China’s Cyber Administration published for public comments a draft of administrative measures for artificial intelligence (AI) services. This is a plan aimed at regulating AI services provided to the public in China, across a wide range of areas such as privacy protection, prevention of discrimination, bias control, content management, security, algorithmic transparency, etc. It applies to all research, development, and use of AI products that are “provided” to the “public” in China. However, it is unclear whether “provided” includes services that are not specifically intended for China and whether “public” includes companies and organizations.

The plan covers content monitoring, which requires service providers to respect social morality, order, and customs in China. Service providers must ensure that any content generated by the AI is true matches socialist values and is not inciting or harmful, prevent discrimination during AI development, respect intellectual property rights, ensure that the content generated by the AI is true, accurate, and precise, and protect the rights of others.

Users are required to report to the Chinese authorities if they find content that violates these requirements, and AI service providers are required to block them. Service providers who violate these requirements may also be subject to criminal liability.

AI service providers are also required to maintain accurate and complete training data, establish a review mechanism for training data, and ensure that the use of training data does not violate laws, regulations, or social morality in China.

In addition, service providers must establish and implement security management systems, clarify the division of security management responsibilities, establish emergency response mechanisms, and report any security incidents to relevant Chinese authorities.

Click here to read the draft administrative Policy for artificial intelligence services (In Chinese).