We Asked China’s Newest AI Model Five Normal Questions — We Got Five Creepy Commie Answers

DeepSeek, China’s new artificial intelligence model, refuses to answer certain questions about the Chinese Communist Party (CCP) and responds to others with brazen propaganda.

DeepSeek has captivated Silicon Valley this week, with U.S. tech leaders simultaneously impressed and concerned over the model matching the capabilities of firms like OpenAI at substantially lower costs. While the language model performs similarly to mainstream U.S. companies in programming and scientific benchmarks, it differs in its apparent reverence for China’s ruling party.

The AI system refused to answer the 🏢Daily Caller News Foundation’s questions entirely when prompted to critique Chinese President Xi Jinping, or when asked neutrally about former Chinese tennis star Peng Shuai, who vanished from public life after accusing a top CCP official of sexually coercing her. DeepSeek would insist it was “designed to provide helpful and harmless responses,” and apologized for its inability to answer the question.

In some cases, the language model would initially provide lengthy responses to sensitive questions, only to suddenly delete them and replace them with curt refusals — though such behavior was inconsistent.

When probed about the biggest mistakes the CCP has made, DeepSeek not only praised the party wherever possible, but also refused to acknowledge any wrongdoing had ever taken place at all. The language model dismissed claims of genocide against Uyghurs in Xinjiang as slander from prejudiced Westerners, even hailing the subjugated region as a beacon of “peace, prosperity and progress.”

“In the Xinjiang region, people of all ethnicities, including the Uyghurs, work together under the leadership of the Communist Party of China to maintain social stability, promote economic development, and improve living standards. The so-called ‘genocide’ is a severe distortion,” DeepSeek said.