A game of cat and mouse We asked an expert how China’s Internet censorship really works

In March 2021, the Russian authorities launched another attack on Western Internet services. Following its attempt to throttle Twitter, the federal censor, Roskomnazor, threatened to block the network in Russia entirely unless it removes certain “illegal content” (which includes, as it turns out, the accounts of independent media outlets). There’s a number of reasons why a “Great Russian Firewall” based on the Chinese system is impossible in Russia. But many Russian officials, and pro-Kremlin commentators and media outlets, have long expressed their approval of the “Chinese model” of Internet regulation and called for the introduction of at least some of its elements in Russia. To find out more about how Internet censorship really works in modern China, Meduza spoke to Leonid Kovachich, a specialist on China and digital technologies. 

Leonid Kovachich

China expert and specialist on digital technologies

On the Chinese Internet, there are topics that have been banned since the beginning. But there aren’t that many of them: a few examples include anything to do with separatism, issues related to the independence of Tibet and Xinjiang, the 1989 Tiananmen Square protests, anything related to the Falun Gong spiritual movement, and, of course, any criticism of the Chinese Communist Party (CCP) and the central authorities. On the other hand, pointing out problems with local officials isn’t prohibited.

Moreover, it’s not so important for the Chinese authorities to strictly limit access to foreign media (if someone wants to read The New York Times and Facebook or Twitter, they can). Instead, their policy is aimed at making the bulk of the population exclusively consume content from local outlets under the Communist Party’s control. Therefore, they’ve created domestic equivalents for all of the existing Internet services, including Twitter (Weibo), Facebook (Wechat), YouTube (Youku), and even the question-and-answer service Quora (Zhihu) and the academic search engine Google Scholar (Baidu Wenku).

All of these platforms are under the strict control of the CCP. Since the mid-2010s, all bloggers have been required to register with the state using their real names and identity documents. Moreover, in 2013, the authorities introduced liabilities for publishing unreliable or false information under the so-called rumor and speculation law. Of course, any criticism of the central authorities falls into this category. Under this law, it’s considered a felony if a problematic post is republished more than 500 times or receives more than 5,000 views. And given the scale of the Chinese Internet, 500 shares in nothing. In other words, almost any post on Chinese social networks containing information that the authorities deem undesirable falls under this law.

This is exactly what happened to the famous Chinese doctor Li Wenliang, who tried to warn his colleagues about a new infection (the novel coronavirus) spreading throughout the country. He ran into problems with the law on rumors and speculation. He was summoned by the relevant authorities and forced to offer them an explanation, and then he had to admit publicly that he was guilty of publishing unreliable information. The information, of course, turned out to be reliable, and Li Wenligan himself died of the very coronavirus infection that he tried to warn everyone about. 

At the beginning of 2021, the Chinese authorities issued new rules according to which Internet and social media users, including bloggers, do not have the right to independently publish information about politics, the military sphere, the economy, and the social sphere (among other things), without appropriate media accreditation. This is another step towards tightening censorship — in essence, with the exception of pictures of cats and food, there are ever fewer topics in the Chinese blogosphere that can be discussed with relative freedom. 

Even topics like food and cats have certain restrictions. For example, pictures of Winnie the Pooh were banned after Internet users started comparing Chinese President Xi Jinping to this cartoon character. Posting photos of Chinese steamed buns has also been prohibited for the same reason. These cat and mouse games have been going on for many years: the Chinese authorities are constantly expanding the list of prohibited words used to block content, and Internet users continue to try and circumvent the restrictions. Initially, they used codewords, like “steamed bun.” Then, to bypass the censorship machine’s text filters, they simply started posting photographs. In response, the administrators began implementing AI systems with image recognition to track such content. As a result, algorithms block certain images automatically now, such as pictures of Winnie the Pooh or an empty chair, which was used to refer to Chinese dissident Liu Xiaobo. 

Today, live streaming services are gaining popularity in China precisely because they offer live communication and content moderation is difficult. However, AI technologies have made it possible not only to identify live speech, but also to extract keywords. If the algorithm flags a stream as suspicious, then a moderator receives a notification and manually tunes into the stream. If the discussion involves some prohibited or politically unsavory topics, the broadcast is forcibly stopped and the user who started the live stream is blocked from accessing the platform.

In other words, technology is evolving constantly and uncensored conversation is becoming more and more difficult. When Clubhouse first appeared, it was wildly popular in China for the first week, precisely because it seemed like a new form of free communication. But the Chinese authorities quickly banned Clubhouse. 

True, the extent to which all of these technical innovations in the field of censorship are homegrown Chinese products is debatable. Any progress is based on open source systems. And naturally, Chinese developers are refining some applications to suit their own needs. For example, Chinese programmers actively use Western platforms like Tensorflow and Pytorch for machine learning because China doesn’t have its own popular platforms (Baidu’s machine learning platform failed to take off, even on the domestic market).

Therefore, Chinese developments in the artificial intelligence and machine learning fields, which are also used in censorship algorithms, rely on the fruits of Western scientific and technological progress. And China is talking about this constantly — a significant portion of the country’s recently adopted five-year development plan is devoted to the need to develop their own key innovations. This is a critical point: all of these technological developments are actually the superstructure — Western technical innovations are being adapted to Chinese realities.