China: Interim Measures Governing Generative AI Issued, AI Law on the Way

China has joined the growing ranks of jurisdictions seeking to regulate artificial intelligence (“AI”). The Cyberspace Administration of China (“CAC”) and six other regulators jointly issued Interim Administrative Measures for Generative Artificial Intelligence Services (生成式人工智能服务管理暂行办法, “Measures”) on July 13, 2023. Meanwhile, the General Office of the State Council published its 2023 Legislative Plan on May 31, 2023, listing preparation of a draft Artificial Intelligence Law (“AI Law”) among its legislative priorities, signaling more rulemaking to come in the AI sector.

The Measures

The issuance of the Measures, which will come into effect on August 15, 2023, has followed in short order the publication of draft measures for public comment on April 11, 2023.  

The April draft contemplated extensive restrictions on the development and use of generative AI in China. In contrast, the final Measures adopt a more balanced approach, reflecting the Chinese government’s keenness to foster the development of AI at the same time as it regulates its development and use.  

The scope of the final version of the Measures provides a key illustration of this change. They apply only to generative AI services that are available to the general public in China and exclude the conduct of internal research on or use of generative AI technology by industrial organizations, companies, educational and research institutions, and other professional organizations. This scope is much narrower than that contemplated in the April draft, which anticipated regulating the development and use of all generative AI, with no carve-outs.

Notable obligations under the Measures include:

• Generative AI services providers (“Service Providers”) must sign service agreements with users, specifying each party’s rights and obligations.
• Generative AI services that are “capable of influencing public opinion or mobilizing the public” are subject to a security evaluation requirement and a filing with CAC in accordance with the Administrative Provisions on Recommendation Algorithms in Internet-based Information Services (互联网信息服务算法推荐管理规定). How broadly the term “capable of influencing public opinion or mobilizing the public” will be interpreted remains to be seen.
• Service Providers must not collect unnecessary personal information, unlawfully retain input data or user history from which a user’s identity can be inferred, or unlawfully supply a user’s input data or user history to a third party. Service Providers must also promptly respond to individuals’ requests for inspection, copying, correction, supplementing, or deletion of their personal information.
• Service Providers must take effective measures to improve the quality of training data, including its authenticity, accuracy, objectivity, and diversity—a softer requirement than that contemplated under the April draft, which would have made Service Providers responsible to ensure the truth and accuracy of both training data and AI-generated content.
• Service Providers are responsible for the content generated by their services. As such, they are required, among other things, to police usage of their services and take down unlawful content and block usage of their services that involves generation or dissemination of unlawful content.
• Service Providers must set up a complaints mechanism providing members of the public with a streamlined process to report illegal content and activities, and must respond promptly to such reports.
• Content generated using “deep synthesis” technology—such as “deepfake” audio or video content—must be labelled as such in accordance with the Administrative Provisions on Deep Synthesis of Internet-based Information Services (互联网信息服务深度合成管理规定).

Cross-Border Impacts

The Measures apply principally to Generative AI services offered by Service Providers in China but also state that Chinese regulators may impose technical measures or other sanctions to deal with Generative AI services offered to the general public in China by offshore providers if the services violate the Measures or other provisions of PRC laws. The Measures do not specify what technical measures might be imposed, but given the practical difficulties of extraterritorial enforcement, such measures are most likely to involve blocking online access to the offshore service from within China.

The Measures generally state that foreign investment in Generative AI services must comply with the applicable laws and regulations. In due course, China’s foreign investment policies may be updated to specify types of AI projects that are variously encouraged for foreign investment, prohibited for foreign investment, or subject to equity caps or other foreign investment restrictions.

Draft AI Law

There is as yet little reliable information on the likely content or schedule for promulgation of the AI Law.

Following China’s standard legislative drafting process, the draft AI Law will be prepared with the involvement of various relevant departments of the State Council and subject to multiple readings by the Standing Committee of the National People’s Congress before final adoption.

It’s difficult to predict how long that will take. For some new legislation, the legislative process took several years. Given the Chinese government’s focus on AI, a much faster process might be involved with the AI Law.

Please note that any information provided herein is not legal advice. Any information concerning the People’s Republic of China (“PRC”) is not an opinion on, determination on, or certification of the application of PRC law. We are not licensed to practice PRC law.