China AI ethics draft translated by Georgetown’s CSET
China’s draft AI ethics measures would require reviews for activities posing risks to health, dignity, public order, or sustainability.
The Center for Security and Emerging Technology (CSET), a policy research organisation within Georgetown University’s Walsh School of Foreign Service, has published an English translation of China’s draft trial measures on ethics reviews for AI technology.
The translated draft says the measures would apply to AI-related scientific and technological activities conducted within China that may pose ethical risks to human health, human dignity, the ecological environment, public order, or sustainable development. It covers universities, research institutions, medical and health institutions, enterprises, and other organisations involved in AI research and development.
Under the draft, organisations with the necessary conditions would be expected to establish AI technology ethics committees, while others could commission specialised ethics service centres to conduct reviews. Review applications would need to include details on the AI activity, algorithms, data sources, data cleaning methods, testing and evaluation, expected applications, user groups, risk assessments, and risk prevention plans.
The review process would focus on fairness and impartiality; controllability and trustworthiness; transparency and explainability; accountability and traceability; and whether the activity has scientific and social value. Committees or service centres would generally have 30 days to approve, reject, or request revisions to an application.
Higher-risk activities would require expert reconsideration. The draft list includes human-computer fusion systems that strongly affect behaviour, psychological or emotional states, or health; AI models and systems able to mobilise public opinion or channel social consciousness; and highly autonomous automated decision-making systems used in safety or personal health-risk scenarios.
Approved AI activities would also be subject to follow-up reviews, generally at intervals of no more than 12 months, while activities requiring expert reconsideration would be subject to follow-up reviews at least every 6 months. Emergency ethics reviews would normally have to be completed within 72 hours.
CSET notes that China released a final trial version of the regulation in April 2026, which it is now translating. The newly published draft translation therefore provides insight into the regulatory structure that preceded the final version, including committee-based ethics review, external service centres, expert reconsideration, and oversight roles for the Ministry of Science and Technology, the Ministry of Industry and Information Technology, and other departments.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!