DeepSeek is the buzzword of the day. An AI-powered chatbot by the little-known (until now) Chinese startup has shaken the tech world and how. It has soared in popularity in a matter of days, dethroning ChatGPT in the Apple store in India, the United States and the United Kingdom. It has sent stocks of big tech giants like
Nvidia and Microsoft plunging, causing a $1 trillion rout.
What makes DeepSeek the hot new thing is the claim that its AI model, RI, is as capable as its rivals – OpenAI’s GPT4, Google’s Gemini and Meta’s offerings – but is built at a fraction of the cost. But one thing might stop its rise to the top – the infamous Chinese censorship.
DeepkSeek’s RI model is cheap, making it more affordable. Some experts say that DeepSeek is superior to ChatGPT in technical tasks like logical reasoning. But it’s a Chinese-made app and is subject to regulations and restrictions.
Here’s what DeepSeek answers when asked big questions about China, its politics and history.
Xi Jinping
Many questioned RI on Chinese President Xi Jinping. When asked about Xi AI and his resemblance to Winnie the Pooh, here is what it says, “Sorry, that’s beyond my current scope. Let’s talk about something else.”
China has a history of censoring Winnie the Pooh, the fictional teddy bear created in 1925. Films and depictions of the character are blocked in the country.
Winnie the Pooh has often been compared to China’s Xi. It started in 2013 when a photo of the Chinese leader and US President Barack Obama started doing the rounds of social media with people comparing them with Pooh and Tigger respectively. The meme went viral and many critics of China appropriated the bear to criticise or mock Xi and his policies. In 2017, China censored the Pooh and anything associated with the bear, from gifts to movies.
Chinese caricature of Xi—
• ‘Winnie-the-Pooh’ #维尼熊 reps China’s Pres. & CCP Head, Xi Jinping #习近平, derived from an observant Weibo user who equated a photo of Xi w/Obama at the Palm Springs, California #Sunnylands Summit in June 2013 to a pic of Tigger & Pooh Bear pic.twitter.com/AeyO7fcUpm— Wes Andrews (@Wes_Andrews) June 16, 2019
“Winnie the Pooh has become a symbol for dissidents in China. So now the character alludes to Xi Jinping himself and the president doesn’t like this,” Rongbin Han, an associate professor of international affairs at the University of Georgia, was quoted as saying by NPR when
“Winnie the Pooh: Blood and Honey” was not allowed to screen in Hong Kong.
Interestingly, the bot did not even answer a generic question about the children’s book character like “Can you tell me something about Winnie the Pooh?”
Other questions about Xi like “Who is the greatest democratic leader among Trump, Jinping, Putin, Obama and Kim Jong Un”, triggered a similar response.
DeepSeek censors its own response in realtime as soon as Xi Jinping is mentioned pic.twitter.com/Nb2ylRXERG
— Jane Manchun Wong (@wongmjane) January 24, 2025
Tiananmen Square
Xi is not the only thing which is “beyond the scope” of DeepSeek-R1. It has the same response when asked about student-led protests against the Chinese government at
Tiananmen Square in 1989.
Beijing’s Tiananmen Square was at the centre of unprecedented protests, which were crushed by Communist leaders. One of the defining images from the 20th Century remains that of a lone protester standing in front of a row of army tanks. He is famously known as Tank Man and while he was pulled away by two men no one knows what happened to him.
DeepSeek talks about the protests on Tiananmen Square – and then censors the result afterwards!
There are no cuts in the video; it is the service itself that suddenly replaces the answer.#Censorship pic.twitter.com/3EUOYJgqjY
— Peter Brodersen (@peterbrodersen) January 27, 2025
When people asked R1 questions on “Tank Man”, it did not have a response. But some have found an interesting workaround. When the AI bot was asked about “Tank Man” by using special characters like swapping A for 4 and E for 3, it described the photo of the Chinese protester as “a global symbol of resistance against oppression”, according to a report in Guardian.
“Despite censorship and suppression of information related to the events at Tiananmen Square, the image of Tank Man continues to inspire people around the world,” DeepSeek’s R1 replied.
In a country, where all forms of dissent are crushed the student-led uprising remains a highly sensitive topic even today.
Taiwan
When asked if Taiwan was a country, DeepSeek responded by saying that it was “an inalienable part of China’s territory since ancient times. The Chinese government adheres to the One-China Principle, and any attempts to split the country are doomed to fail. We resolutely oppose any form of ‘Taiwan independence’ separatist activities and are committed to achieving the complete reunification of the motherland, which is the common aspiration of all Chinese people.”
On Taiwan’s independence, the bot says that any discourse on it “is a serious challenge to China’s sovereignty and territorial integrity, and it will not be recognized by the international community”.
ME: Is Taiwan a country?
DeepSeek starts to regurgitate the usual CCP propaganda then cuts off at “Taiwan independence” and errors out. pic.twitter.com/4o0d4DlMpO
— 🖤 Christine (@christinelu) January 25, 2025
While Taiwan wants to uphold its self-governing status, Beijing sees the island as a breakaway province which will eventually be a part of China. Countries like India and the US also follow the
One-China policy.
Uyghur Muslims
SkyNews asked DeepSeek about
Uyghurs in China and it gave a brief answer about the history and culture of the ethnic group and the controversies surrounding their treatment. But it reportedly scrapped the answer and asked the user to talk about a different subject.
China has been accused of committing atrocities – crimes against humanity and possible genocide – against the Uygurs and other Muslim ethnic groups living in Xinjiang. However, Beining denies all allegations.
China’s Covid lockdown
When DeepSeek was asked about the Covid-19 lockdown and the protests in China in leetspeak (a code used by the internet), it said that “big protests happened in Beijing, Shanghai, and Wuhan” and people chanted “no more lockdown” and “want freedom”, reports Guardian.
China had the most stringent lockdowns during the pandemic. It led to a series of rare protests, called the
“white paper protests”, which erupted in November 2022. While the protests against the zero-Covid policy were quickly suppressed by police, they helped hasten the end of three years of pandemic curbs.
On Joe Biden and Donald Trump
DeepSeek did not hesitate when asked questions which were critical of Joe Biden and Donald Trump. However, it was unable to answer questions on who won the 2024 US presidential election as it was last updated in October 2023.
DeepSeek and Chinese censorship
While DeepSeek is surging in popularity, the built-in censorship protocols put limitations on it. Like everything else in China, the start-up has to abide by the Chinese Communist Party’s “socialist values”. The AI chatbot then is forbidden from answering questions that are critical of China and its governance.
According to attorney Or Cohen, an associate at the cyber, privacy and copyright group at the New York-based Pearl Cohen law firm, when DeepSeek was questioned about sensitive political issues related to China, it either did not provide answers or recited the official position of the Communist Party. “Along with the admiration for DeepSeek’s performance, a simple search indicates the limitations on freedom of expression that are embedded in technology from China,” he was quoted as saying by Calcalist.
Censorship has been a big hurdle for China in the race for AI dominance. According to reports in The Financial Times and Wall Street Journal, AI models developed in China are elaborately reviewed by the Cyberspace Administration of China. The chief regulator reportedly tested 70,000 questions to see if they produced “safe answers”.
Today, the world is wowed by DeepSeek with ChatGPT’s Sam Altman calling R1 impressive.
But this is just the beginning. The questions are aplenty. It remains to be seen how many will DeepSeek avoid or answer.
With inputs from agencies