Original title: Sui Haixu: How artificial intelligence products represented by ChatGPT seek protection under copyright law
Nobel Prize winner Mo Yan said at the anniversary celebration of "Harvest" magazine that he wanted to write an award speech for Yu Hua for several days and could not write it, so he used ChatGPT to input "alive", "tooth extraction" and "Wencheng", and then instantly generated a Shakespearean style, more than a thousand words of praise, so does this beautifully written award speech have copyright? Can Mo Yan sign it as his own work? In order to make a promotional image for the store, a coffee shop owner uses Midjourney to let him generate a picture of making coffee based on the actual picture of the store according to the boss's personal requirements for element combination, color preference and visual effects. Is it a work of art? Can someone else use this image without the coffee shop owner's consent?
To answer the above series of questions, it is necessary to explore the copyright law protection of artificial intelligence products[1]. The key to judging whether artificial intelligence products enjoy copyright is whether they reflect human originality, the criteria for judging the originality of artificial intelligence products are different from other works, and should not be protected as the same type of work only from the appearance, at this stage, artificial intelligence products can be identified as "other intellectual achievements that meet the characteristics of the work" and included in the protection of the Copyright Law, and the user as the creator of artificial intelligence products meets the authorship and enjoys relevant rights. At the same time, we should explore other modes of protection that distinguish them from the protection of existing types of works.
1. Artificial intelligence products that embody human ingenuity should be protected by the Copyright Law
[Whether or not artificial intelligence products enjoy copyright depends on whether they can reflect human originality]
Some domestic scholars compare artificial intelligence products to "macaque selfies"[2], believing that artificial intelligence and macaques do not belong to human authors, so artificial intelligence products and macaque selfies should not be protected by copyright law like macaque selfies[3]. This view ignores the difference in the degree of human participation between the two: macaque selfies are purely accidental and autonomous acts of macaques, and in the process of "creating" a selfie, people do not need to do anything other than provide a camera; In the process of the formation of artificial intelligence products, human participation behavior is not only reflected in the research and development of artificial intelligence - including the screening of massive data information, continuous deduction and iteration of algorithm technology, repeated training of output content, but also reflected in the stage of end-user use - including the way to propose problems, the selection of keywords, interactive feedback with artificial intelligence, etc., artificial intelligence products follow the human will and accept human instructions, arrangements and other auxiliary operations to complete. Of course, a series of human activities in the process of the formation of artificial intelligence products are not necessarily protected by the Copyright Law, and it is necessary to judge whether the human activities in them meet the requirements of the Copyright Law for creation, that is, whether they reflect human originality.
Aside from the artificial intelligence products themselves, the developer's research and development program can be protected as a type of work such as computer software or product design drawings, and the text edited by the user when using artificial intelligence that meets the characteristics of originality can be protected as a written work, and it can be protected as a compilation work when the relevant materials are sorted, arranged, and selected to form an original work, and the above protection methods are relatively clear in the existing copyright law. This article examines the protection of the ultimate manifestation of artificial intelligence products in copyright law.
[The Copyright Law does not protect the creation of artificial intelligence itself independent of humans]
With the development of artificial intelligence represented by ChatGPT, we must admit that artificial intelligence has embodied its own unique creative functions and creativity-related creativity, which has led to the inability to distinguish whether it is a human creation or an artificial intelligence creation from the appearance alone.
However, AI's own human-independent creation and its "creativity" are not protected by copyright law. According to the second paragraph of Article 11 of the Copyright Law, the natural person who creates a work is the author, only the natural person has the author's qualification, and only the "creation" of the natural person is protected by law. Second, the "creativity" of artificial intelligence itself does not meet the requirements of the Copyright Law for the "originality" of works - especially refers to the originality unique to human beings themselves, which carries personality, although artificial intelligence is more skilled than a large part of human beings in terms of word and sentence composition or composition and painting, and also has a creative speed far beyond all human beings. Third, the legislative purpose of the Copyright Law is to encourage, protect and disseminate human creation, if the same protection of artificial intelligence itself creation, the object of encouragement can only be people, people will in order to promote artificial intelligence creative skills, enhance artificial intelligence research and development skills (development side) and their own use of artificial intelligence skills (user side), but there is no motivation to enhance their own creative skills, this incentive consequences from the perspective of scientific and technological development and human control tools are progressive and benign, Protection can also be obtained from other laws, but this is not what copyright law is intended to protect. Just as in the case of Beijing Film Law Firm v. Baidu Company for Copyright Infringement, the court held that the analysis report was formed by the combination of input keywords and algorithms, rules and templates by Wolters Kluwer, and in a sense, it can be determined that Wolters Kluwer "created" the analysis report, and even if the analysis report "created" by Wolters Kluwer is original, it is still not a work within the meaning of copyright law. [4]
[How to judge whether artificial intelligence products embody human originality]
The originality required by the Copyright Law generally includes two aspects, one is independent creation, and the other is that the result of the creation has a minimum degree of creativity. It is not difficult to judge the requirements of independent creation, and there are certain subjective factors in judging the minimum creative characteristics, but it is certain that the minimum creativity is a qualitative requirement, and today, whether in China or the United States, the minimum creativity of the work is required. [5] Since the Copyright Law does not protect the creation of artificial intelligence itself, the "originality" belonging to artificial intelligence itself should not be taken into account when judging whether the artificial intelligence product has the minimum creativity of a person. Specifically, it can be considered from the following three aspects: first, judge whether people have creative behavior in it; secondly, judge whether the creative act is original; Once again, it is necessary to eliminate the creations belonging to artificial intelligence in the artificial intelligence products to determine whether they can still reflect human originality.
Returning to the two examples at the beginning of the article, an award speech of more than a thousand words generated only by entering the words "alive", "tooth extraction" and "Wencheng" does not enjoy copyright, because people do not have creative behavior in it, the three keywords in the instruction, two of which are the name of Yu Hua's work, one is Yu Hua's former professional content, and does not have the minimum creativity required for creation, and the final award words are all the creations of artificial intelligence itself, which does not reflect the originality of people in it. In the second example, if the coffee shop owner uses artificial intelligence tools to upload selected real-life photos of the existing operation tables, coffee beans, coffee machines, tables, chairs and other furnishings in the store, and then describe to Midjourney in detail the image of the barista, the movements and demeanor, the type of coffee, the color, the production progress, how to match with other elements in the image, how to choose the color and light, and finally how to present a series of effects such as well-made, well-selected and unique taste, and may adjust the instructions several times after generation to clarify the requirements, and the final generated image should enjoy copyright. Because the coffee shop owner has creative behavior and a certain degree of creativity in the formation of this picture, the generated picture not only has the part directly created by the artificial intelligence itself, but also reflects the coffee shop owner's own originality in thought and aesthetics, and should be protected by the Copyright Law.
In the case of Tencent v. Shanghai Yingxun in the case of copyright ownership dispute heard by the Nanshan District Court of Shenzhen Municipality, when the court judged whether the article generated by Tencent's main creative team using the Dreamwriter software was original, it had the following expression: its external performance met the formal requirements of the written work, the content of its performance reflected the selection, analysis and judgment of the relevant story information and data of the morning, and the article had a reasonable structure, clear expression logic, and a certain degree of originality. [6] It can be seen that when the court makes a judgment on the originality of artificial intelligence products, it adheres to the same standard as the original judgment of ordinary works, and does not exclude the part belonging to the artificial intelligence itself and then judge whether it is in line with human originality, which is wrong, which means that the Copyright Law gives the same legal protection to human creations as artificial intelligence creations itself, and means that the Copyright Law also protects the creations of non-human authors.
In summary, the part created by artificial intelligence itself should not be the object of copyright law protection, the original judgment of artificial intelligence products should be distinguished from other works, and artificial intelligence products that embody the requirements of human originality should be protected by copyright law. Referring to the practice in the United States, although the United States Copyright Office (USCO) revoked the registration of images generated by Midjourney in Charia of the Dawn on February 21, 2023, retaining only the author's copyright in the story and image arrangement of the creation, on the grounds that US copyright law does not protect works created by non-human authors,[7] but in the Guidelines for the Registration of Copyright in Works Containing Artificial Intelligence Generated Works[8] issued on March 16, 2023, it states that When judging the review of a work containing AI-generated things, they consider whether the work was created by a human author with the aid of a computer or other device, or whether the most basic elements that should have been required by the author to complete the work were actually done by machines... It will be considered whether the part of the contribution of AI is the result of a machine or an external expression derived from the inner heart of the human author, and whether or not to obtain copyright registration depends on how the specific work uses the AI and ultimately creates the work, and needs to be judged on a case-by-case basis. It can be seen that the US Copyright Office does not take a comprehensive negative attitude towards the issue of whether artificial intelligence products constitute works, and does not rule out the future copyright registration of artificial intelligence products that meet the requirements of originality.
Second, how should artificial intelligence products that embody human ingenuity be protected
[What kind of work does the artificial intelligence product that embodies human originality belong to]
Although the Copyright Law does not protect AI itself from human creation, because the AI product that embodies human originality is inseparable from the creation of human and artificial intelligence, the Copyright Law must protect it as a whole, which means that the AI product is a new type of work different from the existing type of work, although judging from the appearance, it is no different from works such as written works or art works. If the law treats it as the same type of work, it is the same as judging originality by treating the originality of artificial intelligence as human originality, which means that the law gives the same legal protection to artificial intelligence creation as human creation, which is equivalent to recognizing the creator identity of artificial intelligence.
In line with the principle that AI products should be protected and distinguished from other works, it is appropriate at this stage to identify AI products as "other intellectual achievements that meet the characteristics of works" as stipulated in the Copyright Law and thus include them in the protection of the Copyright Law. Article 3 of the Copyright Law, in addition to clearly listing eight types of works, including written works, oral works, musical works, and fine art works, also includes the ninth provision on "other intellectual achievements that conform to the characteristics of works", which is a major adjustment made by the Copyright Law in 2020, making the mainland Copyright Law's provisions on the types of works from the previous "other works stipulated by laws and administrative regulations", that is, the legal model of work types, to the existing model of opening up the types of works. Some scholars have criticized this revision, arguing that the mode of opening up the type of work is not in line with the current actual national conditions of the mainland, and can easily lead to problems such as excessive discretion of the courts and the expansion of individual private power. Although special caution is required for the identification of other types of works other than statutory types of works, such determinations do not break through existing legal norms, but can effectively protect people's creative behavior in artificial intelligence products, and are legal innovations that need to be made to adapt to the development of new technologies.
[Who owns the copyright when the artificial intelligence generated works constitute a work]
The formation of artificial intelligence products is inseparable from human intellectual activities at the research and development end and the use end, and the relevant rights of artificial intelligence products when constituting works are nothing more than the two types of subjects of developers or users. Some scholars have proposed that the relationship between OpenAI (ChatGPT developer) - ChatGPT (artificial intelligence) - user (user) can be regarded as a kind of commissioned creation, and the ownership of copyright is determined in accordance with the provisions of Article 19 of the Copyright Law, that is, there is an agreement from the agreement, and if there is no agreement, the copyright belongs to the trustee, that is, the developer. [9] The above view cannot be justified, since it is "using ChatGPT to create a work", it is obvious that the subject of the creation of the work is the user, not the developer, and the definition of the author in the Copyright Law is "the natural person who created the work", and when the artificial intelligence generated by ChatGPT constitutes a work, only the user and ChatGPT have creative behavior, and the law only recognizes the natural person in it, that is, the authorship of the user. The owner of the coffee shop is a user of artificial intelligence software and a producer of promotional images, and the promotional images he produces reflect his own originality, and should become the author of promotional images and enjoy relevant rights.
In fact, developers develop and manufacture artificial intelligence focusing on the development and upgrading of the technology itself, and provide technology to help users obtain the value of the product in exchange for income (usually reflected in sale, licensing or other commercial arrangements), and there is no demand for copyright law protection for artificial intelligence products, which is why the plaintiffs are users in two typical cases of copyright infringement disputes in the field of artificial intelligence in China.
I have to mention the user agreement of artificial intelligence developers, taking ChatGPT in the United States, Midjourney and China's "Wen Xin Yige" as examples, and the three companies stipulate three different models for the ownership of generated things. In summary, ChatGPT stipulates that all product rights belong to the user; Midjourney is divided into paid users and non-paying users, paying users have the rights to all the generated items, and non-paying users only have the right to use the generated products (ownership belongs to Midjourney); Wenxin Yige stipulates that all product rights and interests (including intellectual property rights) are owned by Baidu. From the perspective of copyright law, the relevant provisions of the ChatGPT user agreement are more reasonable, that is, the user enjoys the copyright of the generated thing (if any).
[Exploring other modes of protection that distinguish them from the protection of existing types of works]
Looking at the current legal attitude of major international countries or international treaties to the copyright of artificial intelligence generated things, most regions, including the United States, do not support the copyright of non-human creations, and the United Kingdom is one of the few countries that has passed the law to provide for the copyright of computer-generated objects, and the Copyright, Designs and Patents Act 1988 [10] stipulates that computer generated things refer to the absence of any human author. works generated by computer operations[11]; In the case of a computer-generated textual, dramatic, musical or artistic work, the author is the person who makes the necessary arrangements for the creation of that work. However, the Act distinguishes between computer-generated objects and works created by human authors, and the term of protection for computer-generated objects is only 50 years, while the term of protection for works by human authors is 70 years.
Before the revision of the Copyright Law in 2020, the term of protection of the right of publication and all property rights of "cinematographic works and works created by methods similar to filmmaking" and all property rights was 50 years, while the term of protection for other types of works was 50 years after the life of the author and 50 years after his death, which shows that the term of protection for works filmed with cameras is shorter than that of other types of works, of course, after the revision of the new Copyright Law, the term of protection for photographic works is no longer treated differently. Only audiovisual works have a shorter term of protection than other types of works (for photographic works that have expired before the entry into force of the new law, but are still under the new law, they are no longer protected).
Therefore, referring to the above-mentioned British Act on the differential protection of computer-generated objects, as well as the protection of photographic works in the copyright laws of various countries since the emergence of photographic technology, the mainland's Copyright Law can consider adding artificial intelligence products as a new type of work, and because artificial intelligence products contain a lower proportion of human creativity and use more social public resources, it can also consider appropriately shortening their protection period.
epilogue
The inspiration for writing this article comes from the example of "macaque selfie" mentioned at the beginning of the article, when someone tries to compare artificial intelligence to macaques, the author believes that artificial intelligence is more like a camera, in fact, the question of whether a photograph should be recognized as a "work" and its originality has also been controversial, photographic works were not explicitly listed as protected works by the Brussels Act of the Berne Convention until 1948, and the term of protection of photographic works is also differentiated in the copyright laws of various countries, including the mainland, The reason is that many people think that in the completion of photographic works, the creative act of people may simply press the shutter. Artworks that are very similar in appearance to photographic works carry a higher degree of human creativity when expressing the same or similar objects, although photographic works are higher than art works in terms of creation efficiency and object fidelity, just as today's generative artificial intelligence can far surpass humans in terms of creative efficiency and creative level. However, the copyright law whose mission is to protect the expression of human originality has never been based on success or failure, and even if one day the expression of artificial intelligence in creativity, conception, aesthetics and other aspects has crushed human beings in an all-round way and even grown their own thoughts and souls, the copyright law still only favors every human creator, until artificial intelligence has the ability to abolish this law with its own hands.
● Notes
[1] "Artificial intelligence generated" in this article refers to text, pictures, sounds, videos, codes, etc. directly generated based on algorithms, models, rules and other artificial intelligence technologies, which refers to the definition of "generative artificial intelligence" in the Measures for the Administration of Generative Artificial Intelligence Services (Draft for Comments) of the Cyberspace Administration of China, and the specific form can refer to the generated content of ChatGPT.
[2] Macaque selfie refers to a British photographer photographing macaques in Indonesia, a macaque snatched the camera and took a selfie, and the photographer posted it online to go viral. An animal welfare institute then filed a lawsuit in a U.S. court alleging that the online distribution of photographs infringed the macaques' copyrights, and both U.S. courts rejected the plaintiffs' claims on the grounds that U.S. copyright law only protects human creations.
[3] See Wang Qian, "Can ChatGPT-Generated Content Be Protected by Copyright Law?" "Exploration and Controversy" No. 3, 2023.
[4] See Beijing Internet Court (2018) Jing 0491 Minchu No. 239 Judgment.
[5] The well-known case of Feist decided by the US Supreme Court clearly pointed out that it is not the amount of labor paid by the author that determines the originality of a work, but the quality of labor, negating the traditional principle of "sweating on the forehead".
[6] See Shenzhen Nanshan District People's Court (2019) Yue 0305 Minchu No. 14010 Judgment.
[7] See https://www.copyright.gov/docs/zarya-of-the-dawn.pdf,
Visited: June 17, 2023.
[8] See also
https://www.copyright.gov/ai/ai_policy_guidance.pdf，
Visited: June 17, 2023.
[9] See Yangcheng Evening News, "Is ChatGPT-Generated Content a Work?" Who owns the copyright? An interview with Xu Meiling, a researcher at the Guangdong-Hong Kong-Macao Greater Bay Area Intellectual Property Development Institute at Peking University.
[10] https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/957583/Copyright-designs-and-patents-act-1988.pdf, accessed June 17, 2023.
[11] The definition of computer-generated organisms in the UK Act differs from the AI-generated creatures discussed in this article, which does not require a human author, and the person who makes the necessary arrangements for the creation is considered the author; The latter requires the author to have an actual creation and to reflect originality in the work.
Special Notice:
Dentons strictly abides by its obligations to protect clients' information, and the content of the client projects involved in this article is taken from publicly available information or with the consent of clients. The full content and opinions are for reference only, do not represent any position of Dentons, and should not be regarded as issuing any form of legal opinion or recommendation. If you need to reprint or quote any content of the article, please communicate the authorization by private message, and indicate the source at the beginning of the article when reprinting. No content in such articles may be reproduced or used without authorization.
