The copyright of AI generated content is not clearly defined

Intern journalist Pei Chenwei, reporter of our newspaper Wu ChunxinSince OpenAI launched the generative large language model ChatGPT, generative artificial intelligence has emerged endlessly both domestically and internationally. One of the most concerning issues is whether artificial intelligence generated content (AIGC) has copyright?"The object of copyright protection is works

Intern journalist Pei Chenwei, reporter of our newspaper Wu Chunxin

Since OpenAI launched the generative large language model ChatGPT, generative artificial intelligence has emerged endlessly both domestically and internationally. One of the most concerning issues is whether artificial intelligence generated content (AIGC) has copyright?

"The object of copyright protection is works." On June 16, at the Second East Lake Forum on Copyright Industry Innovation and Intellectual Property Protection held in Wuhan, Hubei Province, Cao Xinming, Vice President of China Intellectual Property Law Research Association and Professor of Zhongnan University of Economics and Law, said that some AI products are not works in the sense of the Copyright Law of the China (hereinafter referred to as the Copyright Law).

Cao Xinming explained that works within the meaning of copyright law should meet the following four conditions: they are created by humans, have originality, are expressions that contain certain ideological content, and are not excluded by copyright laws such as laws and regulations, general data tables, and formulas. At present, there are three forms of AIGC, which are completely independently created by artificial intelligence, assisted by natural persons, and generated based on prompts input by natural persons.

Cao Xinming said that only one of the three forms mentioned above is directly involved by humans, in which case AIGC may have copyright. It is problematic to define AIGC as a 'work' in the remaining two forms. This is because artificial intelligence does not have independent thinking and cannot independently 'create', let alone possess copyright.

Chen Bing, Vice Dean of the Law School of Nankai University and Director of the Competition Law Research Center, said that from the current legislative situation in various countries, Japan, Australia, the United Kingdom, the United States and other countries have not granted the qualification of artificial intelligence as a civil subject, so AIGC is not "copyrighted".

Although the legal definition of AIGC copyright issues is not yet clear, it is still necessary to be vigilant about the risk of infringement during the use of generative artificial intelligence. China has relevant cases in this regard. Chen Bing said that in these cases, the court made a judgment on the generation process of AIGC and clarified whether the generated products were given benefits. This requires a specific analysis of specific issues, but one thing can be clear: the simple form selection of AIGC by humans is not sufficient to constitute originality in copyright law.

Generative artificial intelligence involves the entire process from data capture to content generation. During this process, AIGC, as the final product of generative artificial intelligence, has a vague definition of copyright ownership, and the data captured during its training process also carries the risk of infringement.

There are hundreds of billions of elements and data in artificial intelligence now, many of which are copyrighted materials. It is not yet clear how to protect the rights of the copyright owners of these materials, "Cao Xinming said.

Chen Bing believes that at present, the China National Intellectual Property Administration, Shenzhen, Shanghai and other places have begun to study the confirmation of data rights, but how to confirm the rights needs to be answered in practice. The establishment of copyright is to encourage more innovation, "he said. If data and AIGC are not verified, it will lead to high feeding costs for generative artificial intelligence developers, thereby inhibiting innovation. On the other hand, due to the significant aggregation effect and Matthew effect effect of the large model, AI developers are prone to data monopoly risks when training AI.

In response to the above issues, Chen Bing believes that we should approach AIGC innovation rationally, explore the application boundaries of AIGC, optimize the system design for data crawling, explore scenario based and refined algorithm governance mechanisms for classification and grading, and consolidate the corresponding legal responsibilities of algorithm developers and applications. Regarding the risk of data monopoly, we need to leverage the government's agile and precise regulatory role in the market, promote the 'open source' development of enterprises, and establish a diversified regulatory mechanism, "Chen Bing said at the same time.

In addition, Cao Xinming reminds that generative artificial intelligence may not only infringe copyright. For example, the use of artificial intelligence to imitate their voice, action, posture, gesture, or even use artificial intelligence to "steal faces" without the permission of others violates citizens' Personality rights.

To address these issues, one can choose to install 'guardrail technology' to constrain artificial intelligence, while avoiding attacks on large models by certain users, "Cao Xinming said.

Disclaimer: The content of this article is sourced from the internet. The copyright of the text, images, and other materials belongs to the original author. The platform reprints the materials for the purpose of conveying more information. The content of the article is for reference and learning only, and should not be used for commercial purposes. If it infringes on your legitimate rights and interests, please contact us promptly and we will handle it as soon as possible! We respect copyright and are committed to protecting it. Thank you for sharing.(Email:[email protected])