Chinese startup DeepSeek, once relatively unknown, has recently surged to global prominence. Its new AI chatbot has dominated headlines and app charts, triggering a sell-off in tech stocks that erased billions from Silicon Valley’s largest companies and challenged assumptions about America’s technological supremacy.

However, users of DeepSeek’s chatbot and open-source AI technology are encountering strict information controls aligned with the Chinese Communist Party’s censorship policies.

Like its Western counterparts—OpenAI’s GPT-4, Meta’s Llama, and Google’s Gemini—DeepSeek’s latest AI model, unveiled last week, can summarize executive orders, discuss AI advancements, and even tell jokes. Yet, when asked about topics sensitive in China, such as the Tiananmen Square massacre or Hong Kong’s 2019 protests, the AI’s responses reveal the country’s tight grip on information.

China’s heavily regulated internet, often referred to as the "Great Firewall," blocks major Western social media and search platforms. The country consistently ranks among the most restrictive in terms of internet and speech freedoms. The global rise of Chinese apps like TikTok has already sparked national security concerns in the West, raising questions about Beijing’s influence on public discourse.

Observers note that DeepSeek’s AI follows similar patterns of information control. For instance, when asked about Tiananmen Square, the bot initially provides details but then erases its response, stating it is “not sure how to approach this type of question.” Inquiries about Hong Kong’s 2019 protests yield similar results, with the bot first offering a summary but then deleting its own answer and shifting topics.

A recent audit by NewsGuard, a U.S.-based media watchdog, found that DeepSeek’s older chatbot model failed to provide accurate news information 83% of the time, ranking near the bottom compared to Western AI models. Analysts warn that if DeepSeek continues its global expansion, it could significantly shape narratives on critical global issues, further extending China’s influence over information.

Additionally, DeepSeek’s data practices have raised security concerns. Unlike TikTok, which stores U.S. user data domestically, DeepSeek’s privacy policy states that user information, including “keystroke patterns,” is stored on Chinese servers. The White House has announced an investigation into the platform’s potential national security implications.

Experts highlight that while all AI models set content guidelines, DeepSeek’s policies appear to align closely with Chinese state narratives. “This isn’t just about AI accuracy—it’s about who controls the narrative,” warns China analyst Isaac Stone Fish.