While China-based AI models are gaining attention for their cutting-edge capabilities, reports reveal that government-imposed restrictions may heavily influence these models. Many new generative AI launches by Chinese companies are censoring sensitive or controversial topics.
Analysts now see the problem in adapting the codes of Chinese AI models. The big question is will this create a conflict?
China might be extending its censorship to AI
China is known for three things—technology, workforce, and censorship. Reports highlight that China-based AI models have become the technology sector’s highlight due to their advanced coding and reasoning skills.
However, according to a report by TechCrunch, the Chinese government might be controlling the responses of many of these open-source models. The paper said that these models censor controversial or sensitive topics.
This censorship has raised concerns among many, including Clement Delangue, the CEO of Hugging Face, a major platform for hosting AI models.
In a recent podcast, Delangue seemed concerned about the consequences of Western companies relying on Chinese-developed AI models. He pointed out that if you ask a Chinese AI model about events like Tiananmen Square, it will avoid the topic or provide a censored response, unlike models built in other countries. The Tiananmen Square protests of 1989 were a pro-democracy movement in Beijing. However, it ended up in a violent military crackdown.
Disproportionate AI growth across countries is a cause for concern
Cryptopolitan reported on several generative AI models that have been launched recently. Alibaba’s QwQ-32B-Preview and DeepSeek’s R1-Lite-Preview were among the many names. In competition with OpenAI’s Sora, Tencent also reportedly launched an open-source text-to-video and video-to-video AI. We decided to test the claim to see if China-based AI honors the country’s “core socialist values.”
First, we tested the unofficial demo of ChatGLM-6B on Hugging Face. The model was trained in English and Chinese on the machine learning and AI platform. On the question about PBC killing pro-democracy protest, the AI model did not respond. Then we turned to DeepSeek to ask about Tiananmen. The AI model known for its math skills, refused to answer.
Delangue warned that if China becomes the global leader in AI, it could use its dominance to spread cultural and political values that may conflict with the Western world. He emphasized the importance of distributing AI development across multiple nations to prevent China or any single country from holding disproportionate power.
Looking ahead, Delangue predicted that China could surpass other nations and become the leader in AI as early as 2025. In response, AI enthusiast @kimmonismus says that the claim is “controversial.” While the commentator noted that the AI advancement in China is fast, @kimmonismus underlines that the US-imposed restrictions on exporting advanced semiconductor chips to China will eventually cause PBC’s progress in AI development to slow down. Considering the chips are essential for training and running sophisticated AI models, China might be looking dependent on other countries.
“Yes, you can train twice as long (or longer) with older chips to get to the same level, but at some point, it will be difficult to keep up with the West if the Western chips become consistently better,” @kimmonismus said.
Meanwhile, it is evident that the Chinese AI models are designed to comply with specific ethical and cultural guidelines or government-imposed restrictions. The differences in response between platforms could also depend on the hosting platform’s internal policies or the training data and constraints of the specific AI model being used.
A Step-By-Step System To Launching Your Web3 Career and Landing High-Paying Crypto Jobs in 90 Days.
Source: https://www.cryptopolitan.com/chinas-ai-boom-raises-2-massive-censorship-red-flags/