Generative AI Measures in China
Apply to providers of generative AI services and users of these services
On July 13, 2023, the Chinese government released its finalized regulations on generative artificial intelligence, titled the Interim Measures for the Management of Generative Artificial Intelligence Services ("Interim GAI Measures"), which became effective on August 15, 2023 (Article 24). These measures received joint approval from seven major Chinese government agencies, including the Ministry of Science and Technology, the Ministry of Public Security, and the Cyberspace Administration of China (CAC), the regulatory body in Beijing overseeing Internet-related issues. A preliminary version of the Interim GAI Measures was presented in April 2023 to gather public input.
Scope
The Interim GAI Measures apply to providers of generative AI services and users of these services. Generative AI services are defined as "organizations and individuals that use generative [AI] technology to provide generative [AI] services", including providing these services through application programming interfaces (APIs) (Article 22.2). The measures are extraterritorial, applying to providing services to the public within China (Article 2) and potentially to organizations outside China that provide generative AI services to persons in China. The measure exempts institutions that are engaged in the development and application of generative AI technology, but not offering generative AI services to the public.
Failure of generative AI providers outside China to comply with the Interim GAI Measures and other applicable laws will result in notifying relevant institutions. These institutions will then take technical measures and other necessary actions to address the non-compliance (Article 20).
Key requirements and obligations of Generative AI Service Providers:
- Development and Use:
- Avoid generating illegal content.
- Implement measures to prevent discriminatory content.
- Avoid exploiting advantages leading to monopoly or unfair competition.
- Respect others' rights, including portrait, reputation, honor, privacy, and personal information rights.
- Enhance transparency, accuracy, and reliability of generative AI services.
- Training Data:
- Use data and foundation models from legitimate sources.
- Avoid infringement on legally owned intellectual property.
- Obtain personal data with consent or as per legal requirements.
- Enhance the quality of training data in terms of truthfulness, accuracy, objectivity, and diversity.
- Provision of Services:
- Establish clear service agreements with registered users.
- Guide users on legal use and prevent over-reliance on generated AI services.
- Minimize collection of non-essential personal information and protect user data.
- Respond to data subject requests and settle user complaints.
- Tag generated content as per regulatory provisions.
- Take prompt action to stop and rectify illegal content.
- Warn, restrict, suspend, or terminate services for users engaging in illegal activities.
- Establish a user complaint mechanism.
- Legal Obligations and Enforcement:
- Conduct safety assessments and record-filing for services with public opinion attributes or social mobilization ability.
- Cooperate with supervisory checks by relevant authorities, providing necessary information and assistance.
References:
- China: Generative AI Measures Finalized: https://www.loc.gov/item/global-legal-monitor/2023-07-18/china-generative-ai-measures-finalized/
- China's New Rules For Generative AI: An Emerging Regulatory Framework: https://www.fasken.com/en/knowledge/2023/08/chinas-new-rules-for-generative-ai