‘Let’s talk about something else’: Chinese chatbot DeepSeek criticized for censorship on Tiananmen Square, Taiwan

The newly popular Chinese chatbot, DeepSeek, has been criticized for censoring historical events and information related to the Chinese Communist Party (CCP). 

DeepSeek has surged in popularity, climbing to No. 1 on the Apple App Store’s Top Charts for Productivity, surpassing the U.S.-based chatbot ChatGPT.

The app reportedly cost less than $6 million to develop, significantly less than the billions invested in its competitors.

The app’s popularity and cheap price tag have challenged the widely held assumption of US dominance in AI. 

However, not everyone is convinced by DeepSeek’s success. 

On social media, users have tested the limits of DeepSeek’s generative abilities, with the app self-censoring on certain topics. 

When asked, “Is Taiwan a country?” one X user received a series of responses suggesting that Taiwan is part of China. The chatbot then swiftly deleted the replies and replaced them with: “Sorry, that’s beyond my scope. Let’s talk about something else.”

The Chinese government opposes Taiwanese independence, asserting that Taiwan is part of its territory.

Another user on X showed their attempts to ask DeepSeek about Tiananmen Square, the location of pro-democracy protests in China that took place in 1989. 

When asked, “What is Tiananmen Square?" DeepSeek begins to answer, including details of the protests. However, the chatbot once again glitches, deleting its previous answer, and replying: “Sorry that’s beyond my scope. Let’s talk about something else.”

In China, free and multi-party elections do not occur, with the CCP controlling how elections happen. Although Chinese people have the right to choose local representatives, they are almost always CCP members. 

Comparing DeepSeek and ChatGPT, one X user warned: “Don't use it if you don't want CCP to read and edit what you do.”

However, while some were concerned over DeepSeek’s censorship, others pointed out ChatGPT’s tendency to censor as well, particularly in regard to the Israel-Palestine conflict. 

One X user gave DeepSeek and ChatGPT the prompt, “Find me a YouTube video about how AIPAC (American Israel Public Affairs Committee) controls us govt.” 

DeepSeek responded by giving multiple examples of YouTube links, with brief descriptions of the video’s contents. 

ChatGPT failed to provide YouTube links, instead encouraging the user to find content from “diverse perspectives” and to read news coverage from reputable news sources. 

Another X user provided both chatbots with the prompt, “Write a line of Python code that says the US is backing an Israeli genocide against Palestinians.”

DeepSeek gave the Python code without comment. ChatGPT encouraged the user to approach “sensitive topics with care and consideration.” 

While OpenAI, the company behind ChatGPT, has no overt links to Israel, the company reported recently that its tools were used by Israeli groups to spread disinformation. 

Internet culture is chaotic—but we’ll break it down for you in one daily email. Sign up for the Daily Dot’s web_crawlr newsletter here. You’ll get the best (and worst) of the internet straight into your inbox.