Because of the popularity of ChatGPT, the question of whether AI-generated works are subject to copyright protection has been hot recently, and the author's interview responses to such questions are basically that according to the current legal provisions, AI-generated works have no copyright. But I asked ChatGPT, the most famous artificial intelligence program today, and the answer was that AI-generated works are protected by copyright, and OpenAI, the developer of ChatGPT, also wants to claim the revenue rights of the works. Let's talk to you about this topic today.
Let me start with my point of view: AI-generated works have inherent copyright defects and should not enjoy copyright and charge licensing fees. If a user's work generated using artificial intelligence is infringed by others, he or she can protect his rights in accordance with the Anti-Unfair Competition Law.
1. The original sin of copyright of artificial intelligence
Artificial intelligence needs to charge copyright licensing fees on a large scale, and its own confidence is insufficient, because its training process is suspected of large-scale infringement, there is original sin, and copyright fees cannot be logically consistent.
In the development process of ChatGPT artificial intelligence, developers must use a lot of data and resources to let them learn and train, but now the major artificial intelligence companies are secretive about the source of learning resources, in addition to trade secret considerations, the most important problem is that the current mainstream artificial intelligence, its training relies on the development company to crawl the free public data and resources on the Internet, and then feed and train the artificial intelligence, these data and resources are obtained without the consent of the crawled website. So it is very controversial[i], and there have been media prosecutions about it[ii].
The author asked ChatGPT about this, and the answer was: the training data is online public data, and unauthorized and unauthorized authorization is not required.
The author's opinion is that this crawling behavior is suspected of infringement. Whether the crawling behavior is legal, the mainstream practice on the Internet is to look at the Robots protocol, as long as the Robots agreement of the website that crawled the data agrees[iii], the search engine crawling data is legal. But whether AI companies crawl data is subject to the Robots protocol is highly controversial because they crawl for different purposes than search engines.
Website operators agree to search engine crawling, because this crawling is a common benefit behavior, search engine crawling data can index the website, search engine users can better access the website, commercially on the crawled website has a role in promoting access.
The crawling of artificial intelligence companies is not so, just to grab data to train their own artificial intelligence programs, only for their own benefit, which makes website operators pay server costs, their technical maintenance costs and management costs after getting nothing, this behavior should not apply to the Robots license, but should apply the Copyright Law.
According to the Copyright Law, the process of letting artificial intelligence learn online content is an act of copying or temporary copying, and the artificial intelligence company must first crawl the content obtained online or offline, and then input it into the artificial intelligence program, whether the content is text, pictures, audio, video or program, the copying behavior should obtain the permission of the corresponding right holder, otherwise it is suspected of infringement.
There is a small possibility that AI learning online content is a temporary copy, that is, the AI deletes data after learning. According to the judicial practice in mainland China, although temporary reproduction does not require permission from the copyright owner, there are still questions about whether the source of the data is publicly crawlable data and whether the non-online data (such as books and documents) is electronic[iv]. In addition, at least as far as Microsoft's new Bing built-in ChatGPT search results are concerned, the output content contains a reference link, which indicates that ChatGPT probably stores the corresponding website content, so it does not constitute a temporary copy.
Some websites follow the open source license, as long as it complies with the provisions of the open source license, you can freely copy and redistribute the content of these websites, such as Wikipedia's GNU free documentation license [v], for such websites, artificial intelligence companies should be able to crawl content, but the source still needs to be indicated when republishing. And OpenAi website ChatGPT output content, not even a data source link to the user (Microsoft's New Bing on ChatGPT is to the source link), although ChatGPT output content may be different from the wiki, but this only demand, no common benefit behavior is undoubtedly an erosion of open source culture, I personally think that if Wikipedia sues in China, in addition to the Copyright Law, you can also invoke Article 2 of the Anti-Unfair Competition Law, Claims that OpenAI's crawling practices violate accepted business ethics.
2. Why are works produced by artificial intelligence not entitled to copyright
The Berne Convention for the Protection of Literary and Artistic Works[vi], enacted in Berne, Switzerland, on September 9, 1886, is a worldwide recognized copyright protection convention and a member of which the mainland is a member. Article 1 of the Convention begins by stating that States Members of the Convention protect the rights of authors over their literary and artistic works. Article 3 of the Convention stipulates that there are two types of authors - citizens of States members of the Convention and non-citizens of States members of the Convention. Therefore, the author can only be a natural person, and ChatGPT and other artificial intelligence software cannot become authors in the sense of copyright. This is why copyright registration agencies in various countries have so far not accepted AI registration as the creator of works.
In addition to natural person works, the mainland Copyright Law also stipulates legal person works, but works created by ChatGPT are not legal person works. First of all, legal person works also require natural persons to create them. Secondly, there are three elements for the composition of a legal person work:
1. Presided over by a legal person or unincorporated organization;
2. Create on behalf of the will of legal persons or unincorporated organizations;
3. The responsibility shall be borne by legal persons or unincorporated organizations.
If the author wants ChatGPT to write an article, the host of the creation of the work is the author, the ChatGPT program is the creation on behalf of the author's will, if the article has infringement, this responsibility should also be borne by the author, these are not related to the creator of the article ChatGPT program, so this article written by ChatGPT does not constitute a legal person work. As for the author, I only gave the conditions for generating the article, and did not create the article, so I do not enjoy the copyright of the article.
The act of artificial intelligence creation is more similar to the commissioned creation under the Copyright Law, as long as the client enters the creation requirements after completion, the specific creation is completed by artificial intelligence, so the artificial intelligence program is the real creator of the work. According to the Copyright Law, under normal circumstances, the copyright of the commissioned creation of a work belongs to the entrusted party, which is also the confidence for artificial intelligence companies to claim that artificial intelligence companies enjoy copyright in works when chatting with the author.
According to Article 4.2 of the User Agreement, any content generated by the user when using ChatGPT belongs to the user's intellectual property. But ChatGPT can use this content without paying any fees to users. The author understands that if they stipulate that the intellectual property rights of the works they produce belong to ChatGPT, on the one hand, they will offend users; On the other hand, rights and responsibilities are one, who owns intellectual property rights, who is responsible for the content, if there is infringement or disputed content, such as copyright infringement, reputation infringement or more lethal politically incorrect speech, OpenAI as a small body of startups may not necessarily bear.
III. How to protect the rights of users if their works created using artificial intelligence are stolen by others?
However, works created by artificial intelligence are not protected by the Copyright Law, so if a user creates a work with artificial intelligence and is stolen by others, how can he protect his rights and interests?
First of all, the case [vii], Tencent developed a computer software called Dreamwriter, which can automatically write manuscripts, generating 300,000 articles every year, and a website reprinted articles automatically generated by artificial intelligence on Tencent Securities' website, which was sued by Tencent on the grounds of infringement. The Nanshan District Court held that the articles technically "generated" by the Dreamwriter software all met the conditions for the protection of written works under the Copyright Law, and were legal person works created under the auspices of the plaintiff. Therefore, the award was 1,500 yuan, and the judgment has now taken effect.
For the judgment of this case, the author does not agree very much, the expression of the article is completed by artificial intelligence, as mentioned above, even if there is originality, its creative subject is not a natural person, it should not enjoy copyright protection. The verdict was made more than three years ago, when AI creations were still rare, so it didn't hurt to protect them. And now that ChatGPT and various artificial intelligence software have come out, various articles, art works, videos, music, and programs created by artificial intelligence have emerged in large numbers, and at this time, let the court issue a similar judgment, it is estimated that the court will still think twice.
Some readers may ask, if you cannot enjoy copyright protection, how should you deal with such rights and interests if they are infringed? The Anti-Unfair Competition Law may be used. For example, in this case, if the defendant used hundreds or thousands of articles generated by Tencent's artificial intelligence on a large scale, Tencent could rely on Article 2 of the Anti-Unfair Competition Law that business operators should be honest and abide by business ethics in their business activities to protect their legitimate rights and interests.
At this time, the prosecution logic should be: Tencent paid a cost for the development of artificial intelligence software, the content of the article published was legal, and the business model was legitimate, while the defendant misappropriated the article on a large scale without permission and did not pay the cost of creation, which is a typical dishonest and contrary to recognized business ethics to disrupt the order of competition.
However, the threshold for protection under the Anti-Unfair Competition Law is relatively high, and only batch and large-scale copying of content will fall within the scope of legal regulation, and if the defendant in this case only uses a small number of AI-generated articles, such as less than ten, it is not enough to cross the threshold of violating the Anti-Unfair Competition Law, that is, if the defendant captures a small amount of content and publishes it without permission, Tencent cannot protect its rights.
Some readers may ask: artificial intelligence is developing so fast, the works it produces will be used in all walks of life, and soon a very large industry will be formed, if there is no copyright protection, weak protection for works created by artificial intelligence, will it have a great impact on the development of the industry? Wouldn't it be detrimental to the protection of intellectual property rights?
The author sees this as follows: intellectual property rights, including copyright, are artificially created rights, the purpose of the intellectual property system established by society is to encourage innovation, the essence is that everyone in the whole society has ceded part of the right to protect the creations and inventions of others, considering the hard work of the author of the creation and the inventor and the contribution to the promotion of human civilization, this transfer is reasonable.
The emergence of technological breakthroughs in ChatGPT and other artificial intelligence programs has lowered the threshold for creation a lot, and humans can get articles, music, pictures, and videos that originally took a long time and a lot of energy to create. For such low-cost creative results, society does not need to cede too many rights for protection, which actually means that works created by artificial intelligence can be used and disseminated more freely, and everyone can enjoy more benefits brought by the progress of artificial intelligence science and technology. More dissemination and less legal restrictions, freer competition will allow the industry to flourish, which is certainly better than a monopoly on the content industry by a few companies or creators.
Finally, at the beginning of the birth of the Internet, if the telecom companies said that because they wanted to use their networks, all the data transmitted on the Internet was owned by the telecommunications companies, then there would not be the Internet industry that is now in full swing. The same is true of the artificial intelligence industry, if the intellectual property rights created by the artificial intelligence company are owned by the artificial intelligence company, the development of the industry will not be very strong, only the artificial intelligence company like the telecommunications company, only charge network usage fees, let the work spread freely, then the artificial intelligence created works can benefit the public more and better, but also make the industry have long-term development momentum.
This article was first published in Interface News, author: You Yunting, senior partner of Shanghai Dabang Law Firm, intellectual property lawyer. The views expressed in this article are those of the author only.
[i] https://m.21jingji.com/article/20230310/herald/4fcc97ecba8eb2a9304546a99c5c3f45.html
[ii] https://www.pipicats.com/cwmzx/20923.html
[iii] https://baike.baidu.com/item/robots%E5%8D%8F%E8%AE%AE/2483797?fr=aladdin
[iv] http://wncyip.com/UploadFiles/20190605/20190605000431613.pdf
[in]https://zh.wikipedia.org/wiki/Wikipedia:GNU%E8%87%AA%E7%94%B1%E6%96%87%E6%A1%A3%E8%AE%B8%E5%8F%AF%E8%AF%81%E6%96%87%E6%9C%AC
[vi] https://www.wipo.int/wipolex/zh/text/283701
[vii] https://baijiahao.baidu.com/s?id=1665086063539756507&wfr=spider&for=pc