Thirty years ago today, the Chinese Communist Party used military force to suppress a peaceful pro-democracy demonstration by thousands of university students. Hundreds (some estimates go as high as thousands) of innocent protesters were killed. Every year, people around the world come together to mourn and commemorate the fallen; within China, however, things are oddly silent.

The Tiananmen Square protest is one of the most tightly censored topics in China. The Chinese government’s network and social media censorship is more than just pervasive; it’s sloppy, overbroad, inaccurate, and always errs on the side of more takedowns. Every year, the Chinese government ramps up VPN shutdowns, activist arrests, digital surveillance, and social media censorship in anticipation of the anniversary of the Tiananmen Square protests. This year is no different; and to mark the thirtieth anniversary, the controls have never been tighter.

Keyword filtering on social media and messaging platforms

It’s a fact of life for many Chinese that social media and messaging platforms perform silent content takedowns via regular keyword filtering and more recently, image matching. In June 2013, Citizen Lab documented a list of words censored from social media related to the anniversary of the protests, which included words like “today” and “tomorrow.”

Since then, researchers at the University of Hong Kong have developed real-time censorship monitoring and transparency projects—“WeiboScope” and “WechatScope”—to document the scope and history of censorship on Weibo and Wechat. A couple of months ago, Dr. Fu King-wa, who works on these transparency projects, released an archive of over 1200 censored Weibo image posts relating to the Tiananmen anniversary since 2012. Net Alert has released a similar archive of historically censored images.

Simultaneous service disruptions for “system maintenance” across social media platforms

This year, there has been a sweep of simultaneous social media shutdowns a week prior to the anniversary, calling back to similar “Internet maintenance” shutdowns that happened during the twentieth anniversary of the Tiananmen Square protests. Five popular video and livestreaming platforms are suspending all comments until June 6th, citing the need for “system upgrades and maintenance.” Douban, a Chinese social networking service, is locking some of their larger news groups from any discussion June 29th, also for “system maintenance.” And popular messaging service WeChat recently blocked users from changing their status messages, profile pictures, and nicknames for the same reason.

Apple censors music and applications alike

Since 2017, Apple has removed VPNs from its mainland Chinese app store. These application bans have continued and worsened over time. A censorship transparency project by GreatFire, AppleCensorship.com, allows users to look up which applications are available in the US but not in China. Apart from VPNs, the Chinese Apple app store has also censored applications from news organizations, including the New York Times, Radio Free Asia, Tibetan News, Voice of Tibet, and other Chinese-language human rights publications. They have also taken down other censorship circumvention tools like Tor and Psiphon.

Leading up to this year’s 30-year Tiananmen anniversary, Apple Music has been removing songs from its Chinese streaming service. A 1990 song by Hong Kong’s Jacky Cheung that references Tiananmen Square was removed, as were songs by pro-democracy activists from Hong Kong’s Umbrella Movement protest.

Activist accounts caught in Twitter sweep

On May 31st, a slew of China-related Twitter accounts were suspended, including prominent activists, human rights lawyers, journalists, and other dissidents. Activists feared this action was in preparation for further June 4th related censorship. Since then, some of the more prominent accounts have been restored, but many remain suspended. An announcement from Twitter claimed that these accounts weren’t reported by Chinese authorities, but were just caught up in a large anti-spam sweep.

The lack of transparency, poor timing, and huge number of false positives on Twitter’s part has led to real fear and uncertainty in Chinese-language activism circles.

Beyond Tiananmen Square: Chinese Censorship and Surveillance in 2019

Xinjiang, China’s ground zero for pervasive surveillance and social control

Thanks to work by Human Rights Watch, security researchers, and many brave investigators and journalists, a lot has come to light about China’s terrifying acceleration of social and digital controls in Xinjiang in the past two years. And the chilling effect is real—as we approach the end of Ramadan, a holiday which is discouraged and banned for Party members and public school students to observe—mosques remain empty. Uighur students and other expatriates abroad fear returning home, as many of their families have already been detained for no cause.

China’s extensive reliance on surveillance technology in Xinjiang is a human rights nightmare, and according to the New York Times, “the first known example of a government intentionally using artificial intelligence for racial profiling.” Researchers have noticed that more and more computer vision papers coming out of China are specifically trained to build facial recognition for Uighurs.

China has long been a master of security theater, overstating and over-performing its own surveillance capabilities in order to spread a “chilling effect” over digital and social behavior. Something similar is happening here, albeit at a much larger scale than we’ve ever seen before. Despite the government’s claims of fully automated and efficient systems, even the best automated facial recognition systems they use are only accurate in less than 20 percent of cases, leading to mistakes and the need for hundreds of workers to monitor cameras and confirm the results. These smoke-and-mirrors “pseudo-AI” systems are more than common in the AI startup industry. For a lot of “automated” technologies, we just aren’t quite there yet.

Resource or technical limitations aren’t going to stop the Chinese government. Security spending since 2017 shows that Chinese officials are serious about building a panopticon, no matter the cost. The development of the surveillance apparatus in Xinjiang shows us just how expensive building pervasive surveillance can be; local governments in Xinjiang have accrued hundreds of millions (in USD) of “invisible debt” as they continue to ramp up investment in their surveillance state. A large portion of that cost is labor. “We risk understating the extent to which this high-tech police state continues to require a lot of manpower,” says Adrien Zenz for the New York Times.

Client-side blocking of labor movements on Github

996 is a recent labor movement in China by white-collar tech workers who demand regular 40-hour work weeks and the explicit outlaw of the draconian but standard “996” schedule; that is, 9 am to 9 pm, six days a week. The movement, like other labor-organizing movements, has been to subject to keyword censorship on social media platforms, but individuals have been able to continue organizing on Github.

Github itself has remained relatively immune to Chinese censorship efforts. Thanks to widespread deployment of HTTPS, Chinese network operators must either block the entire website or nothing at all. Github was briefly blocked in 2013, but the backlash from developers was too great, and the site was unblocked shortly thereafter. China’s tech sector, like the rest of the world, rely on open-source projects hosted on the website. But although Github is no longer censored at the network level, Chinese-built browsers and Wechat’s web viewer started blacklisting specific URLs from being accessed, including the 996 Github repository.

Google’s sleeping Dragonfly

Late last year, we stood in solidarity with over 70 human rights groups led by Human Rights Watch and Amnesty International, calling on Google to end their secret internal project to architect a censored Chinese search engine codenamed Dragonfly. Google employees wrote their own letter protesting the project, some resigning in protest, demanding transparency at the very least.

In March, some Google employees found that changes were still being committed to the Dragonfly codebase. Google has yet to publicly commit to ending the project, leading many to believe the project could just be on the back burner for now.

How are people fighting back?

Relatively little news gets out of Xinjiang to the rest of the world, and China wants to keep it that way— journalists are denied visas, their relatives are detained, and journalists on the ground are arrested. Any work by groups that help shed light on the situation is extremely valuable. Earlier this year, we wrote about the amazing work by Humans Rights Watch, Amnesty International, other human rights groups, and other independent researchers and journalists in helping uncover the inner workings of China’s surveillance state.

Censorship transparency projects like WechatScope, WeiboScope, Tor’s OONI, and GreatFire's AppleCensorship, as well as ongoing censorship research by academic centers like The Citizen Lab and organizations like GreatFire continue to shed light on the methods and intentions of broader Chinese censorship efforts.

And of course, we have to take a look at the individuals and activists within and outside China who continue to fight to have their voices heard. Despite the continued rise of crackdowns on VPNs, VPN usage across Chinese web users continues to rise. In the first quarter of 2019, 35% of web users use VPNs, not just for accessing better music and TV shows, but also commonly for accessing blocked social networks, and blocked news sites.

Human rights groups, security researchers, investigators, journalists, and activists on the ground continue to make tremendous sacrifices in fighting for a more free China.