New Regulations on Generative AI in China
An Initial Look
On Thursday, China unveiled a revised regulatory framework for generative AI technology, impacting chatbots like OpenAI's ChatGPT and Google's Bard. The rules can be found at http://www.cac.gov.cn/2023-07/13/c_1690898327029107.htm
These regulations will be administered by the Cyberspace Administration of China (CAC), the nation's principal internet regulator. However, the new rules exempt generative AI research and technologies specifically developed for use abroad.
These regulations also mandate developers to register their algorithms with the Chinese government. Furthermore, if their services are regarded as having a "social mobilization ability" that could potentially sway public opinion, they must undergo a "security assessment". This seems like an extension of existing Chinese efforts to control online discourse.
The new law emphasizes the requirement to "adhere to core socialist values". It details various prohibited uses of generative AI, ranging from those aimed at citizen protection—such as a ban on promoting terrorism and spreading obscene pornography—to those designed to consolidate government authority over this emerging technology. For instance, it prohibits the use of generative AI to subvert state power, damage the country's image, and undermine national unity.
The new rules were formulated by the CAC but were also approved by seven other agencies including the Ministry of Education, the Ministry of Public Security, and the State General Administration of Radio and Television. These developments suggest that China's regulation of generative AI will continue to evolve as the technology itself advances.
Here is an initial English translation (via Claude 2) of the new rules:
Interim Measures for the Administration of Generative AI Services
The Office of the Central Cyberspace Affairs Commission and the National Internet Information Office have approved the issuance of Order No. 15 “Interim Measures for the Administration of Generative AI Services” at the 12th office meeting on May 23, 2023. It has also been agreed upon by the National Development and Reform Commission, Ministry of Education, Ministry of Science and Technology, Ministry of Industry and Information Technology, Ministry of Public Security, and National Radio and Television Administration. The measures are hereby promulgated and will come into force on August 15, 2023.
Director of the Office of the Central Cyberspace Affairs Commission, Zhuang Rongwen
Director of the National Development and Reform Commission, Zheng Jiele
Minister of Education, Huai Jinpeng
Minister of Science and Technology, Wang Zhigang
Minister of Industry and Information Technology, Xin Zhonglong
Minister of Public Security, Wang Xiaohong
Director of the National Radio and Television Administration, Cao Shuhui
July 10, 2023
Chapter I General Provisions
Article 1 These measures are formulated in accordance with the Cybersecurity Law of the People's Republic of China, Data Security Law of the People's Republic of China, Personal Information Protection Law of the People's Republic of China, Law of the People's Republic of China on Progress of Science and Technology and other laws and regulations, in order to promote the healthy development and standardized application of generative AI, safeguard national security and public interests, and protect the lawful rights and interests of citizens, legal persons and other organizations.
Article 2 These measures shall apply to services that provide generated text, images, audio, videos and other content to the public in China using generative AI technologies (hereinafter referred to as generative AI services). Where the state has separate provisions on the use of generative AI services for activities such as news publishing, film and television production, and literary creation, those provisions shall prevail.
Research, development and application of generative AI technologies by industry organizations, enterprises, educational and scientific research institutions, public cultural institutions and relevant professional institutions that do not provide generative AI services to the domestic public shall not be subject to the provisions of these measures.
Article 3 The state adheres to the principle of balancing development and security, and combining innovation promotion and rule of law. Effective measures shall be taken to encourage innovative development of generative AI, generate positive and uplifting high-quality content, explore optimized application scenarios, and build an application ecosystem.
Article 4 Providing and using generative AI services shall abide by laws, regulations, respect social ethics and morals, and comply with the following provisions:
(1) Uphold socialist core values, and shall not generate content that incites subversion of state power, overthrow of the socialist system, endangers national security and interests, damages the national image, incites division of the country and undermines national unity and social stability, advocates terrorism and extremism, advocates ethnic hatred and discrimination, violence, obscenity and pornography, as well as false and harmful information prohibited by laws and regulations.
(2) Take effective measures in algorithm design, training data selection, model generation and optimization, service provision and other processes to prevent discrimination based on ethnicity, beliefs, nationality, region, gender, age, occupation, health status, etc.
(3) Respect intellectual property rights and business ethics, keep trade secrets, and shall not use algorithm, data, platform and other advantages to engage in monopolistic behavior and unfair competition.
(4) Respect the lawful rights and interests of others, shall not endanger the physical and mental health of others, and shall not infringe upon others' rights of portrait, reputation, honor, privacy, and personal information.
(5) Based on the characteristics of the service types, take effective measures to improve the transparency of generative AI services and the accuracy and reliability of generated content.
Chapter II Technological Development and Governance
Article 5 Encourage innovative applications of generative AI technologies in various industries and fields to generate positive, healthy and uplifting high-quality content, explore optimized application scenarios, and build application ecosystems. Support industry organizations, enterprises, educational and scientific research institutions, public cultural institutions and relevant professional institutions in carrying out collaboration in generative AI technology innovation, data resource construction, transformation and application, risk prevention and other aspects.
Article 6 Encourage independent innovation in core technologies such as generative AI algorithms, frameworks, chips and supporting software platforms, carry out fair and mutually beneficial international exchanges and cooperation, and participate in the formulation of relevant international rules for generative AI. Promote the construction of public training data resource platforms and generative AI infrastructure. Promote collaborative sharing of computing power resources and improve the efficiency of computing power resource utilization. Promote classified and orderly opening up of public data to expand high-quality public training data resources. Encourage the use of secure and trustworthy chips, software, tools, computing power and data resources.
Article 7 Generative AI service providers (hereinafter referred to as providers) shall carry out preprocessing, optimized training and other training data processing activities in accordance with the law, and comply with the following provisions:
(1) Use data and basic models from lawful sources;
(2) Where intellectual property rights are involved, do not infringe upon intellectual property rights lawfully enjoyed by others;
(3) Where personal information is involved, obtain individual consent or comply with other circumstances stipulated by laws and regulations;
(4) Take effective measures to improve the quality of training data and enhance the authenticity, accuracy, objectivity and diversity of training data;
(5) Other relevant provisions of laws and regulations such as the Cybersecurity Law of the People's Republic of China and the Data Security Law of the People's Republic of China, and relevant regulatory requirements of competent authorities.
Article 8 For data annotation conducted during the research and development of generative AI technologies, providers shall formulate clear, specific and operational annotation rules in line with the requirements of these measures; carry out annotation quality assessment and sample verification of annotation accuracy; provide necessary training for annotators to enhance awareness of law-abiding, and supervise and guide annotators to carry out annotation work in a standardized manner.
Chapter III Service Regulations
Article 9 Providers shall perform the responsibilities of network information content producers in accordance with the law and fulfill network information security obligations. Where personal information is involved, they shall perform the responsibilities of personal information processors and fulfill personal information protection obligations in accordance with the law.
Providers shall sign service agreements with registered users of their generative AI services (hereinafter referred to as users) to clarify the rights and obligations of both parties.
Article 10 Providers shall specify and disclose the applicable population, occasions, purposes and other aspects of their services, guide users to have a scientific and rational understanding of generative AI technologies and use them in accordance with the law, and take effective measures to prevent excessive reliance or addiction of minor users to generative AI services.
Article 11 Providers shall perform protection obligations for users' input information and usage records in accordance with the law. They shall not collect unnecessary personal information, illegally retain input information and usage records that can identify users, or illegally provide users' input information and usage records to others. Providers shall handle users' requests for access, copy, correction, supplementation and deletion of their personal information in a timely manner according to law.
Article 12 Providers shall identify generated image and video content in accordance with the Management Regulations on Deep Synthesis of Internet Information Services.
Article 13 Providers shall provide safe, stable and sustained services during service provision to ensure normal use by users.
Article 14 If providers discover illegal content, they shall take disposal measures such as stopping generation, stopping transmission, elimination in a timely manner, take rectification measures such as model optimization training, and report to relevant competent authorities. If providers discover users engaging in illegal activities using generative AI services, they shall take disposal measures such as warning, restricting functions, suspending or terminating services to users in accordance with the law and agreements, keep relevant records, and report to relevant competent authorities.
Article 15 Providers shall establish and improve complaint and reporting mechanisms, set up convenient complaint and reporting channels, publicize processing procedures and feedback time limits, and promptly accept, handle public complaints and reports and provide feedback on processing results.
Chapter IV Supervision and Inspection and Legal Liabilities
Article 16 Authorities such as cyberspace affairs, development and reform, education, science and technology, industry and information technology, public security, radio and television administration, press and publication shall strengthen the administration of generative AI services according to their respective responsibilities. Relevant competent authorities shall improve scientific supervision methods suited to the characteristics of generative AI technologies and their application services in relevant industries and fields, and formulate corresponding classified and hierarchical supervision rules or guidelines.
Article 17 Providers of generative AI services with public opinion attributes or social mobilization capabilities shall carry out safety assessments and complete the filing and change, cancellation filing procedures for algorithm recommendation management in accordance with relevant national regulations.
Article 18 Users who discover generative AI services that do not comply with laws, regulations and the provisions of these measures shall have the right to complain and report to relevant competent authorities.
Article 19 Relevant competent authorities shall carry out supervision and inspection of generative AI services according to their duties. Providers shall cooperate in accordance with the law, explain the sources, scale, types, annotation rules, algorithm mechanisms, etc. of the training data as required, and provide necessary technical support, data and other assistance.
Relevant agencies and personnel participating in security assessments and supervision and inspection of generative AI services shall keep confidential the state secrets, commercial secrets, personal privacy and personal information that they learn of when performing their duties according to law, and shall not disclose or illegally provide them to others.
Article 20 For generative AI services originating from outside the territory of the People’s Republic of China that fail to comply with the laws, administrative regulations and these Measures, the national cyberspace affairs authority shall notify relevant agencies to take technological measures and other necessary measures for disposal.
Article 21 Providers that violate the provisions of these Measures shall be punished by relevant competent departments in accordance with the Cybersecurity Law of the People’s Republic of China, Data Security Law of the People’s Republic of China, Personal Information Protection Law of the People's Republic of China, Law of the People's Republic of China on Progress of Science and Technology and other laws and administrative regulations. For violations without stipulations in laws and regulations, relevant competent departments shall issue warnings, make public criticisms, order corrections within a time limit according to their duties. Those refusing to correct or with serious circumstances shall be ordered to suspend relevant services. Those constituting violations of public security administration shall be punished for public security administration according to law. Those constituting crimes shall be investigated for criminal responsibility according to law.
Chapter V Supplementary Provisions
Article 22 The definitions of the following terms in these Measures are:
(1) Generative AI technology refers to models and related technologies with capabilities of generating content such as text, images, audio and video.
(2) Generative AI service providers refer to organizations and individuals that provide generative AI services using generative AI technologies (including providing generative AI services by offering programmable interfaces).
(3) Generative AI service users refer to organizations and individuals that generate content using generative AI services.
Article 23 Where laws and administrative regulations require administrative licensing for the provision of generative AI services, providers shall obtain the license in accordance with the law. Foreign investment in generative AI services shall comply with relevant laws and administrative regulations on foreign investment.
Article 24 These Measures shall come into force as of August 15, 2023.