China Plans to Establish Regulations for Generative AI
The emergence of generative artificial intelligence (AI) models, such as ChatGPT, has sparked severe competition. China’s focus on ensuring compliance with this technology is driving the implementation of regulatory measures.
The Cyberspace Administration of China recently published a draft for the management of generative AI on April 11. The draft, which is currently open for public comment, consists of 21 articles that outline specific requirements for generative AI service providers. These requirements include admission criteria, algorithm design, training data selection, model and content regulations, user real-name authentication protocols, personal privacy protection measures and commercial secrets policies. This marks the first regulatory document in China aimed at regulating the rapidly growing generative AI industry.
The draft explicitly endorses and motivates the advancement of generative AI, a technology that utilizes algorithms, models, and rules to produce text, images, sound, video, code and other content. ChatGPT is an example of generative AI that has triggered a fresh wave of competition among tech giants like Microsoft, Google, Meta, Alibaba, Baidu as well as entrepreneurs.
SEE ALSO: Alibaba Cloud Unveils New AI Model Tongyi Qianwen
The security and ethical concerns surrounding ChatGPT have garnered worldwide attention. According to Yicai, Wu Shenkuo, the deputy director of China’s Internet Society research center, identified three key features of domestic regulatory ideas on generative AI from the draft. These include maximizing existing laws and regulations, prioritizing risk prevention and management, and adopting an ecological and process-oriented approach to regulation.
The proposed regulation mandates that companies offering generative AI products or services must adhere to legal requirements, uphold social morality and intellectual property rights, and follow commercial ethics. They are prohibited from exploiting algorithms, data, platforms, or other advantages for unfair competition. Users of these services must provide authentic identity information. Additionally, the draft emphasizes that content generated by generative AI should be truthful and precise while providers have a responsibility to safeguard user information and usage records.
The Biden administration in the United States is currently exploring the need for inspections of AI tools, such as ChatGPT. As an initial step towards potential regulation, the US Department of Commerce has officially requested feedback on accountability measures, including whether new AI models should be certified before their release. Italy was the first country to impose a ban on ChatGPT. The Italian Data Protection Authority announced a temporary prohibition on March 31 with immediate effect. Following this move by Italy, other EU countries have started considering relevant regulatory measures.