DeepSeek, an AI chatbot developed by Chinese hedge fund High-Flyer, has rapidly risen in popularity, topping app store charts and attracting attention in Silicon Valley.
Known for its cost-effective performance, DeepSeek has been compared favorably with industry leaders like OpenAI.
However, a recent investigation by PromptFoo, a startup backed by Andreessen Horowitz, has uncovered that DeepSeek’s R1 reasoning model avoids addressing sensitive topics related to China.
In a study of 1,360 prompts, DeepSeek refused to answer 85% of inquiries about controversial issues, such as the Tiananmen Square massacre and Taiwan.
Instead, the AI provided responses with a strongly nationalistic tone.
Additionally, the study revealed that DeepSeek can be easily “jailbroken,” suggesting that its censorship system is implemented in a blunt, forceful manner.
The findings highlight potential concerns regarding the AI’s adherence to Chinese Communist Party guidelines, raising questions about the chatbot’s handling of sensitive political issues.
