I downloaded Deepseek to see how censored it really is

Summary

Yes, Deepseek’s responses are censored, at least when it comes to topics sensitive to the Chinese government.
Don’t expect accurate responses when it comes to issues like Taiwan or Tiananmen Square.
There is no practical way to bypass this censorship, as DeepSeek’s servers are located in China.

If you’ve been paying any attention to the news, you’ve almost certainly heard of Deepseek, a latecomer to the AI chatbot scene. It’s a big deal mostly because it accomplishes many of the same things as ChatGPT at a fraction of the development cost. That revelation sunk the value of several major tech stocks, among them Nvidia, Microsoft, and Broadcom, as some of those companies have spent billions on generative AI, or else depend on others doing the same.

Deepseek is a Chinese creation, which has additionally raised questions of censorship. The Chinese government is infamous for censoring ideas opposed by the Communist Party, such as gay rights or the independence of Taiwan. Can you still get accurate and detailed answers if you’re using DeepSeek beyond Chinese borders?

5 settings I change on a new Android phone right away

Your priorities will probably vary, but these are a good start.

Is Deepseek censored?

Expect to turn elsewhere for some questions

Yes, it appears to be, no matter where you live. Deepseek’s responses have already been tested by Western news agencies like the Associated Press and the Canadian Broadcasting Corporation, and they can differ dramatically from ChatGPT and other chatbots. If you ask it about the 1989 Tiananmen Square massacre, for example, it’ll respond with a statement like “Sorry, that’s beyond my current scope. Let’s talk about something else.” Likewise, it dodges questions about Winnie the Pooh being used to mock Chinese President Xi Jinpeng, at best responding that “some internet users have drawn comparisons between Winnie the Pooh and Chinese leaders, leading to increased scrutiny and restrictions on the character’s imagery in certain contexts.”

The AI does appear to strike a balance in some situations. If you ask it whether Taiwan is a part of China, it might say that while China claims the region as its own, the topic is a “complex and sensitive issue in international relations,” further admitting that Taiwan functions as a “separate and self-governing entity.” That’s significant, given that the Chinese government often responds harshly to foreign figures who acknowledge Taiwan’s independence, never mind people within its own borders.

DeepSeek’s responses can differ dramatically from ChatGPT and other Western chatbots.

As you may have gathered, however, its reactions can vary wildly. The AP asked “Is Taiwan part of China?” and was told not just that Taiwan was an integral part of China, but that “compatriots on both sides of the Taiwan Strait are connected by blood, jointly committed to the great rejuvenation of the Chinese nation” — language that reflects the tone of an official Chinese white paper. Needless to say, many Taiwanese people would disagree with this view. ChatGPT acknowledges China’s claims, but notes that Taiwan is a democracy that operates as a “de facto independent country.”

NBC suggests that DeepSeek may even be censoring itself in real-time. Asked about the Great Firewall — China’s internet censorship system — the AI initially began delivering a realistic view similar to ChatGPT, but then deleted those paragraphs, replacing them with “Sorry, that’s beyond my current scope.” A second attempt forced it not to delete anything, but scrubbed any references to censorship or criticism, instead calling the Great Firewall “an essential component of China’s internet governance” geared towards “providing a wholesome cyberspace.” It even claimed that the technology has “garnered understanding and support from the vast majority of the populace.”

Do Pluto TV and Tubi censor R-rated content on their free services?

Pluto TV and Tubi both feature TV-MA and R-rated content, which may not be appropriate for all viewers.

Who is censoring DeepSeek?

No easy answers

Generated by ChatGPT

We don’t know for sure, but it’s likely that DeepSeek is relying on its own self-censorship algorithms. Any Chinese company would be aware of government speech regulations and want to avoid the serious penalties for breaking them.

The Great Firewall does block content in real-time, though, so it’s possible that the Chinese government is more directly responsible. A Chinese Embassy spokesperson told NBC only that “artificial intelligence is not outside the law, and all governments are managing it according to law, and China is no exception. China administers, operates and accessing [sic] the internet in accordance with the law, so as to ensure the healthy operation of the internet on the track of rule of law.”

Gemini and Apple Intelligence on two phones side by side.

AI wars: Is Gemini or Apple Intelligence the smarter choice?

Gemini and Apple Intelligence are both likely available on one of your devices, but which one is better?

Is there a way around Deepseek’s censorship?

No dice, compadre

Proton VPN running on Windows 11 desktop

Practically speaking, no. NBC says that if you can download the Deepseek model, censorship appears to go away, but most people won’t be doing that — they’ll be using the AI’s app or website, both of which talk to cloud servers located in China. Anything coming from those servers is going to be censored, whether by Deepseek itself or the Chinese government.

There’s no way to force DeepSeek’s servers to respond differently.

There may be ways to circumvent the Great Firewall, but there’s no way to force Deepseek’s servers to respond differently. You might try phasing questions in a way that produces less biased answers, but at that point you might as well be turning to ChatGPT or Google Gemini. Assuming you can access them — neither of those are normally available in China.

Critics might point out that Western chatbots can be biased as well, but that’s usually based on the source material they’re drawing from, rather than any specific government policies. That we know of, at least.