AI Photo:VCG

Interim rules for managing generative artificial intelligence (AI) services will take effect in China on Tuesday, marking the nation's latest effort to promote the sound development of the sector while safeguarding national security and the public interest.

AI-generated content (AIGC) providers should carry out data processing including pre-training and optimization training in accordance with Chinese laws, according to interim measures announced by the Cyberspace Administration of China (CAC) and six other ministries in July.

AIGC providers should not infringe on others' intellectual property rights and should obtain consent from individuals if personal data is involved.

The regulations will be applied to AIGC firms that provide text, image, audio and video content to the Chinese public. Enterprises, scientific research institutions and universities that carry out scientific research using the technology will be exempt from supervision. Other AI-related sectors including autonomous driving will not be regulated. 

Promoting the development of new technologies including AI was an important focus during the measures' draft stage, Luo Fengying, an official at the CAC, said at the Internet Security Conference 2023 in Beijing last week.

Luo emphasized that the current work focus is to convert troves of data into forms that can be used for training AI large language models (LLMs), as a major bottleneck in developing AIGC is the issue of data sets - that is, how to guarantee their authenticity and accuracy.

LLMs refer to the technology that is used to train AI chatbots like the popular ChatGPT developed by US firm OpenAI.

"China is the first major economy to announce such a regulation for the fast-developing AIGC sector. The swift response will reassure companies in the sector," Lu Chuanying, director of the Research Center for International Cyberspace Governance under the Shanghai Institute for International Studies, told the Global Times on Monday.

Previously, Chinese technology companies tended to hesitate to enter the sector amid concerns over policy risks, according to Lu.

"Now with the regulatory measures in place, they can have a clear direction and better expectations in developing the technology," Lu noted, adding that some companies plan to expand investment in LLMs.

The implementation of China's AIGC regulations will usher in a wave of LLM commercialization, according to a research by Ping An Securities. 

An array of Chinese firms ranging from search engine giant Baidu and e-commerce platform JD.com to emerging start-ups have unveiled LLMs in recent months.

Liu Xingliang, director of the Beijing-based Data Center of China Internet, told the Global Times on Monday that the number of Chinese enterprises that have announced LLMs or plan to do so may exceed 800, reflecting the rush to follow the ChatGPT frenzy earlier this year.

China and the US are the two leading forces in AIGC development, accounting for more than 80 percent of the world's LLMs, and the US leads the race, while China is catching up quickly, according to Liu.

With more demand for AI chips, US chipmakers have released modified chips specifically for Chinese customers amid the US government's assault on China high-tech companies. 

Nvidia has received orders worth $1 billion from Baidu, ByteDance, Tencent and Alibaba for 100,000 A800 processors, the Financial Times reported earlier this month, citing sources familiar with the matter. These orders are expected to be delivered this year.

On July 11, Intel launched the Habana Gaudi 2, a deep-learning accelerator card designed for the Chinese market that is aimed at enhancing AI training and inference capabilities.