Chatbots pose a particular threat, as their ability to generate their own responses may encourage users to question the authority of the party. In this context, new rules were introduced in November requiring AI to be trained on politically filtered data and to pass an ideological test before entering the market. Additionally, all materials created using AI must be clearly labeled and traceable to simplify the search and punishment of those disseminating undesirable content.
Sources familiar with the government's position claim that the authorities do not intend to tighten regulation too much to avoid stifling innovation and losing competitiveness in the global AI arena, especially in light of the more liberal approach of the U.S. to regulation.
However, Beijing cannot afford to completely let go of the situation with AI. Earlier this year, Xi Jinping noted that AI carries "unprecedented risks," and one of his deputies compared the lack of safety measures to driving on a highway without brakes.
It is observed that China is finding a balance between control and innovation—Chinese models rank highly in international ratings and achieve success in specific areas, including computer programming, despite censorship regarding sensitive topics such as the Tiananmen Square massacre and human rights issues. At the same time, many American AI models remain unavailable in China.
Researchers from outside China studying both Chinese and American models note that China's regulatory approach may have its advantages: its chatbots are generally safer on several parameters, including a lower likelihood of exhibiting violence, pornography, and incitement to self-harm.
