Skip to main content

China wants homegrown AI to reflect the core values of socialism

China wants homegrown AI to reflect the core values of socialism


The CCP is drafting new rules to regulate generative AI systems as Chinese tech giants scramble to launch ChatGPT rivals.

Share this story

An image showing a repeating pattern of brain illustrations
Illustration: Alex Castro / The Verge

Chinese officials have signaled how generative AI may be regulated in the country, writing in recently published draft guidelines that the software “should reflect the core values ​​of socialism,” CNBC reports. The proposed rules were published by the country’s Cyberspace Administration of China as it seeks public feedback and provide key indications of how the country’s heavy-handed speech rules may be applied to locally developed competitors to ChatGPT.

As well as reflecting socialist values, the draft rules (translated via Google Translate) say generative AI should not attempt to subvert state power, overthrow the socialist system, or undermine national unity. Promoting terrorism, extremism, or ethnic hatred is also prohibited, along with distributing “false information” or content that might “disrupt economic and social order.” 

There are guidelines about how algorithms should be designed in a way that prevents discrimination based on race, ethnicity, gender, and other characteristics. And Bloomberg notes that the country plans to conduct security reviews of generative AI services before allowing them to widely launch in the country.

The publication of the draft guidelines comes as Chinese firms have signaled their intention to develop homegrown alternatives to OpenAI’s ChatGPT. Alibaba has Tongyi Qianwen, which it plans to integrate into its productivity software, while Baidu recently announced its own software called Ernie bot

Government restrictions mean China already has a distinct online ecosystem of apps and services. Western search and social media giants are largely barred from the country, allowing domestic firms to dominate the space. Now, it’s looking like this will continue into a new generation of generative AI services. Nikkei Asia previously reported that Chinese users have had to turn to VPNs in order to access the US-developed ChatGPT and that Tencent has banned third-party attempts to integrate the generative AI tool into WeChat. China Daily, the country’s biggest English-language newspaper, has warned of the potential for ChatGPT to “amplify US disinformation.” 

China is keen to promote the development of domestic generative AI alternatives amid mounting tensions with the US over its access to powerful technologies like high-end semiconductors needed to train AI systems. But there’s tension here with the country’s restrictive approach to censorship. “I see the Chinese regulators being quite cautious with its regulatory approach in order to give more room for the development of generative AI in the country,” University of Hong Kong law professor Angela Zhang tells Bloomberg

At least one Chinese generative AI tool has had limitations imposed on its ability to create content relating to political topics. Last year, MIT Technology Review reported that Baidu’s text-to-image generator ERNIE-ViLG refused to create images of Tiananmen Square or other political leaders and would instead label them as “sensitive.”

Such issues of censorship and speech moderation are not unique to China, though. In the US, ChatGPT has been criticized for being too “woke” because of restrictions OpenAI has placed on the software’s ability to generate harmful speech and its perceived liberal bias. The criticism has led to the launch of “uncensored” alternatives like FreedomGPT, whose creators claim isn’t “programmed to further any intellectual agenda.