China’s internet watchdog proposes rules for AI tools

China’s internet watchdog has unveiled a new set of draft rules targeting ChatGPT-like services, as governments around the world move to rein in the rapid development of generative artificial intelligence (AI) tools.

Companies that provide generative AI services in China must take measures to prevent discriminatory content, false information, and content that harms personal privacy or intellectual property, according to the proposed regulation published by the Cyberspace Administration of China (CAC) on Tuesday.

Businesses should also make sure that their products uphold Chinese socialist values, and do not generate content that suggests regime subversion, violence or pornography, or disrupts economic or social order, the CAC said.

All generative AI products should pass a security assessment by the CAC before serving the public, as required by a 2018 regulation covering online information services that have the ability to influence public opinion, the regulator said.

Generative AI – which creates original content based on text, image or audio prompts – has garnered growing enthusiasm in China after Microsoft-backed US start-up OpenAI launched its conversational bot ChatGPT in November.

In China, where ChatGPT remains officially unavailable, domestic companies have been racing to launch technologies similar to ChatGPT.

They include web search operator Baidu, which unveiled its Ernie Bot last month, and e-commerce giant Alibaba Group Holding, which is planning to integrate its Tongyi Qianwen into all of its products. AI firm SenseTime also launched its latest set of large AI models named SenseNova earlier this week.

Alibaba owns the South China Morning Post.

As interest in generative AI surges, however, Chinese authorities have cast a wary eye on the risks posed by the rapidly developing sector. State-run media have repeatedly warned of a “market bubble” and “excessive hype” surrounding generative AI tools, with one newspaper recently cautioning that ChatGPT may corrupt, rather than improve users’ moral judgment.

On Monday, the Payment & Clearing Association of China, governed by the country’s central bank, urged industry workers to be mindful of the risks involved in using ChatGPT-like tools for work and to avoid uploading confidential and sensitive information to such services.

Other countries are also mulling regulations on generative AI.

The US commerce department on Tuesday began inviting public comments on how policymakers should approach increasingly sophisticated AI tools, while Italy this month became the first Western nation to ban ChatGPT over privacy concerns.

The European Union meanwhile is discussing the Artificial Intelligence Act, which was first proposed in 2021 to govern the use of AI products according to their risk levels.

The EU framework, which follows a different approach than the CAC’s draft rules, is likely to impose “extremely onerous compliance burdens on companies”, said Angela Zhang, associate law professor at the University of Hong Kong.

Instead, China’s draft regulation focuses mostly on content moderation, she said.

“These content requirements are not new to Chinese internet companies, so I don’t think the publication of these new rules will add too onerous a burden to Chinese companies,” Zhang said.

Still, some experts have pointed to censorship as a potential roadblock for Chinese companies hoping to create a true rival to ChatGPT.

“Excessive restrictions, content regulation, and censorship could hinder commercialisation and further innovation of such technologies,” Hanna Dohmen, a research analyst at Georgetown University’s Center for Security and Emerging Technology, said in February.

The CAC is soliciting feedback on its proposed rules until May 10. South China Morning Post