Generative Artificial Intelligence (“GenAI”) is a cutting-edge technology rapidly transforming workflows and business models across various industries. Alongside its widespread application, the legal issues surrounding GenAI have become increasingly prominent. This article will analyze the legal status of generative artificial intelligence in China by examining the “Interim Measures for the Management of Generative Artificial Intelligence Services” and relevant cases.
Legal Framework for Generative Artificial Intelligence
Currently, China’s primary laws and regulations regarding artificial intelligence (AI) include the following:
1) Interim Measures for the Management of Generative Artificial Intelligence Services This measure mainly regulates the rights and obligations of providers and users of generative AI services. It stipulates requirements for safety, privacy protection, and algorithm management. For example, providers must clarify their services’ applicable population and purposes, protect user input information and usage records, label generated content, address illegal content, and establish complaint and reporting mechanisms. 2) Cybersecurity Law This law comprehensively addresses network security, data security, and personal information protection, serving as a fundamental law for managing AI services. It requires internet service providers to protect user information and prevent data leakage and misuse. 3) Data Security Law This law stipulates regulations on the security management, circulation, and development of data activities, aiming to ensure the legality and compliance of data processing activities, including the collection, processing, and storage of AI-related data. 4) Personal Information Protection Law
This law comprehensively regulates personal information processing activities, clarifying data processors’ obligations and individuals’ rights, applicable to AI technologies involving personal data processing.
5) Regulations on the Management of Deep Synthesis of Internet Information Services These regulations manage algorithm recommendation services with public opinion attributes or social mobilization capabilities, requiring security assessments, algorithm filing, and emphasizing algorithm transparency and interpretability.
These laws and regulations collectively form China’s legal framework for AI technologies, balancing technological development with legal supervision, protecting user rights, and ensuring data security and technology transparency.
Additionally, China is actively promoting the legislative process for artificial intelligence technology. For example, the “Artificial Intelligence Law” draft has been included in the State Council’s 2023 legislative work plan aiming to improve further the legal framework and regulatory mechanisms for artificial intelligence. The 2024 government work report also mentioned the “Artificial Intelligence+” initiative for the first time, emphasizing artificial intelligence’s deep integration and technological innovation across various industries.
Currently, the “Interim Measures for the Management of Generative Artificial Intelligence Services” (“Interim Measures”) is an essential legal document in China’s GenAI field, issued by the Cyberspace Administration of China on May 23, 2023, and effective from August 15, 2023. The main contents of the “Interim Measures” include the following key provisions:
1) Legality of Data and User Protection
Article 7 of the Interim Measures provides that Service providers must ensure the legality of training data and not infringe on intellectual property rights and personal information. This provision requires that the training data source for generative AI must be legal, ensuring that the rights of others, especially intellectual property rights and personal privacy, are not violated. This regulation aims to prevent legal disputes arising from AI systems’ unauthorized use of data.
Besides, Article 11 of the Interim Measures provides that Providers must protect user input information and usage records during the service process and not illegally retain or provide user information to others. This provision emphasizes the protection of user data, requiring service providers not to store or disclose user input information and usage records illegally. The provision helps enhance users’ trust in generative AI services.
2) Algorithm Transparency and Content Legality
Article 8 of the Interim Measures states that Providers should formulate specific, operable labeling rules, train labeling personnel, and conduct sample checks on the correctness of labeling content. This provision ensures the transparency and accuracy of the training data labeling process for generative AI systems, helping to improve the quality and reliability of AI-generated content.
Article 8 also states that Providers must clearly state and publicly disclose the applicable population, occasions, and purposes of their services and take appropriate measures to prevent users from becoming overly dependent on or addicted to generated content. This provision aims to prevent users, especially minors, from becoming dependent on or addicted to generative AI content by specifying the usage scope and providing appropriate warnings.
3) User Identity Verification
Article 9 of the Interim Measures states that users of generative AI services must provide accurate identity information. This provision requires providers to conduct factual identity verification of users to prevent legal and security issues arising from anonymous use. The provision is significant for tracking and regulating illegal activities.
Case Analysis Involving Generative Artificial Intelligence
1) Infringement of Works Created by Generative Artificial Intelligence
With the rapid development of AI-generated content (AIGC) technology, AI can mimic human creative styles and generate content such as images, texts, videos, and audio at unprecedented speeds and scales. This technology significantly lowers the creative threshold, allowing ordinary users to generate works of particular artistic value through simple prompts. However, this also raises intense debates about the ownership of creative rights, the originality of works, and the scope of copyright protection. On the one hand, creators using AI tools argue that their intellectual effort during the creation process entitles them to corresponding rights over the generated content. On the other hand, using AI to create content involves several parties such as the AI creation platform (AI prominent model provider), creators, providers of material used for AI ample model training, and other aspects. Therefore, it is difficult to identify the copyright.
In April 2024, the creator of the AI-generated movie trailer “Mountain and Sea Mirror,” known as “Xianren Yikun,” filed a lawsuit against the studio “Wenke Creation” The accusation claims that Wenke Creation used AI technology to replicate and distribute a highly similar infringing video without permission. Xianren Yikun filed the lawsuit on April 11, 2024, and the court heard the first hearing on May 15, 2024. It is the first case in China where copyright for AI-created audiovisual content is being claimed, although it is not the first copyright dispute involving AI-generated content.
Previously, on April 23, 2024, the Beijing Internet Court ruled on the nation’s first AI voice infringement case. The plaintiff, Ms. Yin, won the case after AI technology imitated and used her voice commercially without her consent. The court ordered the defendant company to compensate Ms. Yin RMB 250,000 for economic losses and to issue a public apology. This case established that, under the condition of recognizability, the protection scope of natural persons’ voice rights extends to AI-generated voices. The recognizability of AI-generated voices should be assessed considering the user’s context and whether ordinary listeners in the relevant field can identify the voice.
Additionally, in January 2024, China’s first AI-generated image copyright infringement case was concluded (2023 Beijing 0491 Civil First Instance No. 11279). The plaintiff, Li Yunkai, who created images using the Stable Diffusion model, filed a lawsuit after his images were infringed upon. Following five hearings, the court determined that Li Yunkai’s images possessed “originality” and met the definition of a work of art, thus qualifying for copyright protection. The court ruled that the defendant had violated Li Yunkai’s rights to information network dissemination and attribution and was liable for infringement. The defendant was ordered to publicly apologize and compensate for economic losses amounting to RMB 500.
The core issues in this case were whether AI-generated images are “works” and what type of works they constitute, as well as whether AI-generated images possess legal copyright. The presiding judge emphasized that when using AI models to generate images, there is no question of determining who the creator is; fundamentally, humans use tools for creation. The judge noted that human intellectual input, not the AI models, drives the creative process. By correctly applying copyright laws and using appropriate legal measures, more people can be encouraged to create using the latest tools, fostering artistic creation and AI technology development. The ruling acknowledged that images designed by users and generated with AI models, demonstrating originality and intellectual input, qualify as artistic works protected by copyright.
2) Privacy Infringement by Generative Artificial Intelligence Services
In July 2023, the “MiaoYa Camera” application gained widespread attention. Contrary to its name, this application is not a traditional “camera” but an AI-generated content (AIGC) application that regenerates user-uploaded photos. The app requires users to upload at least 20 frontal photos of themselves and pay a fee of RMB 29.9 (with a limited-time discount of RMB 9.9) to create a digital avatar. The user agreement stipulates that users must grant MiaoYa Camera a perpetual, irrevocable, transferable, sub-licensable, accessible, and non-exclusive license to use their content globally, including in virtual spaces like the metaverse. This license allows MiaoYa Camera to use the user’s content in any form, medium, or technology, whether currently known or developed in the future. MiaoYa Camera claims that this clause is necessary to provide, ensure, and improve its services, including using AI-generated content as retraining data.
Within ten days of its launch, the Shanghai Consumer Protection Committee accused the application of potential infringement. The user agreement required users to grant an “irrevocable” license, which conflicts with the Personal Information Protection Law, which states that users can revoke their consent. Additionally, the agreement’s terms of perpetual, transferable, and sub-licensable rights raise further infringement concerns. According to Article 11 of the “Interim Measures for the Administration of Generative Artificial Intelligence Services,” providers of generative AI services must legally protect users’ input information and usage records, avoid collecting unnecessary personal information, refrain from unlawfully retaining identifiable user information and usage records, and not illegally share this information with others. As a result, MiaoYa Camera had to revise its user agreement.
Many enterprises are leveraging generative AI to provide services or enhance productivity and efficiency. However, businesses must ensure the legality of data use and user privacy protection when using generative AI. Adhering to relevant laws and regulations is crucial to ensure data is used legally and compliantly.
Conclusion
China’s legal framework for generative artificial intelligence is still in development—the Interim Measures provide initial legal norms for the field. However, with rapid technological advancements, existing regulations need continuous updates and improvements to address emerging legal challenges. Measures such as legislative clarifications on the copyright ownership of AI-generated content, detailed data processing and protection standards, and comprehensive technical review and certification mechanisms are necessary to ensure the legal and compliant application of generative AI across various industries. These improvements will foster a more complete and robust legal system for the generative artificial intelligence sector.
Legal Framework for Generative Artificial Intelligence
Currently, China’s primary laws and regulations regarding artificial intelligence (AI) include the following:
1) Interim Measures for the Management of Generative Artificial Intelligence Services This measure mainly regulates the rights and obligations of providers and users of generative AI services. It stipulates requirements for safety, privacy protection, and algorithm management. For example, providers must clarify their services’ applicable population and purposes, protect user input information and usage records, label generated content, address illegal content, and establish complaint and reporting mechanisms. 2) Cybersecurity Law This law comprehensively addresses network security, data security, and personal information protection, serving as a fundamental law for managing AI services. It requires internet service providers to protect user information and prevent data leakage and misuse. 3) Data Security Law This law stipulates regulations on the security management, circulation, and development of data activities, aiming to ensure the legality and compliance of data processing activities, including the collection, processing, and storage of AI-related data. 4) Personal Information Protection Law
This law comprehensively regulates personal information processing activities, clarifying data processors’ obligations and individuals’ rights, applicable to AI technologies involving personal data processing.
5) Regulations on the Management of Deep Synthesis of Internet Information Services These regulations manage algorithm recommendation services with public opinion attributes or social mobilization capabilities, requiring security assessments, algorithm filing, and emphasizing algorithm transparency and interpretability.
These laws and regulations collectively form China’s legal framework for AI technologies, balancing technological development with legal supervision, protecting user rights, and ensuring data security and technology transparency.
Additionally, China is actively promoting the legislative process for artificial intelligence technology. For example, the “Artificial Intelligence Law” draft has been included in the State Council’s 2023 legislative work plan aiming to improve further the legal framework and regulatory mechanisms for artificial intelligence. The 2024 government work report also mentioned the “Artificial Intelligence+” initiative for the first time, emphasizing artificial intelligence’s deep integration and technological innovation across various industries.
Currently, the “Interim Measures for the Management of Generative Artificial Intelligence Services” (“Interim Measures”) is an essential legal document in China’s GenAI field, issued by the Cyberspace Administration of China on May 23, 2023, and effective from August 15, 2023. The main contents of the “Interim Measures” include the following key provisions:
1) Legality of Data and User Protection
Article 7 of the Interim Measures provides that Service providers must ensure the legality of training data and not infringe on intellectual property rights and personal information. This provision requires that the training data source for generative AI must be legal, ensuring that the rights of others, especially intellectual property rights and personal privacy, are not violated. This regulation aims to prevent legal disputes arising from AI systems’ unauthorized use of data.
Besides, Article 11 of the Interim Measures provides that Providers must protect user input information and usage records during the service process and not illegally retain or provide user information to others. This provision emphasizes the protection of user data, requiring service providers not to store or disclose user input information and usage records illegally. The provision helps enhance users’ trust in generative AI services.
2) Algorithm Transparency and Content Legality
Article 8 of the Interim Measures states that Providers should formulate specific, operable labeling rules, train labeling personnel, and conduct sample checks on the correctness of labeling content. This provision ensures the transparency and accuracy of the training data labeling process for generative AI systems, helping to improve the quality and reliability of AI-generated content.
Article 8 also states that Providers must clearly state and publicly disclose the applicable population, occasions, and purposes of their services and take appropriate measures to prevent users from becoming overly dependent on or addicted to generated content. This provision aims to prevent users, especially minors, from becoming dependent on or addicted to generative AI content by specifying the usage scope and providing appropriate warnings.
3) User Identity Verification
Article 9 of the Interim Measures states that users of generative AI services must provide accurate identity information. This provision requires providers to conduct factual identity verification of users to prevent legal and security issues arising from anonymous use. The provision is significant for tracking and regulating illegal activities.
Case Analysis Involving Generative Artificial Intelligence
1) Infringement of Works Created by Generative Artificial Intelligence
With the rapid development of AI-generated content (AIGC) technology, AI can mimic human creative styles and generate content such as images, texts, videos, and audio at unprecedented speeds and scales. This technology significantly lowers the creative threshold, allowing ordinary users to generate works of particular artistic value through simple prompts. However, this also raises intense debates about the ownership of creative rights, the originality of works, and the scope of copyright protection. On the one hand, creators using AI tools argue that their intellectual effort during the creation process entitles them to corresponding rights over the generated content. On the other hand, using AI to create content involves several parties such as the AI creation platform (AI prominent model provider), creators, providers of material used for AI ample model training, and other aspects. Therefore, it is difficult to identify the copyright.
In April 2024, the creator of the AI-generated movie trailer “Mountain and Sea Mirror,” known as “Xianren Yikun,” filed a lawsuit against the studio “Wenke Creation” The accusation claims that Wenke Creation used AI technology to replicate and distribute a highly similar infringing video without permission. Xianren Yikun filed the lawsuit on April 11, 2024, and the court heard the first hearing on May 15, 2024. It is the first case in China where copyright for AI-created audiovisual content is being claimed, although it is not the first copyright dispute involving AI-generated content.
Previously, on April 23, 2024, the Beijing Internet Court ruled on the nation’s first AI voice infringement case. The plaintiff, Ms. Yin, won the case after AI technology imitated and used her voice commercially without her consent. The court ordered the defendant company to compensate Ms. Yin RMB 250,000 for economic losses and to issue a public apology. This case established that, under the condition of recognizability, the protection scope of natural persons’ voice rights extends to AI-generated voices. The recognizability of AI-generated voices should be assessed considering the user’s context and whether ordinary listeners in the relevant field can identify the voice.
Additionally, in January 2024, China’s first AI-generated image copyright infringement case was concluded (2023 Beijing 0491 Civil First Instance No. 11279). The plaintiff, Li Yunkai, who created images using the Stable Diffusion model, filed a lawsuit after his images were infringed upon. Following five hearings, the court determined that Li Yunkai’s images possessed “originality” and met the definition of a work of art, thus qualifying for copyright protection. The court ruled that the defendant had violated Li Yunkai’s rights to information network dissemination and attribution and was liable for infringement. The defendant was ordered to publicly apologize and compensate for economic losses amounting to RMB 500.
The core issues in this case were whether AI-generated images are “works” and what type of works they constitute, as well as whether AI-generated images possess legal copyright. The presiding judge emphasized that when using AI models to generate images, there is no question of determining who the creator is; fundamentally, humans use tools for creation. The judge noted that human intellectual input, not the AI models, drives the creative process. By correctly applying copyright laws and using appropriate legal measures, more people can be encouraged to create using the latest tools, fostering artistic creation and AI technology development. The ruling acknowledged that images designed by users and generated with AI models, demonstrating originality and intellectual input, qualify as artistic works protected by copyright.
2) Privacy Infringement by Generative Artificial Intelligence Services
In July 2023, the “MiaoYa Camera” application gained widespread attention. Contrary to its name, this application is not a traditional “camera” but an AI-generated content (AIGC) application that regenerates user-uploaded photos. The app requires users to upload at least 20 frontal photos of themselves and pay a fee of RMB 29.9 (with a limited-time discount of RMB 9.9) to create a digital avatar. The user agreement stipulates that users must grant MiaoYa Camera a perpetual, irrevocable, transferable, sub-licensable, accessible, and non-exclusive license to use their content globally, including in virtual spaces like the metaverse. This license allows MiaoYa Camera to use the user’s content in any form, medium, or technology, whether currently known or developed in the future. MiaoYa Camera claims that this clause is necessary to provide, ensure, and improve its services, including using AI-generated content as retraining data.
Within ten days of its launch, the Shanghai Consumer Protection Committee accused the application of potential infringement. The user agreement required users to grant an “irrevocable” license, which conflicts with the Personal Information Protection Law, which states that users can revoke their consent. Additionally, the agreement’s terms of perpetual, transferable, and sub-licensable rights raise further infringement concerns. According to Article 11 of the “Interim Measures for the Administration of Generative Artificial Intelligence Services,” providers of generative AI services must legally protect users’ input information and usage records, avoid collecting unnecessary personal information, refrain from unlawfully retaining identifiable user information and usage records, and not illegally share this information with others. As a result, MiaoYa Camera had to revise its user agreement.
Many enterprises are leveraging generative AI to provide services or enhance productivity and efficiency. However, businesses must ensure the legality of data use and user privacy protection when using generative AI. Adhering to relevant laws and regulations is crucial to ensure data is used legally and compliantly.
Conclusion
China’s legal framework for generative artificial intelligence is still in development—the Interim Measures provide initial legal norms for the field. However, with rapid technological advancements, existing regulations need continuous updates and improvements to address emerging legal challenges. Measures such as legislative clarifications on the copyright ownership of AI-generated content, detailed data processing and protection standards, and comprehensive technical review and certification mechanisms are necessary to ensure the legal and compliant application of generative AI across various industries. These improvements will foster a more complete and robust legal system for the generative artificial intelligence sector.