China’s Generative AI Measures: Regulating Content, Models, and Algorithms

Article 14 of China’s recently released "Generative AI Measures" aims to safeguard user rights and ensure responsible development, use, and governance of generative AI. The article emphasizes the importance of integrating content governance with model governance, as AI oversight needs to adapt to the rapidly evolving nature of Generative AI.

Key Provisions

1. Immediate Action: Providers must take prompt action upon identifying illegal or discriminatory content, including stopping its generation and dissemination, removing the content, and notifying the appropriate authorities.
2. Labeling AI-Generated Content: Service providers are required to label AI-generated content to inform users that they are engaging with machine-created material. This enables rational filtering and assessment of relevant content.
3. Public-Facing Components: Generative AI services provided to the Chinese public must adhere to provisions protecting user rights, including complying with cybersecurity obligations as online information content producers, personal information protection obligations, stable and secure service implementation, and setting up a complaint system.
4. User Agreement: Service providers must sign a user agreement that outlines the terms of use, privacy policy, and other relevant details.
5. Compliance Enforcement: The Cyberspace Administration of China (CAC) holds the authority to halt or revoke online privileges for foreign Generative AI service providers contravening Chinese statutes, protocols, or article clauses.

Major Takeaways

• Article 14 underscores the crucial role content governance plays in responsible AI development and use.
• Providers must adapt to the rapidly evolving nature of Generative AI by integrating content governance with model governance.
• The article emphasizes the importance of prompt action upon identifying illegal or discriminatory content, ensuring that users are protected from harmful material.
• Labeling AI-generated content is essential for enabling rational filtering and assessment of relevant content.
• Generate AI services provided to the Chinese public must adhere to provisions protecting user rights, including cybersecurity obligations, personal information protection obligations, stable and secure service implementation, and setting up a complaint system.
• The CAC holds the authority to halt or revoke online privileges for foreign Generative AI service providers contravening Chinese statutes, protocols, or article clauses.