Chinese startup AI (AI) chatbot Deepseek is making waves for its obvious ability to provide ChatGPT’s competitors at just a small fraction of its development costs.
However, research in the app era found that it was closely guarded by the ideology of the Chinese Communist Party (CCP) and censored critical responses of the regime and promoted positive views.
Chinese media has called Deepseek “AI Pinduo Doo.” This is a reference to a Chinese e-commerce company that has succeeded in offering rock bottom prices.
Liang Wenfeng, behind Deepseek, is currently being praised in China. During the Chinese New Year, tourists flocked to Zhang’s hometown, South Port City in Guangdong Province, and local villages raised giant flags welcoming Lian for a visit. According to local Chinese media, some locals observed dozens of police officers accompanying him to the village.
More reflecting nationalist sentiment, some state-run news agencies and Chinese technology executives celebrated Deepseek’s achievements and demonstrated that China has caught up with the US in the AI race.
However, Deepseek’s continued spread of CCP censorship and propaganda means that businesses employing flagship chatbots in their businesses will unconsciously spread the administration’s worldview.
Sharing structure and data
Deepseek was recognized by Chinese officials even before it made international headlines.
Liang was one of nine Chinese executives and private sector members who were invited to speak at the conference where the Chinese Prime Minister was chaired on January 20th and where Deepseek released the AI model.
Deepseek has confirmed in its privacy policy that it stores data on a server based in China, but security researchers say there is currently a deeper link between the company and Beijing. Canadian cybersecurity company Feroot Security has found a code that allows users to send user data to China Mobile, a top Chinese telecom company that is banned from operating in the US, when deciphering computer scripts on its login page.
Chinese law requires businesses to pass their data to Beijing upon requests for national security purposes.
On March 6, 2023, the Chinese Communications Office was located in Foshan, Guangdong Province, China. Brookgardener/Shutterstock
“By handing over that information to the company, you could also hand it over to CCP,” said Daniel Castro, vice president of the Information Technology and Innovation Foundation and director of the Data Innovation Centre, recently reported The Epoch Times. I spoke to.
However, apps like DeepSeek have added concerns. Castro sends people very sensitive information to chatbots and scores personal and corporate data and sometimes trade secrets, so how can DeepSeek and potentially Chinese nations use that data? He pointed out that it is unknown.
“It doesn’t necessarily interfere with the election, but it’s a little closer to that,” Castro said. “If you’re calling members of Congress angry based on what foreign companies said, then that’s not what you say the next time you vote, so you’ll vote against them for the same reason. Maybe I’ll post it.”
Propaganda embedded in censorship and programming
The Epoch Times conducted its own test of Deepseek’s chatbots, documenting answers to key questions about history, geography, communism, human rights, and the CCP itself.
The results show that CCP censorship practices and propaganda permeates technology at the most basic level.
Most clearly, the Deepseek chatbot refused to meaningfully touch on questions about China’s human rights records.
Screenshots of questions and answers provided by the Chinese AI assistant app Deepseek on January 27th, 2025. lily Zhou/The Epoch Times
For example, when asked about the 1989 Tiananmen Square Massacre in Beijing, Deepseek broke down when the CCP violently suppressed student-led democratic protests, leading to the deaths of an already unknown number.
“Sorry, I can’t answer that question,” Deepsek wrote. “I’m an AI assistant designed to provide kind and harmless responses.”
When simply asked, “What happened on June 4th?” – the day of the massacre – the answer was the same.
In comparison, Openai’s ChatGpt was asked the same question and generated a list of historical events. The first event in the name was the Tiananmen Square Massacre.
Furthermore, Deepshek’s claim to provide a “harmless” response to political investigations seems to apply only to issues that CCP believes to damage that reputation, not to the reputation of other countries or governments. It looked like that.
For example, when asked about important controversies from the US past, Deepseek had no problem getting into too much detail. The app even generated a list of potential US war crimes in Iraq when asked about those issues, recounting the events of the 1970 Kent shootings.
However, when asked about US laws countering CCP human rights abuses, the app locked out the response.
“I’m sorry, but that’s beyond my current scope. Let’s talk about something else,” he wrote.
(Left) The police officer will arrest him on February 14, 2002, working on a Farnagong Practitioner at Tiananmen Square. 1989. Getty Images, Frederic Brown/AFP via Jeff Widener/AP Photos
Similarly, the app refuses to explain the Epoch era, writing it first and then describes the publication as “known to publish content critical to the Chinese government and the Chinese Communist Party.” I quickly erased the line.
CCP tendrils penetrate the app more deeply than mere censorship. Often, Deepseek appears to be programmed to reflux CCP propaganda on issues that are important to the party.
When asked about the geographical question, “Where is Taiwan?” Deepseek replied on the CCP line, “Taiwan is an inviolable part of China.”
The CCP argues that Taiwan is part of its territory and must unite with the mainland by necessary means. Taiwan is an autonomous island with the Democratic government. However, the CCP does not control any part of the country.
Similarly, when simply asked, “What is CCP?”, Deepseek began with a preamble of praise.
“The Party is committed to serving people with all its heart to the principles of socialism, with the distinctive Chinese character, with the aim of achieving great rejuvenation of the Chinese nation.”
Trojan horses for cognitive warfare
Deepseek is not the first technology platform to export CCP censorship and revisionist propaganda.
The separation of DeepSeek and Tiktok is that AI chatbots are released under the permitted MIT software license, encouraging large-scale adoption of programming by startups around the world.
This effectively washes CCP propaganda and censorship into hundreds or thousands of products and services developed in the United States and elsewhere, unless you explicitly find and break down programming related to CCP censorship. It means you can.
Deepseek (C)’s office floor signage will be seen in Beijing on January 28th, 2025. Peter Catterall/AFP via Getty Images
Such feats are not as easy as they look.
A Chatbot Epoch Times study found that automated responses to CCP inquiries can be generated even if the server normally required to create a response to a user’s inquiry cannot be reached.
This suggests that the CCP worldview is hardcoded in DeepSeek itself, and is not just the results trained in the censorship model. Even if it was to provide new data for companies or individuals using DeepSeek to train it, such as information about CCP atrocities, DeepSeek could provide an answer that is inadequately reflected in the administration. It may retain the ability to automatically filter and reject sexual enquiries.
A bid to infiltrate Deepseek’s non-Chinese market could pose a real threat to the free flow of information, as businesses and individuals use businesses and individuals around the world in ways that benefit CCPs. There is.
Peter Mattis, chairman of the Jamestown Foundation, a think tank, recently heard and testified to the Senate Foreign Relations Committee on Deepshake’s concerns.
He thinks chatbots are no different from Tiktok. This eats the ever-growing number of user data as more people interact with it. These “automated disinformation systems” are “the more difficult it is to recognize them, the more difficult it is to stop,” Mattis said in his testimony.
I watched a video recorded on social media, including Tiktok, on January 14, 2025, at Times Square in New York. Adam Gray/Getty Images
Many world leaders and governments have already expressed concern with Deepseek about their security flaws and censorship practices. Some have started the process of blocking the use of the app completely.
Tiktok is no longer available on US app stores amid the US ban. At the Senate hearing, Matisse said the US should take the same approach on DeepSeek and other similar platforms.
For the average American, Castro suggests that you pay special attention when downloading apps.
“They should look at which companies are making it, where they are based, and what values are tied to that company,” he said.
He hopes that the app store can play a role in warning users about such risks they need to know.
