China’s First Regulation on Generative AI Supervision

I. Introduction

On July 13th , 2023, the National Internet Information Office and other departments jointly issued the Interim Measures for the Administration of Generative Artificial Intelligence Services (“Measures”), which will come into effect and be implemented on August 15th, 2023, and put forward detailed compliancy requirements in terms of content, data and algorithms. It is notably the first regulatory document for artificial intelligence generated content (“AIGC”) in China.

II. Scope of Application

According to Article 2 of the Measures, it shall apply to the services that utilize generative artificial intelligence technology in providing to the public in China generated text, images, audio, video and other content (hereinafter referred to as “generative artificial intelligence services”), and shall not be applicable to relevant institutions that use generative artificial intelligence technology while not providing generative artificial intelligence services to the public within China. Therefore, there are two points to note:

1.The Measures apply to providers who provide services to the public in China. Providers, whether domestic or foreign, as long as they provide services to the domestic public in China, fall within the scope of supervision.

2. The Measures specify that industry organizations, enterprises, educational and scientific research institutions, public cultural institutions, relevant professional institutions and other entities that develop and apply generative AI technologies and fail to provide generative AI services to the domestic public are not within the scope of application, which reflects the inclusive and encouraging regulatory attitude towards new technologies in China.

III. Key Points of Regulatory Compliance

Article 3 of the Measures stipulates that the State implements the principles of inclusive prudence and categorical and hierarchical supervision of generative AI services, that is, it is necessary to encourage innovation and reasonable supervision. The Measures set out different regulatory requirements for different stages of AI generation:

1. Development Phase

For training data processing activities such as pre-training and optimization training, Article 7 of the Measures puts forward the following five compliance points: (1) The data and basic models have legitimate sources; (2) shall not infringe the intellectual property rights enjoyed by others according to law; (3) Where personal information is involved, it shall comply with legal provisions; (4) Ensure the authenticity, accuracy, objectivity and diversity of training data; (5) Comply with laws and regulations and the relevant regulatory requirements of the relevant competent authorities.

2. Preparatory Phase

Before officially providing generative AI services to the public, providers also need to pay attention to the following compliance obligations according to the Measures:

(1) Service agreements: Article 9, paragraph 2, of the Measures stipulates that providers shall sign service agreements with users of generative AI services to clarify the rights and obligations of both parties.

(2) Safety assessment: Article 17 of the Measures clearly states that the provision of generative artificial intelligence services with public opinion attributes or social mobilization capabilities should carry out safety assessment in accordance with relevant national regulations, and perform algorithmic filing as well as change and cancellation filing procedures.

3. Application Phase

The Measures set out a number of obligations for providers, such as: protecting personal information obligations; clarifying and disclosing the applicable groups of their services, guidance and anti-addiction obligations; the obligation to label generated content, the obligation to safeguard the normal use of the user; and the obligation to deal with and rectify illegal content.

IV. Conclusion

At this stage, globally, the regulation of AI in various countries is also in active exploration and practice. The newly introduced Measures establish the regulatory principles of inclusiveness, prudence, and classified and graded supervision, striving to seek a balance between promoting innovation, risk prevention and control for generative AI services.

For more information, please contact us at: info@dandparnters.com.

Disclaimer

The content and the writing of the present document did not enjoy any other input in addition to the ones specifically mentioned. D’Andrea & Partners reserves the right to update the information included in the present document at any time, according to any change in the relevant Rules or updated information. The use of the present document is strictly reserved to the recipient only. Any transfer or any total or partial reproduction is strictly forbidden unless previous authorization from D’Andrea & Partners.