Generative AI such as ChatGPT is still in the era of weak AI, but its products are in line with the composition of a "work" both in form and substance. Judicial practice shows that the right holder of AI-generated objects can obtain civil remedies, but there are still value dilemmas and normative dilemmas in whether the criminal law should protect AI-generated objects. It is necessary to adopt the two-tier legal benefit theory to understand the crime of copyright infringement, avoid the occurrence of the "digital divide", and start from the perspective of legal benefit coverage, argue that ChatGPT and other generative AI works are the targets of copyright infringement, and make a detailed understanding of the crime of copyright infringement. AI-generated works are not an independent type of work, and the criminal law can be expanded to include AI-created works. Based on this, the criteria for what constitutes a work of AI-generated objects should be strictly defined. Taking into account the degree of social danger of infringing on AI-generated objects, AI-generated objects are accorded less protection than natural persons.
introduction
Although we are still in the era of weak artificial intelligence, the emergence and iteration speed of generative AI such as ChatGPT is surprising. On November 30, 2022, ChatGPT, a chatbot program developed by OpenAI in the United States, was officially released, and within 3 months of its release, this new generation of generative AI program gained hundreds of millions of users around the world, becoming the program with the largest number of users and the highest attention in history. ChatGPT is at the forefront of the development of artificial intelligence technology, it can understand and learn human language and expressions through natural language models, actively communicate and dialogue with humans, complete tasks such as writing papers and codes under human instructions, and even complete exam questions that are not included in its database and achieve high scores. The emergence of ChatGPT has subverted people's understanding of artificial intelligence and updated people's experience of human-computer interaction. As a kind of generative artificial intelligence, ChatGPT distinguishes itself from traditional artificial intelligence in the production of generated objects. ChatGPT can generate content such as articles and computer software under the user's operation, and there are also many generative AI that can generate content such as art and music. At present, there is no clear answer as to whether and how the products of generative AI deserve criminal law protection.
1. The scope of application of the crime of copyright infringement in the era of artificial intelligence
(1) The technical path of generative AI
The main crime involved in the criminal protection of AI-generated objects is copyright infringement. The legal interests protected by the crime of copyright infringement under Article 217 of the Criminal Code are copyright and copyright-related rights. Therefore, the question is whether the AI-generated works can become works within the meaning of the Copyright Law, and whether their rights holders can become victims of copyright infringement. The negative argument holds that ChatG-PT's content generation is no different from a "macaque selfie", and it does not have originality and therefore cannot become a work within the meaning of the Copyright Law. At present, the controversy over AI generated products mainly focuses on their nature, and the root cause lies in the difference in the grasp of the technical path of generative AI.
The technical path of ChatGPT-like artificial intelligence output generation is "data training", which is an update and expansion of "code definition". Before the advent of generative AI such as ChatGPT, AI generated creations by giving intelligent robots the way they could think in a human way by giving them a pre-programmed code. Taking artificial intelligence with translation function as an example, the mechanism for generating translated text is to set a single vocabulary concept in advance, and then arrange and combine the sentence structure of long sentences according to the program code, and finally reach the level of conforming to human language habits. Under the technical path defined by code, the generation of artificial intelligence generated objects depends on pre-set code, and it is difficult to call it "creation", which is more embodied as a tool to assist human beings. The uniqueness of ChatGPT-like generative AI lies in "machine learning", which integrates, analyzes, and models massive data resources existing on the network through algorithms, so that machines can learn human thinking methods and form computing power and output results. The key technologies of ChatGPT-like artificial intelligence are: first, the chain of thought. Faced with a problem with complex logic, ChatGPT-like generative AI can split it into several sub-problems to solve, while asking details to avoid inaccurate answers. Second, natural instruction learning. While traditional machine learning relies on tedious data labeling, generative AI can learn from instructions and generalize exponentially after a few prescriptive tasks to cope with unfamiliar tasks. Third, large language models. Generative AI uses neural network-based language models that mimic the way the human brain thinks, enhances machine deep learning, and makes accurate predictions based on massive amounts of data. Currently, GPT-4 already has 1.8 trillion parameters. Big data is driving a major change of the times, just as the advent of telescopes can give us a sense of the universe, big data is changing the way we understand the world, becoming a source of new creation, and more changes are in the pipeline.
(2) The crime of copyright infringement faces the dilemma of artificial intelligence generated objects
1. The value dilemma
The fundamental purpose of criminal law is to protect legal interests, and although there are different descriptions of the meaning of legal interests, it is generally accepted that punishment is only allowed if criminal punishment serves the protection of legal interests. In other words, the punishment provided for in Article 217 of the Criminal Law can only be applied to a certain act of copyright infringement if it infringes on the legal interests guaranteed by the Criminal Law. It can be deduced from this that if it is considered that the infringement of the products of artificial intelligence such as ChatGPT is infringing on legal interests, the criminal law can play a protective function and apply criminal penalties to the perpetrator's harmful behavior. Legal interests are essentially the interests of human life, and judging whether the infringement of artificial intelligence products infringes on legal interests is, in the final analysis, a matter of value judgment. Therefore, one of the dilemmas faced by the crime of copyright infringement in the face of AI-generated objects is the value dilemma, that is, the controversy between the "order theory" and the "private rights theory".
The "order theory" holds that the legal interests protected by this crime are the order of the state's management of copyright, on the grounds that the position of this crime in the specific provisions of the Criminal Law is "the crime of undermining the order of the socialist market economy". The "order theory" is not only a common theory in criminal law circles, but also fits the opening statement of the relevant judicial interpretations of intellectual property cases. The "New Order Theory", which was developed on the basis of the "order theory", constructively further interprets the protection of the legal interests of this crime as "the order of market competition". The "private rights theory" does not completely stand on the opposite side of the "order theory", and it considers copyright infringement to be a crime
The primary legal interest of protection is not a certain management order or market competition order, but the private rights of the right holder to copyright. The provisions of Article 217 of the Criminal Law and the way of thinking in handling copyright infringement cases in judicial practice precisely reflect the rationality of the "private rights theory" and provide sufficient arguments for the argumentation of the "private rights theory". Paragraph 1 of this article clearly stipulates that "if the amount of illegal gains is relatively large or there are other serious circumstances, the sentence shall be up to three years imprisonment", and in judicial practice, the main criterion for judging whether the crime of copyright infringement is established is also the amount of the crime, rather than considering or judging the seriousness of the violation of the management order.
There is no problem with the "order theory" and the "private rights theory" in identifying the traditional copyright infringement crime, but with the development of artificial intelligence technology and the emergence of generative artificial intelligence such as ChatGPT, the two theories are facing the dilemma of value judgment. The "order theory" focuses on the maintenance of the national management order or the market competition order, which means that in a general sense, since the current academic and practical circles have not formed a consistent view on the AI products, the infringement of the AI products does not constitute a violation of the order. The problem facing the "order theory" is the "digital divide". We have already entered the era of big data, the development of big data is related to a country's international competitiveness, related to the level of social and economic development, and because of this, the "14th Five-Year Plan" will "accelerate digital development, build a digital China" as an independent chapter, pointing out that big data is one of the seven key industries of the digital economy. Generative AI based on big data will play a huge role in the future, and in the era of artificial intelligence, almost all types of cultural works can be generated by AI. In addition, poetry, prose, photography, composition and other literary and artistic creations, generative AI can be generated, and in terms of production efficiency, human beings are almost powerless in the face of the generation speed of AI. Artificial intelligence is of great significance to the progress of social productivity, and from the perspective of value judgment, it is indeed necessary to protect the products of artificial intelligence, so as to stimulate the vitality of R&D institutions for generative artificial intelligence research and development, and prosper cultural production in the era of artificial intelligence. If the framework of the "order theory" cannot achieve the protection of AI products, if it cannot curb wanton infringement, counterfeiting, and piracy, the development of generative AI in mainland China will be inferior to that of foreign countries, resulting in a "digital divide". The question faced by the "private rights theory" is whether to give the subject of artificial intelligence the status of legal qualifications, which has been one of the controversial topics in criminal law in recent years. Those who hold a positive view believe that generative AI should be given the status of a subject, and that it should enjoy the moral rights and property rights of the author, while those who hold the negative view believe that generative AI is only a tool of human beings and does not have the status of a subject. From the perspective of the "private rights theory", the legal interests protected by the establishment of the crime of copyright infringement are private rights, so if it is believed that a crime can be established against artificial intelligence generated objects, does the private right infringed belong to the user, the designer, or the generative AI itself? From the perspective of value judgment, the "order theory" is too conservative and cannot adapt to the trend of productivity change in the era of artificial intelligence, and the proposition of some "private rights theory" theorists to give artificial intelligence the status of the subject is too radical and difficult to be widely recognized.
2. Normative dilemma
To explore the application space of copyright infringement crimes against artificial intelligence creations, the topic that cannot be avoided is the application of Article 217 of the Criminal Law. The provisions of the Criminal Law follow the principle of unity of legal order, and the types of works protected are consistent with those of the Copyright Law, including eight types of works, such as written works and computer software. At the same time, Article 217, Paragraph 1 of the Criminal Law also stipulates "other works", and Article 3 of the Copyright Law stipulates "other intellectual achievements that meet the characteristics of the work" as a catch-all clause. The problem caused by the paradigm of the Criminal Law and the Copyright Law is that neither law provides for the type of "AI work", if it is considered that the crime of copyright infringement can be established against an AI product, what specific type of work is the perpetrator infringing? One view is that the type of work "AI work" should be added to the relevant laws and judicial interpretations such as the Criminal Law and the Copyright Law, while the other view is that there is no need to add a new type of work at present, and only the AI creation needs to be interpreted as " There is also a view that it is not appropriate to adopt the provisions of the catch-all clause, but to expand the interpretation of specific types of works, and it is sufficient to classify AI-generated text as written works and generated music as musical works, and AI cannot generate works beyond the eight types of statutory works. If this divergent question is not answered, it is difficult for the crime of copyright infringement to be applied.
2. AI-generated works can become works under the Copyright Law
Defining the nature of ChatGPT-like artificial intelligence products is related to whether artificial intelligence products can become the criminal object of copyright infringement, and is the premise of the criminal law to protect artificial intelligence products. In the author's opinion, as long as the AI-generated works meet the substantive criteria of "originality", and in form conform to the types of works stipulated in the Copyright Law and are difficult to distinguish from works created by natural persons, they can be regarded as "works" and thus obtain the dual protection of civil law and criminal law.
(1) The originality of AI-generated products
Whether an AI-generated product constitutes a work is mainly judged in accordance with the definition of the work in Article 2 of the Implementing Regulations of the Copyright Law, and it is obvious that such a generated product is an intellectual achievement of human beings, and it is also an external expression that can be objectively perceived by others and can be reproduced in some form. Therefore, the key point of judgment is whether the AI generated product has originality. The author believes that AI-generated products are original for the following reasons:
First of all, the AI-generated product meets the requirement of "originality" of originality. "Independence" in the sense of the Copyright Law does not mean "uniqueness" or "uniqueness", and a labor achievement meets the requirements of "independence" in two situations: first, the labor result is an independent creation created by the worker from scratch. Second, the worker creates a new result on the basis of another person's work, and there is a recognizable and obvious difference between the work and the original work. The negative theory holds that the artificial intelligence generated does not meet the characteristics of "unique", but "intelligent search engine + intelligent text analyzer + intelligent manuscript washer", and the results generated by it are actually the product of plagiarism. In the author's opinion, this view is inappropriate. Copyright law protects expressions, not ideas. As long as the content arrangement and text combination of text works, art works and other works generated by artificial intelligence generated are original and can convey relatively complete information, they are not plagiarized achievements and can be protected by the Copyright Law. Based on the above-mentioned technical paths, generative AI such as ChatGPT can "select" and "arrange" database content through its powerful language model algorithms, although it may overlap with the existing content in thought, but it will not overlap with the existing content in objective form. For example, a student with a limited level of knowledge reserve is unable to put forward a unique point of view on a certain issue, and in order to complete the task of writing a thesis, he or she converts the existing literature into sentences and expresses his own thesis results, and the thesis results still constitute a work. The reason is that the homogeneity of ideas does not negate the originality of the paper's results, as long as there are significant differences between the expressions of its words and sentences and the references, and it meets the requirements of originality, it can be protected by copyright law.
Second, the AI-generated products meet the requirements of "creation" of originality. Neither the Copyright Law nor the Regulations for the Implementation of the Copyright Law stipulate the meaning of "creation", and the requirements for "creation" are also different in civil law countries and common law countries, but it is generally recognized that "creation" requires that the fruits of labor have room for intellectual creation. For example, if a student picks up a pen and scribbles on his workbook during class, although this kind of random graffiti meets the requirement of "independence", it cannot be regarded as a work within the meaning of the Copyright Law because it lacks the minimum intellectual creation component. As far as AI creations are concerned, the negative theory holds that existing AI is still essentially a program or algorithm, and even if generative AI can create a work on the basis of a pre-set program or algorithm, its detachment is governed by another program rather than a product of intellectual creation. The author does not agree with the view of the negative theory, in fact, the reason for the negative theory is that it does not have a deep grasp of the technical path of generative AI. As mentioned earlier, the technical path of ChatGPT-like artificial intelligence output products is "data training", which has long subverted the traditional "code definition" method. Although generative AI relies on established programs, it can simulate the functions of the human brain through a large number of parameters, so as to create corresponding content without the rules of the algorithm and without the transfer of human will. There is no substantive difference between the way generative AI is created and the content of works created by humans under neural control. In many cases, AI creations are much more intelligent than human creations. For example, ChatGPT was instructed to create a poem that imitated Du Fu's style, and the result was: "Plug the scenery and look at Qinchuan." Thousands of miles of Yellow River water, rolling east and passing away. The desert is boundless, and the Gobi is like an iron field. The years are desolate and dreamy, and they are not traces of the past. Life and death are impermanent, and several people chase yellow dogs. After returning to dust, a spring breeze remains. The imagery that appears in this poem includes "Saishang", "Qinchuan", "Yellow River", "Desert", "Gobi", etc., which is very rich in Du Fu's depressed and frustrated style, depicts the characteristic scenery of ancient China, and shows the author's feelings about time and life. At the same time, the poem created by ChatGPT is simple and natural, full of history and literati atmosphere. It can be seen that the poems created by generative artificial intelligence such as ChatGPT are not simple text arrangements or trivial graffiti, but have a high degree of intellectual creation, which meets the requirements of "creation".
Thirdly, AI products have a personality element. The negative view is that the works protected by the Copyright Law can only be the intellectual creations of human beings, not the products of machines, so the artificial intelligence generated works are not works within the meaning of the Copyright Law. In the author's view, the arguments relied on to dismiss the argument are not sufficient. One of the reasons for the negators is that Article 2 of the Copyright Law of the People's Republic of China stipulates that the subjects who enjoy copyright are "Chinese citizens, legal persons or unincorporated organizations" and "foreigners and stateless persons", and generative AI is not a legal subject, so its products naturally do not become the object of protection under the Copyright Law. The negative theory actually confuses the issue of the creator of the created work and the issue of the right holder of the created product, and the copyright law and relevant judicial interpretations stipulate that citizens, legal persons, unincorporated organizations and other subjects are only the right subjects of the work, not the creator of the work, and the relevant laws do not clearly stipulate that only the legal subject can create. The second argument of the negators is that AI-generated products do not have a personality element, and the mainland copyright law only protects creations with a personality element. In this regard, the case cited by the negators is the "macaque selfie case", in which the macaque inadvertently pressed the shutter of the camera to produce a photograph of good quality, which is indistinguishable from the photographer's meticulous photograph in form, but since the macaque does not have a personality element, the photograph certainly cannot be a work. In the author's opinion, the "macaque selfie case" cannot be a reason to determine that ChatGPT-like artificial intelligence generated objects do not constitute works, and the artificial intelligence generated objects obviously contain personality elements in them. The so-called personality is not only contained in the inner thought of human beings, but also manifested outside the expression of human thoughts. For example, the expression of ideas, attitudes, opinions, and emotions is the manifestation of the creator's personality elements. The current generative AI is not a pure machine creation, no matter how it evolves, the human element is always contained in it, which is reflected in: First, the technical path of generative AI contains the element of personality. Generative AI such as ChatGPT does not come into being in a vacuum, but relies on pre-set code and is trained on data based on code definitions, which are themselves the embodiment of personality elements. Second, generative AI is created in a way that includes an element of personality. The process of generative AI output creation does not rely entirely on the machine itself, but rather on the process of human-computer interaction. For example, if ChatGPT is asked to compose a poem, it needs to be instructed what kind of poem to compose, and ChatGPT can even be asked to imitate a certain poet's style and limit the poetry genre. In this human-machine synthesis process, ChatGPT's works are still completed by artificial intelligence under the guidance of humans, which significantly reflects the implication of personality elements.
Finally, cases in judicial practice can reflect the copyrightability of AI-generated works. In the copyright infringement dispute case of Company A v. Company B, Company A used an artificial intelligence software called XX to publish a financial report article on its securities website, and at the end of the article, it was noted that "this article was automatically written by Robot XX of A", and the defendant Company B published the article on its website without permission, and the title and content were exactly the same as those published by Company A. In March 2020, the Nanshan District People's Court of Shenzhen explicitly determined that AI-generated articles constituted works in the first instance, which is the first case in mainland China to determine that AI-generated articles constitute works, which is of typical significance. Deniers cite the "Shortest Road to Heaven" painting registration case in the United States to deny the copyrightability of AI-generated works. In that case, Taylor asked the U.S. Copyright Office to register copyright on a painting and claimed that the painting was created entirely by artificial intelligence, but the U.S. Copyright Office refused to register the painting on the grounds that copyright law only protects the fruits of human intellectual labor. However, the author believes that this case cannot deny the copyrightability of the AI-generated works, and the reason for the rejection of Taylor's copyright registration request is not that the AI-generated works do not constitute works per se, but because of a strategic error. At present, the production of artificial intelligence generated by human interaction is the result of human-computer interaction, and there is no product that is completely generated by machines without human intervention. Taylor's application was based on the fact that the painting "The Shortest Path to Heaven" was produced purely by artificial intelligence, which was wrong, and the U.S. Copyright Office was justified in denying his request.
(2) Copyright ownership of AI-generated works
If an AI-generated work meets the standard of originality, it can constitute a work within the meaning of the Copyright Law, and the next issue is the copyright ownership of the AI-generated work. In order to constitute the crime of copyright infringement as provided for in Article 217 of the Criminal Law, it is necessary to meet the conditions of "without the permission of the copyright owner", "without the permission of the producer of audio and video recordings", "without the permission of the performer", and "without the permission of the copyright owner or the right holder related to copyright". Obviously, the copyright ownership of AI-generated works directly affects the applicable space of copyright infringement crimes.
Regarding the ownership of artificial intelligence products, mainland scholars have put forward different opinions from the perspective of generation mechanism and balance of interests. The "neighboring rights theory" argues that only limited protection can be granted to AI products based on the "sweaty forehead principle". In other words, it is believed that the "originality" of AI-generated products is insufficient to be protected by copyright in the narrow sense, but in order to protect the property interests of investors, they are allowed to be included in the protection scope of neighboring rights. The "fruit theory" holds that the AI generated works do not belong to works, but are a kind of "thing", and the AI generated works should be interpreted as "intellectual property fruits" through the interpretation of civil law, and the ownership of the AI generated objects should be determined in accordance with the civil law's own attribution rules. However, neither of these two views on the ownership of AI-generated objects is considered to constitute a work, so the author does not agree with them.
An AI-generated object conforms to the characteristics of the work in form and meets the criteria of originality in substance, has elements of personality, and can constitute a work. Under the condition of human-computer interaction synthesis creation, ChatGPT and other generative artificial intelligence and humans have made substantial contributions to the creation of works, so the ownership of the copyright of the generated works can be handled with reference to the works of legal persons or according to the agreement. In the theory of copyright law, when a natural person or legal person completes the creation of a work, he or she originally obtains the copyright of the work in accordance with the "principle of automatic acquisition of rights". The "principle of automatic acquisition of rights" is not fully applicable in the context of the rapid development of generative AI, where the creator of the work is AI, and AI is not eligible to enjoy copyright because it has no independent ability to will. Therefore, in the era of artificial intelligence, the subject structure of dualism should be adopted to distinguish between the qualifications of creative subjects and the qualifications of rights subjects. In a nutshell, generative AI can be the author of a work, but it does not enjoy copyright and copyright-related rights. The legislative purpose of the Copyright Law is to protect the exclusive rights of authors and prosper cultural creation, and through the construction of a dualistic subject structure, even if the authorship of generative AI is recognized, it will not affect the development of social, cultural and scientific undertakings, because the exclusive rights of rights holders (mainly copyright property rights) will not be affected.
There are three possible ownership possibilities for AI-generated objects: the first is to belong to the user. Through the instruction, the user makes the generative AI create a work that meets his own requirements, and in the process, the user makes a substantial contribution to the production of the generated object and injects his own personality elements. Although the investors and developers of generative AI have also made contributions with personality elements from the perspective of "code definition", in terms of importance, the production of a specific work mainly depends on the user's instructions, and the investors and developers have no control over this process. The second is to vest investors and developers. At present, many companies are leading the development of artificial intelligence technologies, such as "Microsoft Xiaoice". In these cases, the provisions of the Copyright Act on the works of legal persons may apply. The third is as agreed. When using generative AI software, users are generally required to agree to the User Agreement in advance, so the ownership of the Created Objects can be determined based on the content of the agreement. In the case of ChatGPT, OpenAI agreed to transfer all of its rights to the AI-generated artifacts in the User Agreement.
3. Artificial intelligence (AI) generated objects may be the target of copyright infringement crimes
Since an AI-generated object can be considered to constitute a work, if an act infringes upon the legitimate rights and interests of the copyright owner, it must violate the prohibitive norms of the Copyright Law, and the crime of copyright infringement can only be constituted when the infringement of legal interests reaches a certain level. This is a requirement of the principle of uniformity of legal order, and it is also a manifestation of the fact that the crime of copyright infringement is a statutory offense. However, even if it is proved that the AI creation is a work within the meaning of the Copyright Law, it cannot be automatically considered that the creation is the object of the crime of copyright infringement. Only when it is determined that the protection legal interests of the crime of copyright infringement cover the rights enjoyed by the right holder in the AI-generated objects, can it be fully clarified that the AI-generated objects are the criminal objects of the crime of copyright infringement.
First of all, it should be clarified that the legal interests protected by the crime of copyright infringement should be clarified. Both the "order theory" and the "private rights theory" are facing a dilemma in the era of artificial intelligence, and they cannot properly deal with the protection of artificial intelligence generated by copyright infringement crimes. Therefore, it is necessary to adopt the two-layer legal benefit theory to solve this problem. Professor Zhang Mingkai believes that "in criminal law, there is a large number of legislative phenomena that protect the legal interests of A (the back layer) and protect the legal interests of B (the blocking layer) in order to protect the legal interests of A. That is, A's legal interests are individual interests, and B's legal interests are collective interests. In the author's opinion, the crime of copyright infringement provided for in Article 217 of the Criminal Law is a typical case of the legislative phenomenon of the two-tier theory of legal interests. The crime of copyright infringement is stipulated in the chapter "Crimes of Undermining the Order of the Socialist Market Economy", which reflects that the legal interests of the blocking layer of the crime are certain orders, while the legal interests of the underlying layer are the private rights of copyright. If the traditional "order theory" is still adhered to, it will be unable to cope with the challenges brought about by the artificial intelligence era and cause the "digital divide" to be aggravated. Therefore, it is undoubtedly more reasonable to combine the two theories to construct a two-tier protection of legal interests for the crime of copyright infringement. In other words, in order to constitute the crime of copyright infringement, it is necessary to violate a specific order and infringe on the private rights of the right holder.
Second, the rights of the right holder over the AI-generated products should be clarified. First, the right holder does not enjoy moral rights in respect of AI-generated works. Generative AI is the creator of generative works, which means that no one other than generative AI has moral rights in generative works, and although humans have made some substantial contributions in the process of creating works by AI, they play a participatory role rather than playing the role of creators. According to the theory of civil law, moral rights are non-transferable and inheritable, that is, human beings cannot acquire moral rights in some way. If it is determined that these rights belong to generative AI, the perpetrator will constitute a civil infringement when infringing these rights, and it can be deduced that if a civil infringement is constituted, the infringer shall bear civil liability such as damages, and the infringer shall compensate the AI robot. Under the existing legal framework, this is obviously absurd, and AI robots do not have the capacity for civil conduct and civil liability, and cannot become civil subjects. Therefore, there is currently no eligible subject with moral rights in AI-generated works. Second, the right holder can enjoy the property right of the AI-generated works. Section 106 of the U.S. Copyright Act uses the term "copyright owner" to define the subject who can enjoy the property rights of the author, not the "author", which means that the moral rights of the author and the property rights of the author can be separated. Moral rights are a kind of moral rights, and generative AI such as ChatGPT does not have an independent personality, and will not stop creating works because their own works are made public by others or signed by others, and AI does not enjoy moral rights. The denial of the moral rights of the authors of AI-generated works does not mean that there are no qualified subjects of the property rights of the authors. Under the current legal framework, the copyright of AI-created works can be attributed to users, investors, or developers, as appropriate. Third, the right holder does not have neighboring rights over AI-generated products. Article 217 of the Criminal Law not only protects copyright in the narrow sense, but also protects neighboring rights such as performers' rights and audio and video producers' rights. The object of protection of neighboring rights is the disseminator of the work and the creator of the fruits of labor outside the work, and the owner of the right of AI-generated works enjoys the property right of the work, so it is not the object of protection of neighboring rights. Artificial intelligence generated works can constitute works, and performers who perform on the works can enjoy performer's rights, and if the performer's rights are infringed, the crime of copyright infringement can also be established if the constitutive elements are met. However, the object of copyright infringement here is not artificial intelligence creation, but neighboring rights.
Finally, it should be made clear that the legal interests protected by the crime of copyright infringement cover the rights enjoyed by the right holder. Under the framework of the two-tier legal interest theory, the legal interests protected by the crime of copyright infringement include both specific order and private rights. The traditional theory adopts the "order theory" alone, and does not believe that the infringement of ChatGPT-like artificial intelligence products constitutes a violation of the national management order, because it does not cover private rights, is restricted by established legal norms, and is unable to protect artificial intelligence products that have not yet been provided for in laws and judicial interpretations, and lags behind the new needs of judicial practice. The two-tier legal interest theory can effectively correct the shortcomings of the traditional "order theory", and the "new order theory" adopted by it no longer believes that the protection of legal interests of copyright infringement crimes is only the state's management order of copyright, but also includes the market order. The "New Order Theory" is obviously in line with the needs of the development of the AI era, and the perpetrator's infringement of AI products constitutes an infringement of the market competition order. The legal interest behind the "new order" is "private rights", so the actor's behavior constitutes an infringement of the right holder's copyright property rights. Therefore, the protection of the legal interests of the crime of copyright infringement can effectively cover the rights enjoyed by the right holder, and it can be determined that the artificial intelligence generated object is the object of the crime of copyright infringement.
4. The protection path of criminal law for artificial intelligence generated objects from the perspective of copyright infringement
Artificial intelligence (AI) generated works can constitute works within the meaning of the Copyright Law and the object of copyright infringement crimes, so the criminal law can provide protection for AI-generated works. On the one hand, the protection of the criminal law means that the copyright property rights of the right holder are effectively protected, which will have a positive effect on the development of artificial intelligence technology, encourage investors to invest in the field of generative artificial intelligence, and inject strength into the development of socialist cultural undertakings; In essence, they can create works comparable to those created by natural persons, which makes the competition in the market more and more fierce, and reversely stimulates natural persons to create more high-quality works. After all, AI-generated products are new, and there are many ways to protect them for copyright infringement that need to be refined.
(1) Expansive interpretation
Artificial intelligence (AI) generated works can be the object of copyright infringement crimes, and what kind of works belong to this category has not yet been uniformly recognized. Some views hold that in addition to the circumstances already provided for in Article 217 of the Criminal Law, "infringing the copyright of others by other means" should be added as an open criminal provision. There is also a view that the crime of copyright infringement should be added, and along this line, there is also a view that the statutory type of work of "artificial intelligence works" should be added, which is not feasible in the author's opinion. First, the object of the crime of copyright infringement is the work or the object of neighboring rights under the specific mode of conduct of the perpetrator, and the four modes of conduct provided for in Article 217 are arranged in a logical order, and different modes of conduct correspond to different objects of conduct. If "AI works" are added, what is the corresponding behavior mode of "AI works"? Is it "reproduction and distribution" or "publication", or "information network dissemination"? Second, the solution of adding "AI works" is essentially the same as interpreting AI products as "other works". According to the rules of interpretation of the Criminal Law, the "other works" interpreted must be on the same level as the types of works specified in the specification, and in fact, the AI-generated works become a new category of works. Third, the addition of "AI works" is not conducive to the coordination of legal norms. At present, the copyright-related laws and judicial interpretations such as the Copyright Law of the People's Republic of China and the Regulations for the Implementation of the Copyright Law do not stipulate the statutory type of "artificial intelligence works", and if this type of work is added to Article 217 of the Criminal Law, it will lead to a mismatch between the copyright law and the criminal law. Therefore, the opposition to the addition of "artificial intelligence works" is based on the consideration of legislative cost.
In view of the disadvantages of adding "artificial intelligence works", the problem of legal interpretation should be solved by expanding the way of interpretation. The inclusion of AI creations within the scope of protection of criminal law is not a legislative issue, but a matter of legal hermeneutics. From the perspective of interpretation path, the criminal object of Article 217 of the Criminal Law can be expanded to include artificial intelligence creations. For example, the term "written work" in the provision is interpreted to mean "artificial intelligence written work" and "traditional written work". Other AI generated by analogy. On the one hand, interpreting "AI literary works" as "literary works" does not exceed the semantic range of "literary works" and does not violate the basic principles of criminal law interpretation. On the other hand, expansive interpretation refers to the inclusion of concrete things into abstract legal terms, and legal terms must be abstract and open concepts to have room for expansive interpretation. For example, "children" and "young girls" are closed concepts, and there is no room for expansive interpretation. Article 217 of the Penal Code stipulates that the objects of the offences are all open concepts that can be applied to the method of expansive interpretation. For example, "written works" include poetry, prose and other forms, and it is realistic and reasonable to interpret AI-generated text content as "written works".
(2) Defining standards
Breakthroughs have been made in generative AI technology, and the emergence of many "black technology" products has attracted attention. However, in various creative fields, it seems that generative AI has not defeated humans, nor has it caused significant changes in people's production and life. In Company A v. Company B, although the court clarified for the first time that an AI-generated object could constitute a work, the infringement of the AI-generated object was not serious enough to require criminal law intervention. Considering the lack of infringement of legal interests, some scholars believe that the protection of artificial intelligence generated by criminal law is a waste of legislative and judicial resources. However, the response to social risks requires the criminal law to be proactive, and under the concept of positive criminal law, it is indeed necessary to think forward-looking about relevant issues, rather than lagging behind social development.
The protection of AI creations for copyright infringement requires a strict definition of the standard, i.e., not all AI creations are worthy of criminal law protection, and they must meet the standard of originality. It is not difficult to protect AI-generated works through the method of extended interpretation, but how to determine which AI-generated works meet the constitutive requirements of works and have the value of legal protection is a more worthy of in-depth study. In the current judicial practice of copyright infringement, it is usually necessary to go through a professional appraisal to determine whether an act constitutes infringement. Therefore, the author believes that the mainland's appraisal system should be updated to strictly define the criteria for what constitutes a work of AI-generated works.
(3) Retention limit
AI-generated works are different from traditional works in many ways, and the criminal law should grant them less protection than works of natural persons based on their characteristics. On the one hand, there is no subject with moral rights for AI-generated works. Although Article 217 of the Criminal Law mainly protects the property right of the author, the right holder can indirectly protect the moral right of the author through the protection of the property right of the author, and the AI-generated works have nothing to do with the moral right of the author. The social harm of infringing on the copyright of natural persons' works is significantly higher than that of infringing on the copyright of artificial intelligence works. The imposition of the same penalty for infringement of two different works is a violation of the principle of proportionality of punishment and crime. On the other hand, from the perspective of criminal policy, there should be a distinction between infringement of the works of natural persons and the works of artificial intelligence. ChatGPT-like generative AI can generate countless written works every day, and the value of these works is lower than that of human works, and if equal protection is too harsh for the perpetrator, it is necessary to adopt a lenient criminal policy to seek a balance between the prosperity of socialist cultural undertakings and the protection of the legitimate rights and interests of the perpetrators.