Recently, the Cyberspace Administration of China issued new comprehensive provisional measures, which went into effect on August 15, 2023, and govern the development and use of generative AI (GAI) in China. The measures cover “GAI services” (including foreign-invested GAI services) provided within the PRC, including the text, pictures, audios, videos and other content they generate. These rules apply to the use of GAI services provided to the public, and not the internal use of generative AI by a company or research institution.

The importance of this regulation is that it confirms that the GAI service provider (including the platform operator) takes the primary responsibility for compliance. The GAI service provider will be viewed as the “internet content provider” (ICP) under the Chinese laws, which makes them subject to various other regulations and licensing requirements. Based on these measures, the practical effect of this law appears likely to prevent foreign generative AI service providers from providing such services in China.

Article 4 of the measures require providers of GAI services to ensure it complies with:

  • “Upholding socialist core values” – Including not generating any prohibited content, such as incites “subversion of the state power or the overthrow of the socialist system, endangers national security and interests, damages the national image, incites splitting the country, undermines national unity and social stability, advocates terrorism, extremism, ethnic hatred and discrimination, violence, pornography, and false and harmful information.”
  • Taking effective measures to prevent discrimination – With regard to, for example, nationality, religion, country, region, gender, occupation and health, in the design of the algorithm, training data set, model generation, optimization and service provision.
  • Respecting intellectual property rights and business ethics – Including keeping confidential trade secrets and refraining from “carrying out acts of monopoly and unfair competition” regarding proprietary algorithms, data and platforms.
  • Respecting others’ legitimate rights and interests – Refraining from endangering others’ physical and mental health and respecting their honor and privacy.
  • Boosting the transparency, accuracy and reliability of the contents generated by GAI services. 

A stated goal of the measures is to promote independent innovations as to GAI algorithms, frameworks, chips and supporting software platforms, including international exchanges on an “equal and mutually beneficial basis.” 

GAI service providers must conduct pre-training, optimization and other training ensuring that the use of data is:

  • From lawful sources
  • Not infringing IP
  • Obtained with the individual’s consent

GAI service providers are required to improve the quality of the training data set, carry out quality assessments and provide necessary training to the annotation staff in order to respect the law. The provider, which is the entity ultimately providing the service to the public in China, is responsible for the operation of GAI services, including ensuring that underage users (under 18 years old) are prevented from “over-relying on or addicting to GAI Services.” As for the information input by the user, the provider is bound not to unnecessarily or illegally collect, use or share the input (including personal) information.

The provider is required to obtain governmental approvals to the extent that the generated content creates text, pictures, videos or other content that are subject to government labeling requirements (such as requiring watermarks with the name and unique ID of the service provider). In addition, depending on the nature of the GAI services, an internet content provider license may be required. Further, additional licenses, such as press and publication, film and television and public opinion broadcast licenses, may be required. We note that most of the content-related licenses listed above prohibit foreign investment. 

The GAI service provider must timely remove illegal content, report the incident to the relevant authority and retrain the system to avoid such from happening again. For users who utilize GAI services for illegal activities, the provider must give warnings and service restrictions, keep records of the activity and report to the relevant authorities. In addition, the provider must provide an effective complaint- and whistleblower system for its users to report such abuse. Users also have the right to complain directly to the relevant authorities. 

To the extent that a GAI service may create public opinions or is capable of social mobilization, it must undergo a state security assessment, including filing the algorithm. 

GAI service providers that fail to follow the regulations are subject to punishment under any existing laws and, failing that, can be ordered to revise the service, issue public criticism of the service and/or to suspend the service. 

If you would like to discuss these measures or any other regulation applicable to AI, feel free to reach out to any of our market-leading global AI and data group including Scott Warren, Lindsay Zhu, Charmian Aw and Nick Chan (for APAC), David Naylor and Charles Helleputte (for Europe), or Alan Friel, Julia Jacobson and Kyle Fath (for the US)

Disclaimer: While every effort has been made to ensure that the information contained in this article is accurate, neither its authors nor Squire Patton Boggs accepts responsibility for any errors or omissions. The content of this article is for general information only, and is not intended to constitute or be relied upon as legal advice.