entire chain of the AI industry 12 .

Scholars who advocate that the copyright of AI-generated products should belong to the use of AI believe that although the creation of AI relies on machine learning and big data technology, the core of operation is the preparatory work done by AI users using AI software or programs, such as setting programs and selecting keywords. Therefore, the user is the basis for the emergence of AI-generated results, the promoter of the necessary links, and the user presses the "start button" of creation. The user of artificial intelligence informs a large amount of information and issues corresponding instructions with the intention of outputting a specific work, and the artificial intelligence generated objects reflect the user's specific choices, judgments and analysis 13 。 It is more reasonable for the user to become the subject of rights. First of all, empowering users is more conducive to the balance of interests among many subjects. Artificial intelligence itself is a computer software work under the Copyright Law, and the designer and owner have obtained corresponding economic benefits. As mentioned above, AI designers can obtain corresponding economic benefits through copyrights (computer software) and patent rights (intelligent machines) for the AI they design. Artificial intelligence owners receive revenue through advertising booth fees, licensing fees. Therefore, since other entities have recovered the corresponding costs through other means, giving copyright to users can alleviate the costs paid by users in the process of dissemination of AI-generated works, maximize the value of the work, and achieve a balance between the interests of many subjects. In the artificial intelligence industry ecosystem, the user is located on the consumer side, and the interests of the user will be fully protected to activate the consumer side, and artificial intelligence R&D enterprises will earn more profits; The company's earnings will increase, and it will invest more R&D funds to promote the birth of more intelligent artificial intelligence; The improvement of artificial intelligence technology will attract more users to use it, so as to realize the healthy and circular development of the artificial intelligence industry 14 . Second, AI-generated products contain substantial contributions from users. The final style and type of the product is determined by the form of instructions and original materials input by the user, and these instructions and materials reflect the user's creative intention and contribution. These acts may seem simple, but the judgment of the intellectual labor performed by the user should not be judged by the complexity of the act, and the user's contribution should be treated equally by the copyright law. Moreover, after the creation of the product, the user needs to pick the most valuable, further polish it, and embellish it to improve the quality of the product, and the substantial contribution made in this process is not simple at all. Finally, giving it to users is conducive to the realization of the purpose of copyright law. Artificial intelligence users can be roughly divided into the general public and professional subjects, and the general public often uses artificial intelligence for creation and dissemination for reasons such as preferences and interests, such as users borrowing poetry creation software to generate poems for dissemination on the Internet, which helps to enrich the cultural and entertainment life of the public; Professional subjects make a living from the creation of works, and their input of raw materials is more delicate, and the quality of the generated works is often higher, which can obtain benefits and spiritual satisfaction, and at the same time promote the prosperity of society and culture.

  1. Xu Jiali, "Copyright Ownership of Artificial Intelligence Generated Objects", Jinan Journal, No. 4, 2023.

  2. Wang Qian: Is the content generated by ChatGPT protected by copyright law?, Exploration and Controversy, No. 3, 2023.

  3. Lv Bingbin, "Theoretical Reconstruction of the Proposed Producer of Copyright Law in the Era of Artificial Intelligence", Nanjing Social Sciences, No. 10, 2023.

  4. Li Weimin, "The Reconstruction of the Service Works System and the Selection of the Copyright Ownership Path of Artificial Intelligence Works", Law Review, No. 38, 2020.

  5. Dong Yadi: Analysis of the Elements and Attribution Rights of Artificial Intelligence Generated Works, Legal Expo, No. 8, 2024.

  6. Lin Xuebiao, Chen Yuanying: On the Copyright Ownership Dispute of Artificial Intelligence "Creations" and Countermeasures, Straits Law Science, No. 4, 2022.

  7. Yu Lei: Research on the copyright of AI-generated products.

  8. Lin Xuebiao, Chen Yuanying: On the Copyright Ownership Dispute of Artificial Intelligence "Creations" and Countermeasures, Straits Law Science, No. 4, 2022.

  9. Cong Lixian, Li Yonglin: Recognition and Copyright Ownership of Generative AI: A Case Study of ChatGPT's Application Scenarios, Journal of Shandong University (Philosophy and Social Science Edition), No. 4, 2023.

  10. Yu Wenwen: On the Ownership of the Rights and Interests of AI-Generated Content in Copyright Law, Journal of the University of Chinese Academy of Social Sciences, No. 2, 2022.

  11. Dang Xi, Wang Liqun: Research on Copyright Attribution of Artificial Intelligence Generated Objects, Journal of Zhejiang Sci-Tech University (Social Sciences Edition), No. 6, 2021.

  12. Wei Xingchen: Copyright Law protects AI-generated results.

  13. Ma Zhongfa, Xiao Yulu: Copyright Protection of Artificial Intelligence Creations, Electronic Intellectual Property, No. 6, 2019.

  14. Qiu Rungen, Cao Yuqing.: On the Copyright Protection of Artificial Intelligence "Creations", Journal of Nanchang University, No. 2, 2019.



Scientific and technological progress is an irreversible trend, and for a series of problems related to the products brought about by this technology, the copyright system should adopt an open and inclusive attitude, rely on the vitality and compatibility of the system, and actively explore appropriate systems, rather than sticking to the old ways and being afraid of new things so that they are excluded from the copyright system.

Artificial intelligence cannot be the subject of copyright. In other words, no matter what stage of development artificial intelligence reaches, as long as the subjectivity of human beings and the objectivity of artificial intelligence are not reversed, human beings can only be dominated as objects when making institutional arrangements. The theory of property rights and labor in the philosophy of intellectual property law, and the theory of personality are also based on people, and artificial intelligence does not need to be motivated. Copyright confers on both designers and owners some insurmountable drawbacks. Granting to users is more conducive to the balance of interests among many subjects, the healthy development of the artificial intelligence industry, and the realization of the purpose of copyright.