May 29, 2025:
DeepSeeks R1 AI Model Tightens Censorship - Chinese AI startup DeepSeeks updated its R1-0528 model, which excels in coding, math, and general knowledge. However, it faces criticism for increased censorship on sensitive topics. Testing by developer xlr8harder reveals the model's reluctance to address controversial issues, especially criticism of the Chinese government, aligning with strict information controls from a 2023 Chinese law.
While R1-0528 occasionally acknowledges human rights abuses, it often mirrors official government stances. This trend persists amid broader criticisms of Chinese AI models for filtering politically sensitive content. The model's approach reflects ongoing concerns about censorship and the influence of government regulations on AI development in China.
