Censoring the News Before It Happens

Every day in China, hundreds of messages are sent from government offices to website editors around the country that say things like, “Report on the new provincial budget tomorrow, but do not feature it on the front page, make no comparisons to earlier budgets, list no links, and say nothing that might raise questions”; “Downplay stories on Kim Jung-un’s facelift”; and “Allow stories on Deputy Mayor Zhang’s embezzlement but omit the comment boxes.” Why, one might ask, do censors not play it safe and immediately block anything that comes anywhere near offending Beijing? Why the modulation and the fine-tuning?

In fact, for China’s Internet police, message control has grown to include many layers of meaning. Local authorities have a toolbox of phrases—fairly standard nationwide—that they use to offer guidance to website editors about dealing with sensitive topics. The harshest response is “completely and immediately delete.” But with the rapid growth of difficult-to-control social media, a need has arisen for a wide range of more subtle alternatives. For stories that are acceptable, but only after proper pruning, the operative phrase is “first censor, then publish.” For sensitive topics on which central media have already said something, the instructions may say “reprint Xinhua but nothing more.” For topics that cannot be avoided because they are already being widely discussed, there are such options as “mention without hyping,” “publish but only under small headlines,” “put only on back pages,” “close the comment boxes,” and “downplay as time passes.”

We know all this thanks in large part to Xiao Qiang, an adjunct professor at the School of Information at Berkeley, who leads the world in ferreting out and piecing together how Chinese Internet censorship works. Xiao and his staff have collected and organized a repository of more than 2,600 directives that website editors across China have received during the last ten years. Some are only a line or two long; others run to many pages. Some are verbatim, others are paraphrases. Some were collected from Twitter, Sina Weibo (China’s domestic Twitter), and Internet forums, while others were sent to Xiao by editors in China who were frustrated or angry—either at what the directives said or at the fact of censorship itself.

And as Xiao has discovered, the new censorship strategies show the government’s growing awareness of the power of social media. Informal news stories—often accompanied by photos from smart phones—now spread widely and quickly enough that official media lose credibility if they do not at least mention them. In such cases, “on the back page” might be the best option. Moreover, Web users now understand Internet censorship well enough that the issue can itself be one that angers them. (The traditional print and electronic media are censored, too, but directives for them arrive via unrecorded telephone calls, which are much harder to trace and seldom leak. Because the Internet is too large to manage by telephone, its directives go out in writing.) Under the scrutiny of Web users, propaganda officials face the unwelcome task of censoring the Internet while trying to appear as though they are not—or at least not doing it “unreasonably.” This forces them to seek balance. In one instance, a story about two policemen who were killed in an auto accident got out on the social media; censors anticipated an outcry if they “completely and immediately deleted,” so they allowed the story to appear but added the instruction “close comment boxes”—apparently from fear that the boxes might fill with cheers of the kind that normally spring from generalized resentment of the police.

China’s censors use much time and money restricting Internet speech. Consider this summary of directives government officials in Beijing sent to Hunan Province in June 2011 (it was leaked five months later):

All websites should conscientiously grasp the relevant principles and use them to purge any material that:

  1. blackens the image of Party and state leaders or obfuscates the great historical achievements of the Party;
  2. attacks our system or advocates the Western democratic system;
  3. incites illegal assembly, petitioning, or “rights support” activity that harms social stability;
  4. uses price rises, corruption cases, or other controversial events to spread rumors and incite hatred of officials, of police, or of the wealthy that could lead to activity offline;
  5. incites ethnic hatred [of Han Chinese] that harms national unity;
  6. attacks the Party’s systems of managing the media and the Internet by using the slanderous claim that we limit free speech.

But as Xiao has revealed, the censors expend even more effort on the parallel task of “guiding” expression in pro-government directions. When a story reflects well on the Party, Web editors receive instructions to “place prominently on the home page” or “immediately recirculate.” Authorities also organize and pay for artificial pro-government expression in chat rooms and comment boxes. Provincial and local offices of External Propaganda and Party Propaganda hire staff at salaries of about U.S. $100 per month (less, for part-time work) to post pro-government comments. It is hard to say how many salaried commenters exist nationwide, but estimates run to the high 100,000s. Some of this commenting is outsourced as piece-work. A few years ago, people who agreed to do this work were given the satiric label “fifty-centers” because they were said to be paid fifty Chinese cents per post. By now there are commercial enterprises that contract for comment work. Even prisons do it; prisoners can earn sentence reductions for producing set numbers of pro-government comments.

The “fifty-cent” initiative has met with some problems, however. Posts for pay have become so repetitive and mechanical that Web users spot them easily. Such posts also run the risk of undermining opinion that might be genuinely pro-government, because they make any pro-government comment subject to the suspicion that it was done for money. In some circles, mockery aimed at fifty-centers has expanded to include regime apologists of any kind. Someone who thinks that External Propaganda might actually be doing some good by watching the Internet is called a “self-employed fifty-center.” Westerners who praise the CCP are “foreign fifty-centers.”

Xiao Qiang protects his informants by slightly altering dates, names, and word-orders. He needs to do this: in 2005 the journalist Shi Tao was sent to prison for ten years for sending an unapproved document overseas. Xiao and his small staff check the authenticity of what they receive against evidence of actual censorship. Occasionally they detect fake directives and toss them out. The work is tedious and time-consuming, but sometimes they get lucky. One day in spring 2012, as Xiao was using keywords to validate directives, a Google search turned up an item that was oddly labeled “save for evidence.” He opened it to find a very large file that contained a full year of directives to a major province-level news forum. Normally such material is guarded behind firewalls in government servers, but Google found this one in someone’s personal space. Xiao did not know the person, but the label “save for evidence” seemed a hint that he or she might be a website editor who had become disgusted with government directives. Xiao’s archive contains this and three similar large, comprehensive sets of directives.

In late June, ten scholars of Chinese law, politics, society, and language attended a workshop at Berkeley to join Xiao and his staff in analyzing Xiao’s archive. The day sparkled with insights, a few of which were these:

  • One of the principal aims of the government directives is to prevent unapproved groups from organizing through the Internet (noted as “incitement,” “gatherings,” etc.); some of the scholars argued that this goal is even more fundamental than prevention of negative comment about the Party.
  • Because political power and commercial interests are commonly intertwined in China, censorship often merges with something that resembles commercial “reputation management” in the West. In 2008, when a scandal broke over melamine found in the baby formula of the Sanlu milk company, government censors sent out dozens of directives trying to play down the matter. To protect Sanlu? To tamp down general “incitement”? Both? In any case, it is easy to find examples of censorship that protects both political and commercial interests at all levels.
  • In some cases a directive to block an item of news comes out even before the news itself appears. This seems to happen because of worry over what public reaction will be. In August, 2010, Web editors in Hunan received this directive: “The trial in the June 21 murder case in Guizhou will open tomorrow; no medium of any kind is to make any report about either the trial or the sentencing.”
  • In its attempts to garner popular support for censorship, the regime still lumps political speech together with pornography. (The two are similarly “unhealthful to society,” according to the government.)
  • Officials tend to be protective of their own jurisdictions, but not necessarily of others. News of a scandal in Henan, blocked in Henan, might appear in Shanxi. While local officials rarely oppose the central government, that, too, can happen. After Xinhua, the state news agency, reported that a military official in Guangzhou had assaulted a flight attendant in 2012, the Propaganda Department for Guangdong province ordered that Guangdong websites “not republish related Xinhua copy.”
  • Directives aimed at improving the popular image of courts are of mixed quality. Some instruct editors to refer to people facing trial as suspects rather than criminals, and this seems a step forward; but others seem to revert to an earlier mindset, telling editors to “report on the upcoming trial but not on the execution to follow.”
  • News of suicides, the government seems to have concluded, should be blocked, not only to protect the overall image of a “harmonious society,” but also to reduce the credibility of suicide threats, which can be used to leverage concessions from officials.
  • Until about ten years ago, more than half of the reporting in the Xinhua information system was “internal reference” (classified) reporting to the leadership on what people in society were actually doing and thinking. With the dramatic growth of the Internet, this function of Xinhua has shrunk. The Internet makes it far more readily apparent—to anyone—what people are doing and thinking.

In the end, though, none of the parts of Xiao Qiang’s project is as important as the whole that it seeks to reveal. In recent years China’s rulers have been building a gargantuan Internet censorship system. It is many times larger than any comparable effort, in any era. Soviet-era censorship and China’s own Mao-era approach to the press were tighter and also included elements of guiding as well as outright suppressing of information. But no system in the world has been remotely as large in the number of details it attended to or in the number of people devoted to the work. (The recently disclosed efforts of the NSA to store, in secrecy, the metadata of electronic communications of nearly all U.S. citizens, however deplorable, are not nearly as far-reaching. And the NSA activity apparently has been limited to collecting data and occasionally to eavesdropping—not to blocking, manipulating, or manufacturing what is said.)

Most of the Chinese system remains obscure; Xiao’s 2,600 directives show only a corner of it—or, more precisely, several small corners. On June 17 and 18, Xiao attended an international “Freedom Online” conference in Tunis. One of his hosts brought him to visit the nearby Bulla Regia ruins of an ancient Roman city. They observed some walls here, some columns there, a mosaic over there—remnants that spoke of something much grander. Xiao was reminded of his research project—except that, in his case, the huge mysterious picture was slowly coming together, not deteriorating.