The advent of ChatGPT made the already heated artificial intelligence (“AI”) sector even hotter in China. Companies and individuals have reacted to AI developments with mixed feelings. Some have embraced it, some are worried and some have even opposed it. The Cyberspace Administration of China (the “CAC”) has decided to regulate the industry sector. After releasing the Administrative Provisions on Deep Synthesis of Internet-based Information Services last December to particularly address the use of AI in providing internet-based information services, the CAC recently issued a more comprehensive set of draft regulations titled Administrative Measures on Generative Artificial Intelligence Services (the “Measures”) for public comment. A summary of the key points and developments under the Measures that may be worth considering are set out below.
Scope of Application
The Measures apply to research, development and utilization of generative AI products to provide services to the public within the PRC. This means the Measures will have extra-territorial effect which was not the case with previous AI-related regulations. Foreign companies using their AI products to provide services to the Chinese public will be subject to the Measures. They will have to take actions to comply with the Measures when developing any AI products that will be used in serving the Chinese public, or block the Chinese public from access to their AI products in order not to be covered by the Measures.
The Measures define Generative AI as technologies for generating text, pictures, sounds, videos, codes and other contents on the basis of algorithms, models or rules. Again, this expands the scope of application by including models and rules-based AI which was not covered under the previous AI-related regulations.
Major Obligations
Under the Measures, companies engaged in provision of the generative AI products or services should fulfil the following major obligations:
- To ensure that the AI generated contents (“AIGC”) are legal, consistent with the core values of the socialist system and free from discrimination and fake information. Previous regulations only require the AIGC not to contain any fake news. It is not clear how fake information will be interpreted under the Measures.
- To ensure that the pre-training data and optimized training data are from a legal source and are authentic, accurate, objective and diverse.
- To tag the AIGC in accordance with the Administrative Provisions on Deep Synthesis of Internet-based Information Services.
- To provide safe, stable and continuous services to ensure users’ normal use within the life cycle. This is a new requirement that will need to be clarified.
- Before using the AI product to provide services to the public, to apply to the CAC for security assessment in accordance with the Provisions on the Security Assessment of Internet Information Services Characterized by Public Opinions or Capable of Social Mobilization, and within ten working days after the launch of the AI product, to complete the filing through the internet-based information services algorism filing system in accordance with the Administrative Provisions on Recommendation of Internet-based Information Service Algorithms. The Measures expand the application of the foregoing requirements from application to AI products/services used by the public for voicing public opinions or mobilizing the society to all generative AI products/services.
- To take appropriate measures to prevent users’ over-reliance upon or addiction to the AIGC. This is a new requirement compared to the previous regulations which only required the provider not to induce minors into internet addiction.
- To guide users to scientifically understand and rationally use the AIGC. This is also a new requirement not found in previous regulations.
- To establish a user complaint mechanism to deal with users’ requests on personal information protection and complaints about illegal AIGC.
- To cease generation and also prevent re-generation through model optimization training or otherwise which shall be completed within three months, upon discovering or becoming aware of any illegal content.
- To bear the responsibilities as the producer of the AIGC generated by a third party by utilizing their AI products or services (including programmable interfaces or otherwise) and bear the statutory responsibilities and fulfil the obligations as the data processor where personal data is involved. This may be the most concerning of the foregoing obligations. It deviates from the general rules under PRC law that the provider of the internet services is only liable to the extent that it fails to take appropriate measures in a timely manner after receiving an infringement notice from the allegedly infringed party or take necessary measures when it knows or should have known its services have been used to commit an infringement. It may also be problematic for the AI product/service provider to fulfil its obligations as the data processer in relation to the personal data that the public collects or otherwise deals with when using such AI product/service. These provisions may need to be clarified or refined so as to impose reasonable obligations and responsibilities on the AI product/service providers.
Liabilities for Violations
Violations of the Measures will be punished according to the PRC Cyber Security Law, the PRC Data Security Law, the PRC Personal Information Protection Law and other relevant laws and regulations. In the absence of relevant provisions in the foregoing laws and regulations, violators may be given warning or circulation of criticism and/or ordered to make rectification within a prescribed period by the CAC. In case of failure to obey the rectification order or under serious circumstances, the violators may be ordered to cease or suspend provision of services by using their AI products and fined in an amount from RMB 10,000 to 100,000.
The Measures may be finalized and officially released soon as the proposed effective date contained in the Measures includes a reference to the year 2023. It is obvious that the PRC government wishes to regulate the AI sector, but striking the delicate balance between regulating it and supporting it will be a huge challenge. We expect a rapid expansion in the regulations impacting this sector. Deacons will consistently monitor the legislative developments in this area and provide updates to our clients. Please do not hesitate to contact us if you have any questions.