China sets stricter rules for training generative AI models

cointelegraph.com:

China has released draft security regulations for companies providing generative artificial intelligence (AI) services, encompassing restrictions on data sources used for AI model training.

On Wednesday, Oct. 11, the proposed regulations were released by the National Information Security Standardization Committee, comprising representatives from the Cyberspace Administration of China (CAC), the Ministry of Industry and Information Technology and law enforcement agencies.

Generative AI, as exemplified by the accomplishments of OpenAI's ChatGPT chatbot, acquires the ability to perform tasks through the analysis of historical data and generates fresh content such as text and images based on this training.

The committee recommends performing a security evaluation on the content utilized to train publicly accessible generative AI models. Content exceeding "5% in the form of unlawful and detrimental information" will be designated for blacklisting. This category includes content advocating terrorism, violence, subversion of the socialist system, harm to the country's reputation and actions undermining national cohesion and societal stability.

The draft regulations also emphasize that data subject to censorship on the Chinese internet should not serve as training material for these models. This development follows slightly more than a month after regulatory authorities granted permission to various Chinese tech companies, including the prominent search engine firm Baidu, to introduce their generative AI-driven chatbots to the general public.

Since April, the CAC has consistently communicated its requirement for companies to provide security evaluations to regulatory bodies before introducing generative AI-powered services to the