China recently made history by becoming the first country to introduce regulations regarding generative artificial intelligence (AI) models. The new law considered the most comprehensive of its kind worldwide, focuses on controlling generative AI tools like OpenAI’s ChatGPT and other similar technologies.
Led by the Cyberspace Administration of China, the provisional regulations were jointly published by seven Chinese regulators that are scheduled to take effect on August 15 and will be applied to all generative AI content services including text, pictures, audio, and videos provided to the Chinese public.
To regulate the deployment of ChatGPT-style services, Beijing emphasizes the importance of promoting wholesome content and upholding "core socialist values" in its latest law. In April, a preliminary version of the regulations was made public, and Chinese citizens were invited to provide feedback.
The updated regulations now exhibit a more supportive stance towards cutting-edge technology. Chinese authorities have committed to implementing measures that foster the innovative development of generative AI. Additionally, penalties such as fines ranging from 10,000 to 100,000 yuan for AI-related offenses have been removed from the final version of the law.
Generative AI tools available in China
Chinese companies have not yet fully launched their ChatGPT-style services for public use. Baidu's Ernie Bot and Alibaba Group Holding's Tongyi Qianwen are also in trial mode for now and are limited to business usage. In contrast, OpenAI's ChatGPT and Google's Bard models are not accessible in China at this time.
What are the Instructions for AI service providers?
According to the latest regulations, companies providing generative AI services must abide by "core socialist values." The law explicitly warns AI service providers against generating content that may incite subversion of state power, endanger national security and interests, harm the country's image, etc. Furthermore, the law mandates that companies ensure AI models and chatbots do not produce false or harmful information.
The Chatbots and AI models developed in China are required to incorporate features that ensure the generated content is free of unwanted information. As per a report by South China Morning Post, Zhong Hongyi, the chairman of 360 Security Technology, highlighted that their firm's chatbot includes a self-censorship function. During a presentation, Hongyi demonstrated that if a user enters a "sensitive word" in their prompts, the chatbot will promptly terminate the conversation.
Chinese regulators want AI service providers to be cautious in designing and selecting training data for their algorithms to avoid any form of discrimination based on ethnicity, faith, country, region, and other factors.
According to Xia Heilong, a lawyer based in Shanghai, China's focus is primarily on regulating AI services intended for public use. These regulations extend beyond AI-based services intended for industrial or internal corporate use.
The new rules also require generative AI service providers to promptly stop and report any illegal activities if users exploit AI to generate illicit content on their platforms. These providers also need to implement measures that prevent minors from becoming addicted to their AI services.