How are the Review Measures relevant? 

While the earlier AI Measures are aimed at technology companies providing solutions to the public in China, many of the obligations thereunder do not necessarily apply to those who are using AI technologies, e.g., in R&D activities. However, the Review Measures are different because they cover almost all R&D activities broadly. Science and technology activities - which in the context of China may be interpreted as equivalent to R&D activities - that involve any of the elements below are subject to the Review Measures: 

• activities involving human participants, including research activities such as tests, surveys, observations, etc., and use of human biological samples, personal information data, etc.; 
• activities involving laboratory animals; and 
• activities that do not directly involve human beings or laboratory animals but may pose ethical risks and challenges in areas such as life and health, ecological environment, public order, and sustainable development. 

Without these elements, the Review Measures may still be applicable according to other laws or regulations.  

According to the above, all R&D activities in the sectors relating to life sciences (e.g., pharmaceutical and medical devices including their upstream players) will be subject to the Review Measures. Besides, due to the very close links between AI and human behaviors (e.g., user data analysis and learning), the impact of these measures will be broader than it appears. As far as AI solutions are used, almost all R&D activities will fall into the scope of these Review Measures. This is further affirmed by Article 4 of these measures which, as agreed upon by universities, research institutes, and hospitals, explicitly mentions that 'enterprises' are subject to the measures. AI is clearly highlighted by the same Article where sensitive ethics areas are mentioned. 

Review committee 

The Review Measures require that scientific and technological activities follow the principles below, where many elements echo those under other global regulations, such as the EU guidelines on AI ethics: 

• ensure a good balance between innovation and risk prevention; 
• objectively assess and manage uncertainty and risks due to technological application; 
• promote human health, respect human rights, uphold fairness and justice, reasonably control risks, and maintain openness and transparency; and  
• abide by laws and regulations as well as industrial norms.  

Besides the above high-level principles, the Review Measures oblige companies to establish a Science and Technology Ethics Review Committee (Committee) within their organizations, particularly those conducting activities in life sciences, medical, or AI-related activities that involve 'sensitive areas of science and technology ethics.' Although this term has not yet been precisely defined, a prudential and recommended approach is to follow this requirement rather than rely on the argument that the concerned activities are not sensitive. Necessary resources like staffing, space, and funding shall be provided to the Committee to enable its independent performance of duties. Although a company may also entrust an external Committee to conduct the respective ethics review (as per Article 13 of the Review Measures), international companies still need to create this using internal resources in order to stay compliant with these measures, particularly considering that they became effective on December 1, 2023. Furthermore, the establishment of such a Committee shall be filed in an online platform set up by the Ministry of Industry and Information Technology (MIIT). Activities of the Committee must also be submitted to this platform for filing on an annual basis. 

The Review Measures have high requirements for the Committee as it is expected to establish a complete review structure and system for the company, including compliance with the following rules. It must have at least seven members, including a chairperson and several vice-chairpersons. The Committee is required to be diversified and include industry experts and experts with ethics and legal backgrounds, different genders, and external members. There is also a strict term limit for members of up to five years. The Review Measures address several code of conduct rules, covering issues such as independence and impartiality, confidentiality, avoidance of conflicts of interest, and regular training.  

Procedural guidance 

There are two tracks of review under the Review Measures.  

The review under the general procedure should not take more than 30 days and should be conducted by at least five members of the Committee with diversified backgrounds who shall be present in an on-site review meeting. Besides the high-level principles mentioned above, the review must also focus on, among other things, the following aspects of the planned activities: 

• sufficient qualifications and resources; 
• scientific and societal value, positive impact on the enhancement of human health, and the achievement of sustainable social development; 
• where human participants are involved, the guarantee of privacy and ethical protection as well as personal interests and safety; 
• minimal use of animals, if this use is necessary, and the guarantee of animal welfare; 
• good control over data and algorithms to ensure data security and the protection of personal information, with a contingency plan in place; and 
• a mechanism for managing conflicts of interest. 

A follow-up review of up to 12 months later shall apply to cases that receive an ethical review clearance. A simplified review procedure will then take place, as long as there is no major change in the case, and it may also replace the general procedure when a case involves low risks or in the case of a minor adaptation. 

The other review track applies to sensitive cases, which will trigger a so-called 'expert review procedure.' This applies to those cases in the (dynamically updated) list attached to the Review Measures, including the research activities that may pose significant ethical risks (i.e., such a case shall - upon review by the Committee - be submitted for further review by another expert panel to be organized by the competent industrial watchdog). All such expert reviews should be included in the annual filing report to be submitted to the MIIT online platform. 

Practical implications 

Generally, policy tendency in China focuses on first promoting the development of AI and later building the legal framework to regulate it, as can be determined from the policy priorities announced under its Next Generation Artificial Intelligence Development Plan published by the State Council on July 20, 2017. As in many other areas, the Chinese Government is taking an 'agile approach' and using 'patch strategy' to tackle issues and problems arising from the fast development of AI when it collides with human ethics and interests. Looking at the whole legislation process, one can only see fragmented regulations addressing certain types of AI technology or certain aspects of AI (like the Review Measures). Compared with those reflected in a unified legal framework, such as the EU AI Act substantiated with details, the regulatory attitude towards AI in China appears to be fairly tolerant and accepting.  

Although many existing rules may not directly affect companies using AI solutions in business (notably, the AI Measures) and a structured mechanism (i.e., the risk classification system), as well as the high-level principles mentioned above still lacking implementation details, the impact of these Review Measures should not be overlooked. They do not explicitly refer to the term AI, but they already include actionable obligations that apply to a wide range of companies. This is particularly important for R&D intensive businesses where the use of AI solutions is the trend.  

This is certainly not the end of new obligations. A pragmatic approach of the policymaker as such means that more rules addressing other specific aspects of AI will most likely be formulated in the near future, depending on the industrial development as well as whether these concerns are viewed as priorities for society. 

Dr. Michael Tan Partner 
[email protected] 
Julian Sun Senior Associate 
[email protected] 
Taylor Wessing, Shanghai