Fact Checked by 
A chatbot trained on the thoughts of China’s President Xi Jinping has been created by the country’s top internet regulator.
The large language model (LLM) has been launched by the China Cyberspace Research Institute which operates under the Cyberspace Administration of China, the national regulator.
With people in China unable to use the ever-popular ChatGPT tool, this attempt is aiming to keep residents in on the AI action without using a non-Chinese model.
The draconian nature of the country’s controls has made its way into the chatbot as the closed AI system is based on Xi Jinping’s political philosophy and is aligned completely with the official government narrative.
The philosophy is known as ‘Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era’ and this includes all instructions on political, social, and economic life. It became part of China’s constitution in 2018.
Since then, his philosophy has become enshrined in Chinese culture as school children as young as 10 even have to study his politics.
According to a post on WeChat by the administration, as reported by the South China Morning Post, the LLM is described as being able to meet “a wide range of users needs” and can answer questions, outline reports, summarize information, and translate between Chinese and English.
The system has been created on the servers within the China Cyberspace Research Institute where “all data is processed locally, ensuring a high level of security.”
The model isn’t yet available to a wide and public audience as it’s still undergoing testing, but some designated users have been invited to try it out.
Chinese-based developers control generated content relating to Xi Jinping
Other technology giants like Baidu and Alibaba have generative AI chatbots and any generated content about the President is strictly controlled.
Both of the systems tend to ask users to restart chats when sensitive topics are brought up.
In light of this, the non-profit Cyber Security Association of China released a public database of 100mn entries of data that groups can use in model training.
