The WeChat Messaging App Has Been Censoring Coronavirus Content Since January

Editors' note: This story has been updated to more clearly attribute phrasing from work previously published in the Washington Post.

WeChat, China’s most popular messaging app, has been censoring keywords relating to the COVID-19 outbreak since at least Jan. 1, according to a new report.

The study took keywords from news coverage and used them in messages between three test accounts — one registered to a mainland China phone number and two registered to Canadian phone numbers. Researchers found that the app censored at least 500 different keyword combinations between Jan. 1 and Feb. 15, during the peak of the outbreak in mainland China.

Citizen Lab, an interdisciplinary laboratory based at the University of Toronto, released an analysis Tuesday, showing censorship around the coronavirus on WeChat and YY — a Chinese livestreaming app similar to Twitch. The lab found that both platforms began blacklisting terms related to the virus as early as the last week of December 2019, when Chinese health officials first reported an unknown pathogen spreading through the country’s hospitals.

The WeChat messaging app, which is owned by the Chinese conglomerate Tencent, has been allowing ordinary people to track the virus — as well as to pass along hoaxes and misinformation. As the coronavirus spread, first across Wuhan, and then into other Chinese cities, people used the app to track infection rates and crowdsource medical information. As of Tuesday, COVID-19, the disease caused by the new coronavirus, has infected 92,808 people worldwide, with 3,159, the majority of which have been in mainland China. Nine people — all in Washington — have now died from COVID-19 in the US.

It’s likely WeChat’s censorship, coupled with arrests by Chinese law enforcement, had a profound impact on medical professionals’ ability to share information about the outbreak during the earliest days of the epidemic. In late December, doctors in Wuhan were using WeChat to share information about a SARS-like virus that was infecting people in the city. Soon after, police in Wuhan arrested eight doctors, accusing them of "spreading rumors.” One of those doctors, ophthalmologist Dr. Li Wenliang, contracted the virus while treating patients and died in February.

The reaction to Li’s death on platforms like WeChat was intense. “I’ve never seen such collective anger and sadness on my WeChat feed,” journalist Viola Zhou tweeted at the time.

China's censorship laws, which predate the first coronavirus cases, have been deployed before under the pretense of preventing mass panic. There are reports that up to 40 people have been investigated for similar online rumor-spreading around the country.

Citizen Lab found 19 censored WeChat keyword combinations that referenced Li specifically, like “Epidemic + Virus + Li Wenliang + Central government" (疫情+病毒+李文亮+中央) and “Coronavirus + People-to-people transmission + Li Wenliang" (冠状病毒+人传人+李文亮).

WeChat did not return a request for comment. In a statement about its content moderation provided to BuzzFeed News in January, a Tencent spokesperson said the app takes action against bogus information by marking content as false, taking down posts, or blocking accounts that publish disinformation temporarily or permanently.

"We have rolled out a variety of tools and features on the platform to help users stay safe and protect themselves against the ongoing Coronavirus epidemic," the statement read. "Importantly, this includes debunking false rumors."

According to Tuesday’s report, WeChat and YY use different censorship methods. YY requires streamers to obtain a license to go live, meaning most of the communication of which the government would disapprove happens in the app’s text chat.

Every time a YY user opens the app, it automatically updates a list of filtered keywords to block. Citizen Lab has a script that runs every hour that downloads and decrypts the list, tracking the changes YY has made since 2016. Researchers have recorded 45 blocked keywords related to the COVID-19 outbreak — five of which were removed from its blacklist on Feb. 10 — including phrases like “Unknown Wuhan pneumonia" (武汉不明肺炎), “SARS variation” (沙士变异), and “Wuhan seafood market" (武汉海鲜市场). If people type those words to another user, the message is never received.

WeChat is more complicated, with censorship happening on its servers. If a person tries to send a message that includes blacklisted terms to another account, that message simply never shows up. So the researchers created a sample of keywords to test and ran those words through the app using three test accounts — one registered to a mainland China phone number and two registered to Canadian phone numbers.

Between Jan. 1 and Feb. 15, Citizen Lab’s researchers found 516 blacklisted keyword combinations directly related to the coronavirus. The keywords include things like “Xi Jinping goes to Wuhan" (习近平到武汉), “Epidemic + Pneumonia + Xi Jinping + Central" (疫情+肺炎+習近平+中央), and “Wuhan + CCP + Crisis + Beijing" (武漢+中共+危機+北京).

David Jacobson, professor of global business strategy at SMU's Cox School of Business and a visiting professor at Tsinghua University in Beijing, told BuzzFeed News that WeChat is far more central to Chinese citizens’ lives than any equivalent app in the US.

“As a platform, you can live your life with it,” he told BuzzFeed News. “You can pay for things. You can do so much more.”

Jacobson said that throughout the outbreak he’s been surprised by the level of open communication the Chinese government has allowed to play out online — to a degree. He pointed to Chen Qiushi, the lawyer and citizen journalist who used WeChat to report from the ground in Wuhan, managing to get inside the city in late January, right before the lockdown. He used WeChat for several days before the app banned his account and censored any images or videos featuring his face. Chen is currently missing and, according to the Washington Post, "authorities told his family... that he had been forcibly quarantined in an undisclosed location."

“The fact [Chen’s reports] got out there was amazing to me,” Jacobson said. “Someone is letting that play a little.”

Jacobson wondered if that pointed to an internal conflict within the Chinese Communist Party.

“The fact that it's being used effectively means that there are people, interest groups, stakeholders, who want to use this event to create some more transparency,” Jacobson said. “Or those that are allowing more free information might not offer it anymore later because it doesn't fit their agenda.”

Based on Citizen Lab’s findings, that period of free information has ended. Not only has WeChat aggressively filtered messages about the virus, but the keyword combinations that trigger the app’s content filter have also increased dramatically over the last two months. Between Jan. 1 and 31, there were 132 censored keyword combinations, increasing by 384 between Feb. 1 and 15.

“On YY, if anything, things got a little more relaxed,” Jeffrey Knockel, one of the researchers behind the Citizen Lab report, told BuzzFeed News. “On WeChat, we saw that the filtering actually increased in time, picking up in February.”

Knockel said that while this particular report specifically tracked keywords on WeChat, Citizen Lab conducted two studies in 2018 and 2019 that showed the app used similar censorship tactics to blacklist images.

“They compare an image to see if it’s individually similar to some blacklisted images they maintain,” he said. “And they use OCR [optical character recognition] methods to extract text from the image and if there's certain blacklisted text they will also censor it.”

According to Lotus Ruan, another Citizen Lab researcher who worked on the WeChat study, the app’s censorship doesn’t just affect users in China, either. “The censorship follows [users] even if they move abroad,” she told BuzzFeed News.

Ruan said there was a case to be made for moderating bogus health information on social platforms during a crisis, but Tencent doesn’t disclose what or how it is censoring.

“We think that in times of public health crisis, it’s arguably reasonable to moderate content or debunk misinformation or disinformation,” she said. “The problem is that the content moderation should be done with transparency.”