China Sets Security Requirements for Generative AI Services
China has unveiled proposed security requirements for companies offering generative artificial intelligence (AI) services. These regulations include a blacklist of sources that cannot be used to train AI models. The National Information Security Standardization Committee, which includes officials from the Cyberspace Administration of China, the Ministry of Industry and Information Technology, and the police, published these requirements on Wednesday.
Generative AI, which has gained popularity through OpenAI’s ChatGPT chatbot, uses past data to learn how to take actions and generate new content such as text or images. To ensure the safety and integrity of these AI models, the committee proposes conducting a security assessment of the content used for training. Any body of content that contains more than 5% of illegal or harmful information will be blacklisted. This includes information that advocates terrorism or violence, undermines national unity and social stability, or damages the country’s image.
The draft rules also emphasize that models trained with content censored on the Chinese internet should not be used. It is crucial for companies training AI models to seek the consent of individuals whose personal information, including biometric data, is used for training purposes. The regulations also provide detailed guidelines to prevent intellectual property violations.
The publication of these security requirements follows the recent approval by regulators for several Chinese tech firms, including Baidu, to launch generative AI-driven chatbots to the public. Since April, the Cyberspace Administration of China has been requesting security assessments from companies before launching generative AI services. Analysts note that the measures outlined in the draft rules published in July were less burdensome than those initially proposed in April.
Countries worldwide are grappling with establishing regulations for generative AI. China sees AI as a field in which it aims to rival the United States and become a global leader by 2030. These security requirements contribute to ensuring the responsible and controlled development of generative AI services in the country.
In conclusion, China’s proposed security requirements provide guidelines for companies offering generative AI services, including strict regulations on content used for training AI models. By prioritizing security and integrity, China aims to promote responsible and controlled development in the field of generative AI.