Beijing mandates internal ethics committees for all Chinese AI companies

Chinese companies engaging in artificial intelligence activities are required to set up internal “AI ethics review committees” under new rules released by Beijing on Thursday, effective immediately. The notice comes as policymakers look to ensure that fast-paced AI progress can continue in a “healthy” manner amid growing consumer and enterprise adoption. Jointly released by 10 government bodies and institutions including the Ministry of Industry and Information Technology, National Development and Reform Commission and Chinese Academy of Sciences, the notice calls for the committees to review AI-related activities for their impact on human wellbeing, including the fairness of AI algorithms and whether AI systems are “controllable” and “explainable”. The rules also apply to universities, research bodies and health institutions. Beijing introduced a unified science and technology ethics review system in 2023, which required companies and organisations engaged in “high-risk” activities such as developing AI models that “influence social discourse” or involve “highly autonomous decision-making” to conduct ethical reviews before commencing. However, the system has been criticised for failing to mitigate risks in the fast-evolving AI era due to ambiguity in its operating scope and review standards, as well as the lack of effective enforcement mechanisms. “Some review committees are merely going through the motions... making it difficult to achieve the intended goals of an ethical review system,” Zhao Jingwu, associate professor of law at Beihang University, wrote in a commentary last year. Since 2022, China’s big tech firms including Alibaba Group Holding and Baidu have established internal science and technology ethics review committees. Alibaba owns the South China Morning Post. Beijing began soliciting public comment on new draft rules for an AI ethics review system last August, which it said was part of its Global AI Governance Initiative aimed at pursuing “high-quality development and high-level security” for the AI industry. A key difference between the final and draft versions was the removal of a clause that previously stipulated that an AI ethics review committee should be established, “conditions permitting”. Another addition to the final version was “personal privacy protection” as a core principle of the review system. Chinese companies are also encouraged to participate in the setting of AI standards and to open-source “high-quality AI ethics review data sets”. Meanwhile, greater support should be provided to small and medium-sized enterprises to conduct internal AI ethics reviews, according to the regulators.