China is a country where artificial intelligence (AI) is rapidly developing and widely used in different spheres. The size of Chinese AI market is expected to reach more than 34 billion USA dollars by the end of 2024. China goes second after the USA in the ranking of top 20 AI countries (countries with the highest research capacity). Thus, there is no surprise that China is one of the first countries in the world which adopted legal regulation of AI.
During 2023 Chinese government made major efforts to create legal basis for AI training and provision of services. In this review we are going to study the regulation itself and the most important court decisions to the date.
Generative AI Regulation
In August 2023 “Interim Measures for the Management of Generative Artificial Intelligence Services”[1] (The Interim Measures, Regulation) came into force in China. In general, the Interim Measures impose certain obligations on the providers of the services, connected with AI. Providers must:
- Provide secure services;
- Offer users detailed and clear license agreements which describes all the features of the AI service and rights and obligations of the parties;
- Post a feedback form on the platform, provide quick response to users’ requests;
- Keep users’ data confidential;
- Delete or edit personal data on users’ requests;
- Delete illegal content immediately after revealing it;
- Where providers discover that users are using generative AI services to engage in illegal activities, they shall limit functions or suspend provision of the service and report to the relevant authorities.
Both providers and developers of the AI services must comply with provisions of personal data laws (use personal information on legal grounds, such as consent of the data subject) and copyright laws; respect honor, dignity and goodwill rights.
Before providing services to the users, developers of the AI shall train AI, using data for training in accordance with law. They should increase the quality of training data, including accuracy, objectivity and diversity of such data. Measures should be taken to prevent the creation of any kind of discrimination (when training AI and providing services).
So, both process of providing services and process of training AI are now regulated by law in China. Requirements for the service providers are strict: they should monitor and immediately delete any illegal content, monitor whether the services are used for illegal purposes and even employ measures to prevent minor users from addiction to generative AI services.
Deepfake Regulation
Moreover, earlier last year Chinese government adopted deepfake regulation (Deep Synthesis Provisions (the Provisions), came into force in January 2023). The deep synthesis is determined as “technology utilizing generative and/or synthetic algorithms, such as deep learning and virtual reality, to produce text, graphics, audio, video, or virtual scenes”. The definition is very broad, that may be seen as a flaw of the existing regulation. However, such definition may allow the Provisions to adapt to any new form of deepfake technologies in the future.
The main requirements to the providers of deepfake services are the following:
- Comply with data security laws;
- Establish criteria and processes to determine false information;
- Monitor fake news created with the use of deepfake services and report to relevant authorities;
- Implement real-identity authentication system for users;
- Conduct regular security assessment.
As we see, all these measures are aimed to make the process of using deep synthesis services transparent, prevent spreading of false information and deanonymize users of the services. These provisions can prevent using deepfake technologies in illegal activities.
Draft AI Law
Chinese government also announced working on the Draft AI Law that would provide complex regulation. At the moment the text of the Draft Law is available.
The draft contains crucial provisions regarding copyright. It establishes that based on the extent of the user’s contribution to the final version of the content, AI generated content may be object of copyright or patent rights.
Other important provisions are connected with the users’ security. The draft specifies system of risk assessment that should be performed before giving users access to AI technology and measures to provide security of AI technologies.
Case Law
While lawmakers in China attempt to create legal framework around AI services, judges already have to consider cases and make decisions. In the recent months 2 important decisions on AI cases were issued by the Chinese courts. These cases reflect existing regulation of AI in China and may also influence the Draft AI Law and other future laws[2].
Copyright to AI generated work
In November 2023 the Beijing Internet Court of China issued a milestone decision on case involving copyright infringement.
The plaintiff created a picture of woman using Stable Diffusion (a text-to-image generative AI model). He gave all the necessary prompts to AI model to create the image, and also modified the picture, using app settings. After that he posted the image on social media platform, tagging it as #AI Drawing. After finding out that other user published the image on another platform without seeking consent, the alleged author of the image turned to court.
The Beijing court had to determine 1) whether copyright to AI generated image exists at all; 2) if the answer to the first question is positive, who owns such copyright.
When making decision, the court considered 2 characteristics of copyright: intellectual achievement of the author and originality. Court ruled that in that case intellectual input was enough, because the plaintiff used detailed prompts to generate the image. As for the originality, the court established that using prompts, settings and filters the plaintiff created a unique image, no one else would create. Accordingly, the court decided that the image in question was artistic work protected by copyright.
Further, court ruled that AI itself cannot be the author, neither can the designer of Stable Diffusion, who only created the AI model. So, the user who entered prompts to generate the picture (the plaintiff) was considered the author and copyright owner.
Right to Voice and AI
In April 2024 the same Chinese court recognized that non-authorized use of person’s voice for training AI is illegal.
The plaintiff in this case was a dubbing artist, who discovered that his voice was widely used on Internet-platforms, including those that allow text-to-voice converting. It turned out that a software developing company used his voice to train AI model.
In-court examination was held, during that it was proved that the AI voice, though modified, was still very similar to the voice of the plaintiff. Court used such characteristics as timbre, intonation, pronunciation style to determine the similarity. Though the software company received the recordings with plaintiff’s voice from a media company which had copyright to such recordings, the court decided that the media company was still not entitled to authorize others to use plaintiff’s voice for AI training (debatable conclusion).
Thus, court ruled that unauthorized use of a person’s voice constitutes a violation of personal right to voice and established compensation in the amount of 250 000 RMB (around 35 000 USD).
This landmark decision protected rights of a person, who’s voice was widely used on different platforms without any permission. Court in this case recognized such category as “personal right to voice”, that can be used in further decisions on similar cases.
Conclusion
Artificial Intelligence is a topic, connected with a lot of uncomfortable and complicated issues, from “will the robots replace us?” to “will everybody be able to fake my photos?” While in many countries there is no specific regulation of AI, Chinese lawmakers and law-enforcement authorities already prepared some answers to these questions. Chinese existing regulation pays special attention to the protection of personal data from the stage of training AI, copyright protection, revealing fake information, transparency and security of the services.
All in all, China is one of the world’s pioneers in the sphere of AI legal regulation. As in Russia there is no special regulation of AI yet, when creating such legal basis we may borrow some tips from China. We will continue monitoring news of Chinese legislation regarding AI and write about the AI Law when it comes into force.
[1] English translation: https://www.chinalawtranslate.com/en/generative-ai-interim/
[2] Though China does not have case law system, like USA or Great Britain, judicial practice may affect legislative decisions.