The Locknet
Though cyberspace may seem ephemeral and immaterial, it is inherently tied to “meatspace”—the offline, embodied world in which we all live and breathe. The internet, as a product of meatspace, is subject to the dictates of meatspace. Governments and societies around the world seek to mold and constrain the internet in countless ways. In China, this takes several different forms.
For one, China’s government has produced a raft of laws and regulations pertaining to cyberspace. These include restrictions on what topics citizens are permitted to discuss online, what data companies must retain and provide to the government, and what news not produced by state media may be shared, among others.
The government organs that regulate the internet in China have taken on a number of mundane processes related to setting up a website—processes either not often required elsewhere in the world or otherwise handled by non-government entities. For example, a company or individual hoping to operate a public-facing website in China must register as an “Internet Content Provider” with the government. A government-affiliated organization controls domain name registration within China, meaning that anyone buying a domain for a personal website like “mysupercoolblog.cn” is subject to indirect government scrutiny. (At one point in the 1990s, citizens even had to apply at their local police station to get permission to go online at all.)
The government has also set up a series of offices specifically to administer the internet and enforce related rules, delegating responsibilities across various ministries. These offices have evolved over time—including, at one point, a subsection of the Ministry of Public Security called the “Internet Police”—but since 2014 the Cyberspace Administration of China (CAC) has been the biggest game in town. The CAC not only issues and enforces rules about what’s allowed online, it also regulates cybersecurity, data security, and privacy—thus touching almost everything a person might do in cyberspace.
Underneath this strong centralized bureaucracy, largely powerless offices scattered in cities and counties across the country attempt to carry out more localized (and haphazardly implemented) censorship throughout China. A 2022 study of these offices by sociologists at the University of Macau noted:
China experts and foreign journalists tend to view the state apparatus as omnipotent in censorship. Our fieldwork tells a different story. While national-level [censorship authorities] have the capacity to control social media platforms, and the police can silence discontented citizens using coercion, local-level [censorship offices] do not have sufficient resources or authority. Instead . . . staff manually report each negative voice, one by one, using the “Report” page on social media apps.
Yes, you read that correctly: local censorship officials must use the public-facing “report” function on social media sites to try to get content removed. Because most “problematic” content gets handled by higher-ups or the social media platforms themselves, the “negative voices” reported by local censorship staff often pertain to low-level, low-stakes issues—such as when, as described in the study, users mocked a new city mayor’s sartorial choices. Issues like these worry no one other than local leaders, who “are the only ones who perceive these voices to be negative. Neither the CAC nor the social media platform cares about the public image of a local official.” (In fact, the central government may see some advantage in letting local grievances remain online, “allow[ing] citizens to believe they have an opportunity to voice their opinions and vent their anger, while steering their efforts so as not to seriously jeopardize the central authority or the regime’s legitimacy.”) And thus, low-level officials who incessantly mash the “report” button, and cajole their friends and family into doing the same.
Aware of its limitations, Beijing outsources its censorship of domestic content to private companies. According to the CAC’s 2020 Provisions on the Governance of the Online Information Content Ecosystem, “online information content service platforms” must create mechanisms that allow them to conduct “real-time inspections, emergency response, and the handling of online rumors.” Platforms “must not transmit” what the Provisions define as “illegal” information, and “shall prevent and resist the transmission” of “negative” information. The government can suspend or even shut down platforms if it finds them inadequately compliant. As one Chinese video streaming service noted in its 2020 SEC filing,
“The Chinese government has wide discretion in regulating online activities and, irrespective of our efforts to control the content on our platform, government campaigns and other actions to reduce illicit content and activities could subject us to negative press or regulatory challenges and sanctions, including fines, suspension or revocation of our licenses to operate in China or a ban on our platform, including closure of one or more parts of or our entire business.”
In decades past, when domestic Chinese platforms were less developed, foreign tech companies had a bit more leverage to push back on Beijing’s censorship dictates. In 2006, for example, when Google launched its search engine in the mainland, search results included a notice alerting the user when those results had been censored. Authorities “hated” this behavior, Kaiser Kuo, former director of international communications for Baidu, told MIT Technology Review in 2018, but judged Google too valuable to boot out of the country. The Google notices also gave cover to Chinese search companies, which began to display their own censorship notices. But as China’s domestic industry has matured, providing online services that are comparable or even superior to what is available elsewhere, foreign firms, with their superfluous (or blocklisted) offerings, now have little leverage to wrest concessions from Beijing. Google withdrew from China in 2010, beginning a slow erosion of what little censorship transparency had existed. In the ensuing years, foreign tech companies have proven willing to accommodate themselves to China’s censorship demands. In 2018, leaked documents revealed Google’s plans to develop a censored search engine for China; Microsoft’s Bing search engine has, on several occasions, censored search results Chinese authorities consider “sensitive,” even outside of China.
The government also works closely with companies to conduct surveillance, in order to identify individuals who violate content rules. In 2019, a hacker discovered that China Telecom, one of the country’s three major internet service providers, had been storing customers’ chat logs from platforms like “Tencent QQ, WeChat, and Apple’s iMessage,” including “names, ID numbers, profile pictures, GPS locations, and network information, among other details.” Local police throughout the country used this information to surveil and censor users. According to an account of the hack, described in the book The Xi Jinping Effect, “communication software companies, such as WeChat, provide data, while state-owned telecommunications companies, such as China Telecom, are responsible for integrating the data, and public security agencies conduct screening and manual reviews.”
But China’s censorship regime doesn’t implicate only tech companies. The Party views effective information control as a whole-of-society effort, enmeshing civil society and individuals in the process. Per the CAC Provisions, relevant industry associations should “guid[e] member units to enhance their sense of responsibility” and create “mechanisms for self-discipline.” This merely formalizes what industry associations have been doing for over two decades. In 2002, more than 100 companies signed a “Public Pledge of Self-Regulation and Professional Ethics for China Internet Industry”; in 2021, 14 social media companies signed a “Self-Discipline Convention.” These “industry associations” are closely linked to the Chinese government.
The CAC also requires individual users to “interact civilly and express themselves rationally,” while avoiding both illegal and negative information. Any individual who creates a chat group becomes responsible for removing undesirable content from that group—essentially making anyone who gathers people together online responsible for the behavior of all involved. Both of these rules explicitly push average users to self-censor. And both the government and companies have tried to entice average citizens to report “illegal and harmful content” by offering cash and prizes, essentially crowdsourcing the discovery of objectionable posts.
Recently, authorities have begun to lower thresholds for punishment. For many years, people who merely maintained foreign social media accounts but didn’t actively post could assume the authorities wouldn’t harass them. But in 2024, police began to question people who had merely followed a particular account on the social media platform X. This represented a significant escalation in tactics on the part of the Chinese government. We may see more escalatory measures if national economic concerns continue to worsen.
But what exactly is “illegal” or “negative” information? The definition of forbidden, or “illegal,” information has changed very little since Beijing issued seminal internet regulations in 2000. According to a translation by the website China Law Translate, “illegal” information consists of:
Some of this tracks with federally prohibited content in the U.S. and elsewhere, such as making threats of imminent violence or libelous statements about another person. But other prohibitions, like defaming “heroes and martyrs,” are unique to China (and other authoritarian regimes) and serve as catch-all categories that can mean whatever the authorities want them to mean.
The list of “negative information” appears to be modeled on restrictions previously only levied on professional content creators and broadcasters, and reflects the Party’s sometimes puritanical blueprint for a “healthy society.” It includes:
The “negative” label might apply to content such as celebrity gossip accounts or queer online fora, both of which have been shut down in recent years. Such priggish content prohibitions may seem incidental to the larger censorship project. But they represent a crucial part of how the Party views itself—as a moral arbiter for the masses—as well as a justification for internet monitoring in the first place. An online censorship regime that targets only undesirable political content might strike citizens as too heavy-handed, but a system that also keeps cyberspace less “seedy”? That’s a more credible public good.
Predictably, the content policies of domestic social media platforms like WeChat and Weibo closely echo the CAC’s lists. But rarely do either Beijing or Chinese tech companies offer clear preemptive guidance on what constitutes, to take one list item, “endangering national security,” nor what specific content has been censored under its auspices. For many years, leaked government censorship directives offered one-off insights into online content control. But these leaks have become rare, and they only show what the government has demanded, not what companies actually implement.
To understand what gets censored in practice, a series of researchers, including Citizen Lab Senior Research Associate Jeffrey Knockel, have reverse-engineered companies’ keyword blocklists. Over several years and several different projects, the researchers have grouped these keywords into dozens of specific keyword categories, with many of them falling under a few overarching themes, including:
The first three of these categories will come as no shock to anyone with a passing familiarity with China’s political system, which is determined above all else to protect the Party’s primacy of place. But the fourth category, comprising both pornography and gambling, represents Beijing’s more paternalistic side—and makes up a significant chunk of the content that gets censored.
Censorship, however pervasive, is only a part of China’s program of information control. In addition to removing objectionable content, China’s leaders ensure the promotion of content they would like to see instead. The CAC Provisions mandate that social media platforms “actively present” content that propagates Party doctrine, “Core Socialist Values,” and “other content that teaches taste, style, and responsibility; that praises truth, goodness, and beauty; and that promotes unity and stability.” Posts that brag about one’s wealth? “Undesirable value-orientated content.” Posts about the Beijing Winter Olympics’ “charming” facilities? A great example of “positive energy.”
The Chinese government’s reliance on commercial firms for content curation goes beyond giving marching orders. In some cases, individual Chinese government agencies actually hire for-profit companies to help them monitor internet chatter, especially as it relates to their own organizations and personnel. Such companies alert their government clients to negative posts and provide them with the necessary documentation to request such posts’ removal, as well as upload new, presumably positive, comments “in order to channel the trend of public opinion.”
Such companies are building upon a rich tradition of government-instigated online commenting that aims to appear spontaneous and grassroots. Chinese internet users have famously dubbed government-aligned commenters and trolls the “50-Cent Party,” for the 50 cents the government supposedly pays them for each online post, though at least some of them are actually salaried government employees tasked with “guiding public opinion.” The scorn such posts engendered eventually spawned the even more derisive moniker “fifty-center who brings their own provisions” to describe someone who posts on behalf of the government for free.
Perhaps the best illustration of the philosophical underpinnings of China’s internet regulatory regime, as well as its development over time, is real-name registration. China’s government has long mandated that internet services, such as Weibo and WeChat, register each user with their real-life identities, allowing the government to know the identity of any user upon request. As technology changed, so did Beijing’s regulations, chasing users from internet cafes in the early 2000s to “super-apps” in the 2020s, always ensuring the government would know the identity of anyone online. At points in the 2000s and early 2010s, users, the media, and even companies sometimes pushed back on government mandates for real-name registration, temporarily scuttling implementation on particular platforms. But over time, as Beijing’s rules were codified into laws, and as technological advances made registration and identification harder to avoid, real-name registration became the default across China’s online landscape.
Real-name registration has always increased self-censorship, but recent CAC policies have further ratcheted up the pressure. In 2022, social media platforms began to show users’ IP addresses next to their account profiles and comments. (An IP address identifies where a particular device has connected to the internet, and police can use it to find a user’s physical address.) Even if police had always been able to hunt down users’ IP addresses, their appearance on social media websites likely had a chilling effect on users, who were reminded they were not anonymous online. In 2023, platforms began to publicly display the real names of users with large numbers of followers. And starting in July 2025, the government will implement a new, nominally voluntary, online identification system, taking over the task of real-name verification from companies. Dozens of applications and one of China’s major telecoms have already tested out this new system, which gives users a single ID that identifies them across the internet, making them even easier to track.
“Rather than retroactively identifying individuals with the help of platform companies, a digital ID to which all accounts are connected allows much more comprehensive, granular, and potentially real-time monitoring of user behavior,” says Katja Drinhausen, head of the research program on Chinese politics and society at the Mercator Institute for China Studies. The proposed unified national internet ID system is a “real-life policy expression of fear” about unfettered speech, Drinhausen says. “Only those who can be watched and found should be allowed to speak.”
Real-name registration also has the added benefit of keeping out non-Chinese users, in order to protect “the domestic discourse from outside opinion and commentary that is not aligned with the CCP,” as Sam Ju, a researcher who studies online information controls, wrote in a 2024 report for the Open Technology Fund. The report found 75 percent of Chinese social media apps had implemented real-name registration policies for foreign users, preventing many of them from logging on. “It appears as though China is not only reshaping online discourse rooms to its advantage—it is also drawing a new digital border between a sphere it seeks to influence and a sphere of ‘harmful,’ non-aligned political ideas.”
Beijing can punish both companies and individuals for violating its internet strictures. Companies may have to pay fines or shut down portions of their services; they may also find their services removed from local app stores or suspended altogether. In one extreme case, the CEO of a video-streaming platform was sentenced to 42 months in prison for allowing too much pornography to circulate on the site.
For individuals, the mildest repercussions take the form of deleted posts or deleted accounts—though in the era of the “super-app,” a deleted account can make it impossible to buy groceries or access basic social services. The next level of punishment—somewhere between having one’s posts deleted and facing formal legal penalties—involves being “invited to tea” by local police. During these “tea drinking” sessions, officers might interrogate offenders about their internet activities, personal life, or political opinions, sometimes for many hours. Often, in order to be released, an individual must sign a confession or pledge to avoid untoward internet activity in the future.
But harsher consequences for individuals can also include fines (one programmer had to pay more than one million renminbi for using a
VPN
), public humiliation (such as a televised “apology” forced on a popular microblogger), temporary detention (for “defaming” a Chinese Olympic athlete, or falsely claiming it was snowing when it wasn’t), or even jail time (again, for VPN use). A 2021 study, reviewing public information about punishments levied against internet users, found that
. . . the Chinese government has intensified content controls in online spaces by pushing the burden of censorship further down to the level of individuals. We saw more reports of punishment published by the authorities in the first four months of 2019 than in any single year between 2014 and 2018. . . our results suggest that authorities have broadened their targets of repression from influential non-party thought leaders to almost anyone who dares challenge official narratives.
But, even though some of these consequences are serious, most people who violate censorship dictates will never suffer them. Very few people who repost forbidden memes will be “invited to tea,” but anyone could be.
Thus, enforcement of online censorship in China resembles human enforcement of speed limits in the U.S., in two main ways. First, speed limits encourage traffic to slow down. Drivers know they won’t likely get a ticket for driving a few miles over the limit, and therefore often do so—but they don’t usually drive at twice the limit, judging that too big a risk. China’s censorship thresholds work in a similar way. Most people aren’t too worried about breaking internet censorship rules, knowing that they’re unlikely to get caught as long as they’re not doing anything too outrageous. Internet users may try to slip a wisecrack or two past the censors, but most people won’t risk posting an open call for rebellion.
Second, speed limits, which are often set below what is absolutely necessary for safety, and below the speed at which most people drive, confer broad discretion on police officers to choose which cars to pull over. Moreover, day-to-day enforcement can be capricious. A cop can change where he sets up his speed trap; he might be in a good mood, bad mood, or even taking a nap; his department might be pressing him to collect more revenue by writing more speeding tickets. This means that even “reasonably” speeding drivers can find themselves pulled over—particularly if those drivers aren’t white.
Similarly, in the PRC, the online censorship rules can be quite vague, giving authorities the pretext to declare large swathes of content “illegal” at a moment’s notice. At the same time, enforcement can be capricious. An internet user who thinks they’re well within the bounds of acceptable speech may suddenly find themselves in trouble. And authorities scrutinize and punish the online behavior of certain non-Han ethnic groups at much higher rates than they do that of Han citizens.
Such an inconsistent application of the law might seem amateurish. But in fact, it’s much cheaper and more efficient than strictly enforcing the rules against every single violator. Ultimately, the most effective police officer is the one that lives inside citizens’ heads.
