Interim Measures for Generative AI Services Management
China · Regulation · State Internet Information Office, National Development and Reform Commission, Ministry of Education, Ministry of Science and Technology, Ministry of Industry and Information Technology, Ministry of Public Security, National Radio and Television Administration
High impact. A reviewer confirms each change before it is listed here.
Our summary
This regulation establishes rules for services that use generative AI to create text, images, audio, and video content for the public in China. It applies to providers and users of such services, requiring them to comply with laws, respect intellectual property rights, prevent discrimination, protect personal information, and ensure content quality and transparency. Providers must conduct safety assessments, handle user complaints, and accept government supervision, with violations subject to penalties including fines, service suspension, and criminal liability.
Written by the Index in our own words. It is not legal advice; the official text is the authority.
Inside this policy
Every section of this policy in order. Our notes summarise each one in plain language and do not quote the official text, which is the authority.
Drafted by the Index An automated summary of the official text, not yet reviewed by a person.Reviewed Checked by a person against the official text.
29 sections
Jump to a chapter
第一章 General Provisions
4 sections
This chapter covers the foundational principles, scope of application, regulatory approach, and general obligations for generative AI services in China.
第一条 Legislative basis and purpose
The regulation aims to promote healthy development and standardized use of generative AI, protect national security and public interest, and safeguard rights of individuals and organizations under network security, data protection and intellectual property laws.
第二条 Scope of application
Applies to services providing text, images, audio or video content generated by AI technology to the public within China, except where other state regulations govern news publishing, film production or artistic creation, and excludes internal research and development not offered to the public.
第三条 Guiding principles
The state pursues balanced development and security, combines innovation promotion with lawful governance, and implements inclusive, cautious and tiered oversight of generative AI services.
第四条 General requirements
Providers and users must comply with law, uphold ethics and core values; prohibit generating content related to subversion, national harm, separatism, terrorism, discrimination, violence, pornography or false information; prevent discrimination in algorithm design and training; respect intellectual property and fair competition; protect others' rights including privacy and image; and enhance transparency and accuracy of generated content.
第二章 Technology Development and Governance
4 sections
This chapter covers encouragement of innovation, technology development, training data handling, and quality assurance measures for generative AI services.
第五条 Innovation encouragement
Encourages innovative applications of generative AI across sectors to produce quality content, and supports collaboration among industry, research, cultural and professional institutions on technology, data resources, application and risk prevention.
第六条 Basic technology and infrastructure
Encourages independent innovation in algorithms, frameworks, chips and software; international cooperation; and supports development of AI infrastructure, computing resources, and public training data platforms, including open public data and secure technologies.
第七条 Training data requirements
Providers must use legally sourced data and models; not infringe intellectual property; obtain consent for personal data or comply with other legal requirements; ensure training data quality, authenticity, accuracy, objectivity and diversity; and comply with cybersecurity, data security and personal information protection laws.
第八条 Data annotation standards
Providers conducting data annotation must establish clear, specific and operational annotation rules; assess annotation quality and verify accuracy through sampling; train annotators on legal compliance; and supervise proper annotation work.
第三章 Service standards
7 sections
This chapter covers provider obligations regarding content and information security, user guidance, data protection, content labeling, service reliability, illegal content handling, and complaint mechanisms for generative AI services.
第九条 Provider responsibilities
Providers must assume legal responsibility as content producers and comply with information security obligations. They must handle personal information in compliance with data protection law. Providers must sign service agreements with registered users that clarify both parties' rights and obligations.
第十条 Service transparency and user guidance
Providers must clearly disclose the applicable audience, scenarios and purposes of their services and guide users to use generative AI lawfully and rationally. Providers must implement effective measures to prevent minors from becoming excessively dependent on or addicted to such services.
第十一条 Protection of user data
Providers must lawfully protect user input information and usage records, must not collect unnecessary personal information, and must not unlawfully retain or share information identifying users. Providers must timely handle requests from individuals to view, copy, correct, supplement or delete their personal information.
第十二条 Labeling of generated content
Providers must label generated images, videos and other content in accordance with the regulations for internet information service deep synthesis.
第十三条 Service reliability
Providers must provide safe, stable and continuous services that ensure normal user access.
第十四条 Handling of unlawful content and activities
When discovering unlawful content, providers must immediately stop generation, halt transmission and remove it, then undertake model optimization and report to competent departments. When users misuse services for unlawful activities, providers must warn them, limit functions, suspend or terminate service, retain records and report to relevant authorities.
第十五条 Complaint and reporting mechanism
Providers must establish complaint and reporting systems with easily accessible entry points, publish processing procedures and response timelines, and promptly handle complaints with feedback to the public.
第四章 Supervision and legal liability
6 sections
This chapter covers oversight of generative AI services by multiple government departments and establishes legal responsibilities and enforcement mechanisms.
第十六条 Government supervision framework
Departments including internet, development, education, science and technology, industry and information, public security, broadcasting and publishing authorities shall strengthen management of generative AI services according to their respective responsibilities. Competent authorities shall develop adaptive, scientific supervisory approaches and classification-based regulatory rules or guidance for different industries and sectors.
第十七条 Safety assessment and algorithm registration
Providers of generative AI services with public opinion or social mobilization characteristics must conduct safety assessments according to national regulations and fulfil algorithm registration, modification and cancellation procedures under the Algorithm Recommendation Management Regulation.
第十八条 User right to complain
Users who discover that generative AI services do not comply with laws, administrative regulations or this measure have the right to lodge complaints and reports with relevant regulatory authorities.
第十九条 Supervision and confidentiality obligations
Regulatory authorities may conduct oversight inspections; providers must cooperate and explain training data sources, scale, types, annotation rules and algorithmic mechanisms. Organisations and personnel involved in assessments and inspections must keep state secrets, commercial secrets and personal information confidential.
第二十条 Cross-border non-compliant services
When services provided from outside the People's Republic of China into the country do not comply with laws and regulations, the national internet regulator must instruct relevant organisations to take technical and other necessary measures for handling.
第二十一条 Penalties and enforcement
Violators are subject to penalties under network security, data security and personal information laws; if not regulated elsewhere, authorities may warn, criticize publicly, order corrections, or suspend service provision. Violations of public security management rules incur management penalties; criminal violations are prosecuted.
第五章 Final rules
3 sections
This section contains definitions, administrative requirements and enforcement provisions in the final chapter.
第二十二条 Definitions
Generative AI technology means models and technologies capable of generating text, images, audio or video content. Providers are organisations or individuals offering such services including through application programming interfaces. Users are organisations or individuals generating content with such services.
第二十三条 Administrative approvals and foreign investment
Providers must obtain required administrative permits under law. Foreign investment in generative AI services must comply with laws and regulations governing foreign investment.
第二十四条 Effective date
This measure enters into force on 15 August 2023.
Controls
What this policy requires, forbids, grants or penalises, one control at a time and in our own words. Each control links to the official text, which is the authority. This is not legal advice.
Drafted by the Index Drafted automatically from the official text, not yet reviewed by a person.Reviewed Checked by a person against the official text.
Themes covered 22 of 35
Choose a theme to see only its controls. About the themes
Risk and scope 3 of 4
- Impact assessment, not covered
Data and privacy 4 of 4
Transparency and records 3 of 4
- Technical documentation, not covered
Technical safeguards 4 of 4
Assurance and lifecycle 2 of 4
- Conformity assessment and certification, not covered
- Post-market monitoring and corrective action, not covered
Rights and protected groups 2 of 4
- Automated decision-making rights, not covered
- Biometric identification and surveillance, not covered
Roles and organisation 3 of 6
- General purpose and foundation models, not covered
- Public sector use and procurement, not covered
- Innovation support and sandboxes, not covered
Enforcement and sectors 1 of 5
- Sector rules: health, not covered
- Sector rules: finance and insurance, not covered
- Sector rules: employment, not covered
- Sector rules: education, not covered
The full list of controls could not be loaded, so only the first 40 are shown.
第三条 1 control
- ObligationApplies to: Public authorityDrafted by the Index
The state must balance development with security and combine innovation promotion with lawful governance, using inclusive and differentiated oversight of generative AI services.
第三条
第四条 5 controls
- ProhibitionApplies to: AnyoneDrafted by the Index
Providers and users must not generate content that incites subversion of state power, harms national security or image, incites separatism, promotes terrorism or extremism, foments ethnic hatred or discrimination, or spreads violence, obscene content or false harmful information.
第四条 - ObligationApplies to: ProviderDrafted by the Index
Providers must take effective measures during algorithm design, training data selection, model generation and service delivery to prevent discrimination based on ethnicity, faith, nationality, region, gender, age, occupation or health status.
第四条 - ObligationApplies to: AnyoneDrafted by the Index
Providers and users must respect intellectual property, business ethics and trade secrets, and must not use algorithmic, data or platform advantages to engage in monopolistic or unfair competitive practices.
第四条 - ObligationApplies to: AnyoneDrafted by the Index
Providers and users must respect others' legal rights and must not harm others' physical or mental health, or infringe on others' portrait rights, reputation, honor, privacy or personal information rights.
第四条 - ObligationApplies to: ProviderDrafted by the Index
Providers must take effective measures suited to their service type to enhance transparency of generative AI services and increase accuracy and reliability of generated content.
第四条
第七条 5 controls
- ObligationApplies to: ProviderDrafted by the Index
Providers must use training data from lawful and legitimate sources.
第七条 - ProhibitionApplies to: ProviderDrafted by the Index
Providers must not infringe intellectual property rights when using data involving intellectual property in training.
第七条 - ObligationApplies to: ProviderDrafted by the Index
Providers must obtain individual consent or meet other legal grounds when using personal information in training data.
第七条 - ObligationApplies to: ProviderDrafted by the Index
Providers must take effective measures to improve training data quality, truthfulness, accuracy, objectivity and diversity.
第七条 - ObligationApplies to: ProviderDrafted by the Index
Providers must comply with network security, data security and personal information protection laws and regulatory requirements.
第七条
第八条 3 controls
- ObligationApplies to: ProviderDrafted by the Index
Providers must establish clear, specific and operational data labeling rules that comply with this regulation.
第八条 - ObligationApplies to: ProviderDrafted by the Index
Providers must conduct quality assessments of data annotation and sample-verify the accuracy of labeled content.
第八条 - ObligationApplies to: ProviderDrafted by the Index
Providers must provide necessary training to labeling personnel and monitor their compliance with lawful and proper annotation practices.
第八条
第九条 3 controls
- ObligationApplies to: ProviderDrafted by the Index
Providers must comply with network information security laws and fulfill the legal responsibilities of content producers.
第九条 - ObligationApplies to: ProviderDrafted by the Index
Providers must comply with personal information protection laws and fulfill the legal responsibilities of personal information processors.
第九条 - ObligationApplies to: ProviderDrafted by the Index
Providers must establish and sign written service agreements with registered users that clearly set out the rights and obligations of both parties.
第九条
第十条 3 controls
- ObligationApplies to: ProviderDrafted by the Index
Providers must publicly specify the intended user groups, appropriate contexts and intended purposes of their services.
第十条 - ObligationApplies to: ProviderDrafted by the Index
Providers must guide users to understand and lawfully use generative AI technology in a rational and scientific manner.
第十条 - ObligationApplies to: ProviderDrafted by the Index
Providers must implement effective measures to prevent underage users from becoming excessively dependent on or addicted to their services.
Applies when: minors
第十条
第十一条 3 controls
- ObligationApplies to: ProviderDrafted by the Index
Providers must legally protect user inputs and usage records and must not collect unnecessary personal information.
第十一条 - ProhibitionApplies to: ProviderDrafted by the Index
Providers must not unlawfully retain identity-linked user inputs and usage records or share them with third parties without authorization.
第十一条 - ObligationApplies to: ProviderDrafted by the Index
Providers must promptly handle user requests for access to, copying of, correction of, supplementation of or deletion of their personal information.
第十一条
第十二条 1 control
- ObligationApplies to: ProviderDrafted by the Index
Providers must mark generated images, videos and other content according to applicable regulations on deepfake and synthetic media management.
第十二条
第十三条 1 control
- ObligationApplies to: ProviderDrafted by the Index
Providers must deliver safe, stable and continuous service operations to ensure users can access the service normally.
第十三条
第十四条 6 controls
- ObligationApplies to: ProviderDrafted by the Index
Providers must immediately stop generating and transmitting unlawful content and remove it from their service.
第十四条 - ObligationApplies to: ProviderDrafted by the Index
Providers must retrain and optimize their models when they discover unlawful content.
第十四条 - ObligationApplies to: ProviderDrafted by the Index
Providers must report unlawful content to relevant government authorities.
第十四条 - ObligationApplies to: ProviderDrafted by the Index
Providers must issue warnings, restrict functions, or suspend or terminate service for users engaged in illegal activity.
第十四条 - ObligationApplies to: ProviderDrafted by the Index
Providers must keep records of actions taken against users engaged in illegal activity.
第十四条 - ObligationApplies to: ProviderDrafted by the Index
Providers must report to authorities when they detect users using the service for illegal activity.
第十四条
第十五条 4 controls
- ObligationApplies to: ProviderDrafted by the Index
Providers must establish and maintain complaint and grievance mechanisms.
第十五条 - ObligationApplies to: ProviderDrafted by the Index
Providers must create accessible channels for complaints and reports.
第十五条 - ObligationApplies to: ProviderDrafted by the Index
Providers must publish the procedures and response timeframes for handling complaints.
第十五条 - ObligationApplies to: ProviderDrafted by the Index
Providers must promptly receive, handle and provide feedback on public complaints and reports.
第十五条
第十六条 2 controls
- ObligationApplies to: Public authorityDrafted by the Index
Multiple government departments must enforce applicable regulations according to their respective responsibilities.
第十六条 - ObligationApplies to: Public authorityDrafted by the Index
Government authorities should develop oversight methods and categorized management rules suited to the technology's characteristics and sectoral applications.
第十六条
第十七条 2 controls
- ObligationApplies to: ProviderDrafted by the Index
Providers offering generative AI services with editorial or social mobilization capability must conduct safety evaluations under national regulations.
Applies when: editorial or social mobilization capability
第十七条 - ObligationApplies to: ProviderDrafted by the Index
Providers of generative AI services with editorial or social mobilization capability must comply with algorithm registration, modification and cancellation procedures under internet information service algorithm recommendation rules.
Applies when: editorial or social mobilization capability
第十七条
第十八条 1 control
- RightApplies to: OthersDrafted by the Index
Users may file complaints or reports with relevant authorities if a generative AI service violates applicable laws, administrative regulations or this regulation.
第十八条
Change history
No changes recorded since we began monitoring this policy.