China’s top internet watchdog announced Tuesday new draft rules for generative artificial intelligence — the government’s first major regulatory response to the rapidly advancing technology that has taken the country by storm.  

The draft rules from the Cyberspace Administration of China come as Chinese tech giants have rushed to develop intelligent chatbots that simulate human dialogue, after OpenAI’s ChatGPT triggered a new AI craze in the country. On the same day that the draft rules were released, Chinese tech giant Alibaba unveiled its large language model, Tongyi Qianwen, a month after rival Baidu’s Ernie Bot made its lukewarm debut.

The explosion of AI-generated content has been met with mounting public concern over privacy and the spread of misinformation online, with legal experts warning of risks from massive data sets and a lack of fact-checking mechanisms. The government is pouring billions into AI research and pushing for its adoption across society.    

According to the draft rules, available for public comment until May 10, all generative AI products “that utilize algorithms, models, and other rules to produce content such as text, images, audio, videos, and code” must undergo a security review by the internet watchdog before they are released to the public. Content must be “true and accurate,” and measures should be introduced to prevent discriminatory content. 

Service providers must require users to provide “real identity information” and introduce measures to prevent users from becoming “overly dependent or addicted” to the content. Violations could lead to criminal charges and a maximum fine of 100,000 yuan ($14,500). 

The new rules quickly garnered attention on Chinese social media following their announcement, with many surprised by the quick response from the authorities and some voicing concerns about the difficulty of implementing the rules. 

“The rules are correct but the real issue is that (artificial) neural networks are like a blackbox,” a user commented on microblogging platform Weibo. 

Although ChatGPT and other AI chatbots have enjoyed widespread popularity in recent months, public concerns have grown following a string of scandals linked to the technology. In February, a ChatGPT-generated text imitating a government announcement resulted in public confusion in the eastern city of Hangzhou, forcing police to investigate. Copycat services have also proliferated online. 

In January, the CAC introduced a new regulation targeting deepfake technologies amid rising popularity and misuse. The internet watchdog directed platforms to set up a real-name verification system for users and obtain consent for services that use biometric information. 

The Biden administration also announced Tuesday that it is exploring ways to regulate generative AI like ChatGPT. 

Editor: Vincent Chow. 

(Header image: Yuichiro Chino/Getty Creative/VCG)