The Cyberspace Administration of China (CAC) has published interim measures for the regulation of Generative Artificial Intelligence (GAI) services. These measures will apply to organizations and individuals providing GAI services in China, including content generation (text, images, audio, and video). Developers who do not offer services based on GAI will be exempt from these rules.

The measures require the prevention of discrimination based on ethnicity, belief, nationality, region, gender, age, occupation, and health during the process of algorithm design, data selection for GAI training, model creation, and service delivery. The rules also emphasize the importance of transparency in GAI services and respect for intellectual property rights, business ethics, and privacy. GAI service providers are also required to label content generated through GAI as such.

Under the rules, GAI training warrants special obligations, such as use of data and models from legitimate sources, the need for consent from data subjects when using their personal data, and effective measures to improve the quality of training data, authenticity, accuracy, objectivity, and diversity.

GAI service providers must use service agreements that clarify their rights and obligations when using their services. Special attention is given to preventing minors from over-relying on AI services. Additionally, if unlawful activities are identified, providers are required to warn, limit, suspend, or terminate the provision of services. These interim rules will come into effect on August 15, 2023.

Click here to read the full Interim Measures for the Management of Generative Artificial Intelligence Services (In Chinese).