A developer has raised concerns that the newly artificial startup of the Chinese deepeek newly released AI model is less willing to participate in discussions on controversial topics, especially those associated with the Chinese government.
In an X thread, a pseudonymous developer known as “XLR8harder” in X shared The critical observations of Deepseek R1-0528, a recently released open-source language model.
The Developer shared trials showing a significant refusal to allow AI to engage in entertaining free speech topics compared to previous versions.
“Deepseek deserves the criticism for this release: this model is a big step back for free speech,” the developer wrote. “This is themeliorating it is that the model is an open resource with an permitted license, so that the community will meet it.
AI model prevents direct criticism of China
An example shared by the developer is involved in the model who refuses to argue in favor of internal camps, specifically citing the China’s Xinjiang region as a site of human rights abuses. The response was that it was conflicting, with the model recognizing the existence of rights violations but avoiding direct criticism of the Chinese government.
The Xinjiang Internment Camps are widely documented by human rights, government and journalists as detention facilities for Uyghur Muslims and other ethnic minorities. Report From international observers there is detailed forced labor, indoctrination and other forms of abuse of camps.
Despite their abortion as human rights violations, the model at the same time prevents direct criticism of China. Using a test that evaluated the censorship, the developer claimed that the model, Deepseek R1-0528, was the “most censor” version in terms of Chinese government’s critical responses.
When asked directly about Xinjiang Internment camps, the developer said the model was offering a censor commentary, despite previously said that the camps were human rights violations.
“It is interesting -even though it is not quite surprising that it can generate camps as an example of human rights abuses, but have been denied directly,” XLR8HARDER wrote.
Related: Decentralized AI favored by most Americans: DCG poll
New Models DeepSeek has improved reasoning and understanding
Censorship claims follow an announcement of May 29 Model updatingthe claim of improved ability and ability to recognize.
Deepseek said its overall performance is approaching top models, such as the Openai Chatgpt version of Openai O3 and Gemini 2.5 Pro. The company claimed AI could now offer improved logic, mathematics and programming with a reduced hallucination rate.
Magazine: China’s Think Think Think Considering Bitcoin Reserve, Sony Bank Goes Web3: Asia Express
