The new draft measures clarify the scope of AI-related activities they cover, the entities responsible for ethical review, and the procedures reviews must follow.
China has unveiled draft measures on AI ethics, proposing more targeted and specialized review requirements for the technology.
Jointly issued by 11 government agencies and research institutes, the draft serves as a detailed follow-up to the 2022 Opinions and the 2023 Measures on sci-tech ethics by targeting the AI sector.
In 2022, China issued its first opinions to strengthen ethics governance in science and technology, mandating an ethical review regime.
In 2023, the central government issued administrative measures setting out a unified framework for ethics reviews in scientific research, covering studies involving human participants, laboratory animals and projects posing risks to health, the environment, public order and sustainable development.
The new draft clarifies the scope of AI-related activities it covers, the entities responsible for ethical review, and the procedures reviews must follow.
The draft applies to AI research and development in China that may involve ethical risks to life and health, human dignity, the environment, public order or sustainable development, as well as other activities requiring review under national laws and regulations.
The draft adds new principles for entities — including respect for intellectual property, ensuring controllability and trustworthiness, and strengthening accountability — building on the universal principles in the 2023 measures, such as respect for life, fairness, prudent risk management and transparency.
Universities, research institutes, medical organizations and companies engaged in AI research bear primary responsibility for their own AI ethics management services, with qualified entities required to establish internal AI ethics review committees. In addition to meeting standards on staffing, premises and funding, these committees should include experts with diverse backgrounds in AI technology, ethics and law.
Entities without committees, or with inadequate ones, may rely on specialized AI ethics service centers set up by local authorities and regulators through qualified institutions. These centers can provide ethics reviews, training and consulting for AI development on a commissioned basis.
Compared with the 2022 Opinions and 2023 Measures, the draft sets more detailed requirements for the review materials, including disclosure of algorithms, the type, source and collection method of data, cleaning methods, testing and evaluation processes, the products to be developed, and intended fields and user groups.
Applicants must also assess in detail the potential ethical risks from intended use, misuse or abuse of the technology, and present a plan for prevention and control.
Public comments will be accepted until Sept. 22.
Source: MLex
