February 3, 2025:
DeepSeeks Censorship Persists Locally - Investigations by Wired and TechCrunch reveal that the DeepSeeks AI model includes censorship at its core, contradicting claims that it might be uncensored locally. Both the application and training levels impose restrictions, avoiding or minimizing topics such as the Cultural Revolution and Tiananmen Square.
A test of the locally run model from Groq showed a willingness to discuss U.S. events like the Kent State shootings but a refusal to comment on the 1989 Tiananmen Square protests. This demonstrates the built-in limits of DeepSeeks AI, regardless of its operating environment.
