Amid Tightening Controls and “Rumor” Crackdown, People’s Daily Touts AI-assisted Censorship Guidance

China’s Ministry of Public Security has initiated a new campaign against online rumors following Xi Jinping’s call for a “security barrier” around the Chinese internet. At The South China Morning Post, Kawala Xie reported on the crackdown

This year the public security ministry already ran a 100-day crackdown targeting online rumors and said it had achieved “remarkable results”.

The ministry said that during the crackdown, which started in April, its officers investigated more than 1,600 cases of online rumors and shut down more than 10,000 social media accounts for spreading rumors.

[…] Local governments were quick to tout their success in the crackdown.

In a post on the social media site Weibo, Shanghai police said 258 people in its jurisdiction had been targeted for spreading rumors online, and 460 illegal accounts shut down for violations. [Source]

In press releases, police stress that the “rumor mongers” targeted in the crackdown are those spreading malicious rumors about others, including a case in which someone used AI to generate rumors that a kindergarten teacher was moonlighting as a prostitute. At times, though, the crackdown can border on farce: internet police in Hebei trumpeted the detention and “strict education-through-criticism” of a young man who posted a video of snowfall when it was not, in fact, snowing. “Rumor mongering” has also been used to silence criticism of the state—particularly infamously, during the Wuhan and Shanghai lockdowns.

In the state’s telling, past online clean-up operations have been massively popular. A survey purporting to have sampled 360,000 netizens from 96% of China’s prefecture-level cities, released before a “internet civilization” conference in Xiamen, claimed that 82.24 percent of netizens are satisfied with “net integrity.” At the same forum it was announced that in 2022 censors took down 54.3 million “illegal and harmful” posts, “dealt with” 6.8 million accounts, took down 2,890 applications and mini-programs, disbanded 260,000 groups and forums, and shut down 7,300 websites. 

As part of the broader push to tighten internet controls, cyberpolice in Sichuan have been tasked with monitoring the internet for videos that “incite conflict” between ethnicities. A document published in advance of the Chengdu University World Games, an international collegiate athletic competition, translated by Jeremy Daum of China Law Translate, set out provisions on a monthlong campaign to wipe the internet of discourse on local politics, food quality issues, and ethnic discrimination

1) Disseminating false information on regional public policies and the areas of society and people’s livelihood fields related to the University World Games, fabricating rumors about disasters, crimes, food product quality issues, and so forth, that might cause panic, especially where rumors continue to be spread after the release of authoritative rumor refuting them.

(2) Biased and distorted dissemination of information on emergencies, making statements by posing as a party or related person, making connections to stir up old news and events; maliciously hyping negative topics involving college students, etc.

(3) Publishing texts, pictures, or audio-video related to ethnic or regional discrimination and inciting conflict between groups; staging scenes of local “impoverished” life and maliciously peddling pity in the name of documenting life or giving aid. [Source]

While Chinese censors often turn a blind eye to anti-Black racism on social media sites—trolls on Douyin have taken to mimicking American racism to attack Black China-based creators—content that touches on internal Chinese ethnic strife is often swiftly censored. In 2022, censors repeatedly shut down attempts to draw attention to Uyghur suffering under lockdown, calling the censorship effort a “smokeless war.”

Censorship has not only struck at ethnic issues and rumor mongers. Financial news, once considered a relatively safe topic for public discussion, has been repeatedly censored amid a potential recession. Youth unemployment figures have been sporadically censored, likely for similar reasons. Other recent targets of censorship include cremation statistics (which hint at COVID’s true death toll), reports on the pocket crime “picking quarrels and provoking trouble,” health and science blogs, and even Alibaba listings of toys that were a tongue-in-cheek reference to a recent instance of official mendacity. 

What to censor and when poses a major problem to companies, organizations and individuals—making self-censorship is a highly profitable business. All published information must be screened for potential political fault lines, necessitating a knowledge of red lines many are ignorant of. This is the “Li Jiaqi Paradox,” so named after a famous livestreamer who hawked an ice cream cake that resembled a tank on June 3rd, the eve of the anniversary of the Tiananmen Massacre, seemingly unwittingly. Li mysteriously stopped live-streaming for over a year after the incident. (Young internet censors reportedly take a crash course in sensitive history to ensure they are aware of the events the Party has attempted to banish from collective memory.) Artificial intelligence is now being applied to potentially solve the paradox. At The South China Morning Post, Vanessa Cai and Sylvie Zhuang reported on People’s Daily’s new artificial intelligence censorship product “Renmin Shejiao.” The product’s selling point is the Party’s flagship paper’s ostensibly unparalleled knowledge of what must not be published:

Clients upload material to the platform for review by AI and a team of censors. Content that could be flagged as risky includes material related to ideology, religion, purged government officials, Chinese dissidents, and maps related to disputed border areas.

While Renmin Shenjiao’s algorithm generally might not be as strong as its tech company rivals, the idea is that it will be better at filtering political content, according to a person with knowledge of the product.

This was because Renmin Shenjiao staff were part of the People’s Daily group, and may have knowledge of new taboos that were otherwise not known to the public, the person said.

[…] Neil Thomas, a fellow in Chinese politics at the Asia Society Policy Institute’s Centre for China Analysis said, “Companies that want to maximize profit will rationally acquire content moderation services that minimize the risk of overstepping political boundaries or being out of sync with Xi Jinping Thought.

“I think this shows that China is the global center of censorship and China is at the forefront of innovation when it comes to new ways to strengthen the censorship process.” [Source]


Subscribe to CDT


Browsers Unbounded by Lantern

Now, you can combat internet censorship in a new way: by toggling the switch below while browsing China Digital Times, you can provide a secure "bridge" for people who want to freely access information. This open-source project is powered by Lantern, know more about this project.

Google Ads 1

Giving Assistant

Google Ads 2

Anti-censorship Tools

Life Without Walls

Click on the image to download Firefly for circumvention

Open popup

Welcome back!

CDT is a non-profit media site, and we need your support. Your contribution will help us provide more translations, breaking news, and other content you love.