China sets trial ethics rules for AI science and technology activities

China’s Ministry of Industry and Information Technology and nine other departments have issued the ‘Measures for AI science and technology ethics review and services (Trial)’, setting out rules on scope, support measures, implementing bodies, working procedures, supervision, and legal responsibility.

The text says the measures are intended to regulate ethics governance for AI science and technology activities and to support fair, just, safe, and responsible innovation.

The measures apply to AI scientific research, technology development, and other science and technology activities carried out in China that may raise ethics risks relating to human dignity, public order, life and health, the ecological environment, or sustainable development.

The text states that ethics requirements should run through the whole process of AI activities and lists principles including promoting human well-being, respecting life and rights, fairness and justice, reasonable risk control, openness and transparency, privacy and security protection, and controllability and trustworthiness.

On support measures, the document calls for improving the AI ethics standards system, including international, national, industry, and group standards. It also calls for stronger risk monitoring, testing, assessment, certification, and consulting services, more support for small and micro enterprises, work on ethics review research and technical innovation, the orderly opening of high-quality datasets, development of risk assessment and audit tools, public education, and ethics-related talent training.

The measures state that universities, research institutions, medical and health institutions, enterprises, and other entities engaged in AI science and technology activities are responsible for ethics review management within their own organisations and should establish AI science and technology ethics committees.

Local authorities and relevant departments may also establish specialised ethics review and service centres that provide review, re-examination, training, and consulting services on commission, but may not both review and re-examine the same AI activity.

The text sets out application and review procedures, including general, simplified, expert re-examination, and emergency procedures. It says review should focus on human well-being, fairness and justice, controllability and trustworthiness, transparency and explainability, traceability of responsibility, and privacy protection. Review decisions are to be made within 30 days after acceptance, subject to extension in complex cases. An emergency review is generally completed within 72 hours.

The measures also provide for expert re-examination of listed activities. The attached list covers human-machine integrated systems with a strong influence on human behaviour, psychological emotions, or health; algorithmic models, applications, and systems with the capacity for social mobilisation or guidance of social consciousness; and highly autonomous automated decision systems used in scenarios involving safety or health risks. The text says the list will be adjusted dynamically as needed.

The document further states that violations may be investigated and handled under laws, including the Cybersecurity Law, the Data Security Law, the Personal Information Protection Law, and the Science and Technology Progress Law. According to the text, the measures take effect upon issuance.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!