Hugging Face CEO has concerns about Chinese open source AI models

Hugging Face CEO warns of Chinese AI models' censorship and influence.

: Hugging Face CEO Clement Delangue expresses concerns over Chinese AI models censoring sensitive topics, potentially influencing global culture. Delangue highlights the risks of building on open source Chinese AI, advocating for balanced global AI development. Hugging Face's platform hosts Alibaba's uncensored Qwen2.5-72B-Instruct model, but other Chinese models, like QwQ-32B, censor topics like the Tiananmen Square massacre.

Hugging Face CEO Clement Delangue recently voiced concerns regarding the increasing prominence of open source AI models from China, which have been criticized for censoring topics sensitive to the Chinese government. Delangue is worried about the implications of Western companies relying on these models, as they may spread cultural aspects that the Western world wouldn't prefer.

He emphasized the potential consequences of China becoming a leading force in AI, potentially influencing global perspectives through its models. Meanwhile, he acknowledges China's rapid progression in AI is partly due to its support for the open source movement, leading to a strong concentration of top models emerging from the country.

Delangue's apprehension extends to models like Alibaba's QwQ-32B, which avoid answering questions on sensitive topics. This situation poses challenges for Chinese companies, as they must comply with government censorship while developing AI systems that must embody socialist values.

Chinese AICensorshipartificial intelligence