The Chinese State Cyberspace Administration has recently ordered major Internet platforms to further strengthen monitoring of content in chat applications, requiring comprehensive integration of keyword searches and secondary data analysis systems. According to several interviewed netizens and industry insiders, authorities are extensively deploying artificial intelligence technology on backend servers of mainstream social platforms such as WeChat, QQ, and TikTok to continuously scan and deeply identify users’ chat records, voice messages, images, and other information. One netizen was questioned by the police for several hours simply for posting the phrase “one statue”.
The control over online speech on Chinese cyberspace platforms has tightened once again. Beijing netizen Mr. Chen revealed to a reporter on July 23rd that he was summoned twice by the Haidian Public Security Bureau for posting “one statue” in a WeChat group. He recalled, “The cyber police at the police station directly asked me, ‘Who are you referring to? Why mention him (Xi)? I didn’t even mention Old Xi, but I was still questioned for five hours. They even said, ‘As long as you mention him, we have to take action,’ I felt they were helpless as well.”
Mr. Chen also disclosed that the police inadvertently revealed during the process that Beijing already has tens of thousands of dedicated cyber police for network monitoring and data analysis, with work intensity “even exceeding criminal and police officers.” “These people look at screens from morning till night, analyzing tens of thousands of chat logs and account activities. If Beijing is like this, other cities should be similar to Beijing.”
Internet security expert Mr. Zhou (pseudonym) stated that platforms like WeChat have been integrated into the Cyberspace Administration’s unified deployment of the “intelligent public opinion monitoring system,” which not only monitors public posts in real-time but also rescans stored content like voice, text, and images. He said, “Previously, chat records were just backups, but now they undergo repeated analysis. Even records from several years ago are flagged if they trigger sensitive words.”
Overseas-based Mr. Zhou further explained that the Cyberspace Administration has been promoting multiple new AI inspection systems since this year, capable of identifying homophones, pinyin, emoticons, and text within images, with backend systems operating around the clock. Every suspicious message generates a report and is reviewed by specialists. He said, “Now even likes, reposts, browsing frequency, and other behaviors are recorded and quantified. The backend will track what you said in the groups, who liked your posts, what content you reposted, all stored in the database.”
A netizen from Henan, Mr. Lv, stated that he has been suspended from multiple platforms more than ten times: “Sometimes my account gets suspended before I even say anything. Friends leave comments that I can see, but others can’t. This indicates that ‘invisible censorship’ is in place, preventing your posts from reaching others.”
This mechanism of “visible to oneself but invisible to others” has become a common occurrence on social platforms like Weibo, TikTok, and Xiaohongshu. While accounts appear normal on the surface, they are actually subjected to “limited reach” or “buried,” making it nearly impossible to disseminate information.
IT industry insider Mr. Gao (pseudonym) pointed out that this round of inspections far exceeds previous ones, involving several major platforms including Tencent, ByteDance, Kuaishou, and Alibaba. He said, “Whenever there is the slightest movement, the Cyberspace Administration immediately issues directives demanding post deletion, account suspensions, and submission of user information. Some platforms even have dedicated teams cooperating with the government round the clock.”
Mr. Gao warned, “Now, it’s not about what you say at the moment but what you have said in the past; platforms actively backtrack and analyze. Even if it’s just a repost or a like, algorithms will tag it, forming a behavioral profile.”
Observations by reporters revealed a significant decrease in political statements on Chinese social platforms in the past three months. Methods previously used to evade scrutiny, such as homophonic words, pinyin abbreviations, and emoji substitution, are now identified and blocked. Users are now more inclined to share overseas screenshots, content from self-media pages, or simply remain silent to avoid risks.
A former employee of a social platform, who preferred not to disclose their name, told reporters, “We receive a daily list of keywords and topics to be handled from the Cyberspace Administration. The platform doesn’t care if what you say is reasonable or not; as long as it touches on sensitive words, your account is automatically suspended or restricted.”
According to the 55th China Internet Network Development Status Statistical Report released by the China Internet Network Information Center (CNNIC) by December 2024, the total number of Chinese internet users reached 1.108 billion, with an internet penetration rate of 78.6%, of which mobile internet users accounted for 98.6%. This means that almost every netizen’s social activities are constantly under technical scrutiny.
Mr. Gao finally cautioned the public, “It’s not as simple as just having one account suspended now. If your speech is classified as ‘organized’ or ‘widely disseminated,’ you may not only face account suspension but also summons, administrative detention, and even criminal prosecution.”