China: CAC publishes Interim Measures on Generative AI
On July 13, 2023, the Cyberspace Administration of China (CAC) published the Interim Measures for the Management of Generative Artificial Intelligence Services. This follows the request for public comments on the Measures on April 11, 2023.
'Generative artificial intelligence technology' under Article 22 of the Measures includes models and related technology that have the ability to generate content such as text, pictures, audio, and video.
'Generative artificial intelligence service providers' includes, under Article 22 of the Measures, organizations and individuals that use generative artificial intelligence (AI) technology to provide generative AI services, including by providing AI programmable interfaces.
'Generative artificial intelligence service users' refers to organizations and individuals who use generative AI services to generate content.
The Measures apply to the use of generative AI technology to provide services for generating text, pictures, audio, video, and other content to the public within China. Notably, industry organizations, enterprises, educational and scientific research institutions, public cultural institutions, and relevant professional institutions that develop and apply generative AI technology, but do not provide generative AI services to the domestic public, are not subject to the Measures.
Article 4 of the Measures outline principles for the provision and use of generative AI including, among others:
- taking measures to prevent discrimination based on ethnicity, belief, country, region, gender, age, occupation, and health in the process of algorithm design, training data selection, model generation and optimization, and service provision;
- respecting intellectual property (IP) rights, business ethics, keeping business secrets, and not using algorithms, data, platforms, and other advantages to implement a monopoly and unfair competition;
- respecting the legitimate rights and interests of others, not endangering the physical and mental health of others, and not infringing on the privacy rights and personal information of others; and
- based on the service type, taking effective measures to improve the transparency of generative AI services and improve the accuracy and reliability of generated content.
Further, the Measures establish that generative AI service providers must carry out training data processing activities such as pre-training and optimization training, including:
- using data and basic models with legitimate sources;
- not infringing on the IP rights of others;
- obtaining the consent of individuals when using personal information, or complying with other administrative regulations in accordance with the law;
- taking effective measures to improve the quality of training data, and enhancing the authenticity, accuracy, objectivity, and diversity of training data; and
- complying with other provisions and administrative regulations, such as the Personal Information Protection Law (PIPL).
Further, when data labeling is carried out during the research and development of generative AI, the generative AI service provider must formulate clear, specific, and operative labeling rules that meet the requirements of the Measures, carry out a data labeling quality assessment, and conduct a sampling verification of the accuracy of labeling content.
In addition, generative AI service providers must assume responsibility as network information content producers, and where personal information is involved, assume the responsibility of a personal information processor.
Generative AI service providers must sign service agreements with service users, clarifying the rights and obligations of both parties. Equally, generative AI service providers must clarify and disclose the applicable population, circumstances, and use of their service, and guide users to understand and use their generative AI technology in accordance with the law. Notably, the Measures detail that generative AI service providers must prevent minor users from excessively relying on the use of generative AI.
Generative AI service providers must also mark pictures, videos, and other generated content in accordance with the Regulations on Administration of Deep Synthesis of Internet Information Services.
Generative AI service providers must also, under Article 11 of the Measures:
- establish protection mechanisms for users' input information;
- not collect non-essential personal information;
- not illegally retain input information and use records that can identify users; and
- not illegally provide users with input information and usage records.
Regarding users, the Measures clarify that generative AI service providers must accept and process individuals' requests for reviewing, copying, correcting, supplementing, and deleting their personal information in a timely manner in accordance with the law. Similarly, generative AI service providers must establish and maintain a complaint and report mechanism, including a specific portal, with information on the process and feedback time limit.
Finally, when the generative AI service provider discovers users are using the service to engage in illegal activities, it must take action to warn, restrict, suspend, or terminate the provision of the service.
The Measures enter into force on August 15, 2023.
You can read the Measures, only available in Chinese, here.