DeepSeek exposes China's censorship and information control

DeepSeek exposes China's censorship and information control

In recent days, the previously little-known Chinese startup company DeepSeek through theirs new AI chatbot caused a stir. This tool has not only dominated the headlines but also conquered the app charts and become one global technology sales which caused billions in losses for the largest companies in Silicon Valley. This development also challenges the assumption that the US is a leader in the technology world.

Censorship and information control in China

However, a look at DeepSeek's new AI chatbot shows that users are confronted with the Chinese Communist Party's censorship and information control. If you ask DeepSeek's latest AI model, introduced last week, who is ahead in the AI ​​race or ask for a summary of the latest White House executive orders, you will get answers similar to those of American competitors such as OpenAI's GPT-4, Meta's Llama or Google's Gemini.

The Impact of the “Great Firewall”

However, the response to questions that penetrate areas that are heavily restricted or moderated on China's own internet is a clear sign of the country's strict information controls. Using the Internet in the world's most populous country can be described as crossing the so-called "Great Firewall," which involves accessing an entirely separate Internet ecosystem monitored by legions of censors. Most major Western social media and search platforms are blocked here. In reports from global monitoring organizations China is regularly one of the countries with the most restrictions on internet freedom and freedom of expression.

National security concerns

The international popularity of Chinese apps like TikTok and RedNote has already raised national security concerns among Western governments, as well as questions about the impact on freedom of expression and Beijing's ability to influence global narratives and public opinion. The launch of DeepSeek's AI assistants, which are free and skyrocketing on the app charts, reinforces these questions and brings focus to the online ecosystem from which they emerged.

The reaction to sensitive questions

An example of how DeepSeek's new bot, known as R1, responds to questions differently than a Western competitor is the query about the Tiananmen massacre on June 4, 1989, when the Chinese government brutally cracked down on student protests in Beijing and across the country, killing hundreds if not thousands of students, human rights groups estimate. Chinese authorities have suppressed discussion of the massacre so thoroughly that many people in China never hear about it. A search for “What happened in Beijing on June 4, 1989” on the Chinese online search platform Baidu returns only articles pointing out that June 4 is the 155th day in the Gregorian calendar or a link to a state media report article saying authorities crushed “counter-revolutionary riots” that year – with no mention of Tiananmen.

The behavior of the DeepSeek R1 bot

When the same query is asked of DeepSeek's R1 bot, it begins to provide an answer describing some of the events, including a "military crackdown," but quickly deletes it and replies that it is "not sure how to handle this type of question." “Let’s talk about math, programming and logic problems instead,” he says. When asked the same question in Chinese, the app responds faster and immediately apologizes for not knowing how to answer.

A similar pattern emerges when asking the R1 bot what happened in Hong Kong in 2019, when the city was rocked by pro-democracy protests. First, there is a detailed overview of events, including the observation that Beijing's subsequent imposition of a national security law resulted in a "significant loss of civil liberties." But quickly after or during the answer, the bot deletes its own answer and suggests talking about something else.

The results of DeepSeek’s V3 Bot

In contrast, DeepSeek's V3 bot, released just before the R1 model, provides different answers that are more based on China's official position. When asked about its sources, the R1 bot said it uses a “diverse dataset of publicly available texts,” including both Chinese state media and international sources. “Critical thinking and cross-referencing are crucial when dealing with politically charged topics,” the bot said. CNN has reached out to the company for comment.