From Cointelegraph

May 30, 2025 08:27 AM:

A developer criticizes Chinese AI startup DeepSeek’s new model for avoiding controversial topics, particularly those related to the Chinese government. Tests show a decline in the AI’s willingness to engage in free speech topics. Despite acknowledging human rights abuses, the model restrains from direct criticism of China, deemed the “most censored” version by the developer.

DeepSeek’s updated AI model claims enhanced reasoning and inference capabilities, approaching leading models like OpenAI’s ChatGPT version o3 and Gemini 2.5 Pro. The company touts improved logic, math, and programming abilities with a reduced hallucination rate. The update follows the developer’s criticism of censorship regarding topics like internment camps in Xinjiang.

Read more at Cointelegraph: Deepseek AI model faces criticism over censorship and free speech limits