Overview and Legal Framework
China’s Interim Measures for the Administration of Generative AI Services, enacted on August 15, 2023, establish a regulatory framework for public-facing generative AI services within the country. The measures were approved by the Cyberspace Administration of China (CAC) and six other national regulators, including 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 regulations aim to promote the healthy development of generative AI while safeguarding national security, public interests, and the legal rights of citizens and organizations.
The measures explicitly define their scope as applying to services that utilize generative AI technology to produce text, images, audio, or video content for the public within China. Internal applications—such as those used solely by enterprises, research institutions, or other organizations without public exposure—are excluded from these requirements. The regulatory approach emphasizes a balanced strategy of development and security, combining innovation with legal compliance and adopting a tiered, risk-based supervision model.
Core Compliance Requirements for Service Providers
Providers of generative AI services must adhere to strict content and operational standards. The regulations prohibit the generation of content that undermines national security, promotes terrorism, extremism, violence, or obscenity, or violates socialist core values. This includes content that incites subversion of state power, disrupts social stability, or propagates discrimination based on nationality, ethnicity, gender, religion, or other protected attributes. Providers are also required to implement measures to prevent algorithmic bias and ensure fairness in AI-generated outputs.
Data integrity and transparency are critical components of compliance. Training data must be lawfully sourced, high-quality, and accurately labeled. Providers must establish clear labeling rules, conduct quality assessments, and train personnel involved in data annotation to ensure adherence to legal and ethical standards. If personal information is used in training or service provision, explicit user consent must be obtained, and data protection obligations must be strictly observed. The regulations also mandate that providers minimize unnecessary collection of personal data and avoid unauthorized retention or disclosure of user inputs or usage records.
Technical and Operational Safeguards
This includes deploying mechanisms to detect and prevent the generation of illegal or harmful content, such as real-time monitoring, automated filtering, and model retraining. If illegal content is generated or users exploit the service for unlawful activities, providers must take immediate action—such as halting generation, issuing warnings, restricting access, or terminating service—and report such incidents to relevant authorities. Additionally, providers must establish accessible complaint and reporting channels to address public grievances efficiently, with defined processing timelines and feedback mechanisms.
The regulations also emphasize the protection of minors by requiring providers to clearly define appropriate usage scenarios and implement safeguards against excessive or harmful reliance on AI services. Transparency in service capabilities and limitations is encouraged to foster responsible user engagement.
Regulatory Oversight and Enforcement
Oversight of generative AI services is distributed among multiple regulatory bodies, including the Cyberspace Administration of China, Ministry of Public Security, and other sector-specific agencies. Services with significant influence on public opinion or social mobilization—such as those capable of shaping narratives or mobilizing large audiences—are subject to mandatory security assessments and algorithm registration in accordance with existing regulations like the Management Measures for Algorithm Recommendations.
Non-compliance with the measures can result in penalties ranging from warnings and public reprimands to service suspensions or legal action under relevant laws. The regulations explicitly authorize government intervention against foreign-based services that fail to meet the requirements, including the implementation of technical measures to block or restrict access. Providers found in violation may also face fines or other administrative sanctions, depending on the severity and nature of the breach. The framework underscores the importance of collaboration between providers, regulatory bodies, and international partners to foster a secure and innovative AI ecosystem.
Encouragement of Innovation and International Cooperation
While imposing stringent compliance requirements, the measures also encourage innovation in generative AI technology. The regulations promote the development of autonomous AI infrastructure, including secure chips, software platforms, and public training data resources. Collaboration among industry players, research institutions, and cultural organizations is supported to advance AI capabilities, expand high-quality datasets, and optimize application scenarios. International cooperation is fostered through equitable partnerships and participation in global AI governance initiatives, provided that such collaborations adhere to national security and legal standards.
Providers are encouraged to explore ethical applications of generative AI, such as generating positive, uplifting content that aligns with societal values. The measures aim to create a balanced regulatory environment that supports technological progress while mitigating risks to public safety, privacy, and social harmony.
Required actions
- Public-facing generative AI services with public-opinion attributes must complete a security assessment and algorithm filing before launch - Use lawful training data and respect intellectual property; improve accuracy and objectivity of outputs; content must uphold socialist core values - Label generated content, protect personal information, respond to user complaints and take measures against illegal content within the service - Sign service agreements with users and take extra care with minors