Research into WeChat censorship and Chinese online rumors
The power of online rumors extends far beyond their contents. Even when they are not specifically aimed at rallying masses to a cause, the transmission of even the most dubious of claims is still indicative of another kind of collective movement: an attack on the pervasive censorship system, which has encouraged online users to develop an extreme form of skepticism wherein, as Hu Yong, a professor at Peking University's School of Journalism and Communication, suggests, "news looks like rumor and rumor looks like news." In such an environment, the spread of rumors on the Chinese Internet is not surprisingâand considering their implicit criticism of the credibility of authorities, neither is the effort by government officials and social media companies to restrict them.
ââChinaâs Rumor Mill:Â Why Beijing Is Cracking Down on âUnverifiedâ Information Online,â Foreign Affairs
The above is from an article I wrote last month in Foreign Affairs based on my research into what kind of posts are censored on the public accounts platform of WeChat, a mobile app that is incredibly popular in China right now. Among my conclusions (you can read the full report, âPolitics, Rumors, and Ambiguity: Tracking Censorship on WeChatâs Public Accounts Platformâ on the Citizen Lab blog) are:
Sensitive keywords are found in a greater percentage of censored posts than normal (uncensored) posts.
The data contains evidence of automatic review filters preventing posts with certain blacklisted keywords from being published. Examples include ć ć (June 4), ć€Șćć (princeling), and keywords relating to Falun Gong.
An analysis of 150 censored posts reveals that rumors, speculation, and political commentary were also being censored. Censored content included posts which contained outright falsehoods, tabloid gossip, and sensationalismâa number of which appear fairly harmless. This may be a reflection of the ongoing âanti-rumor campaignâ sweeping Chinese social media.
The report discusses the collective power of rumors and ambiguity in censorship, issues raised by how WeChat controls information on its public accounts platform.
The last two points are the focus of the Foreign Affairs article. In conjunction with the article, I had a brief discussion with Emily Parker on the new social media site Parlio. Itâs been fascinating studying how people use WeChatâs public accounts/blogging platform, and I look forward to tracking all this going forward.
Anya is live and ready to show you everything. Watch her strip, dance, and perform exclusive shows just for you. Interact in real-time and make your fantasies come true.
â Live Streamingâ Interactive Chatâ Private Showsâ HD Quality
Anya is LIVE right now
FREE
Free to watch âą No registration required âą HD streaming
Fit for Public Display: Rethinking Censorship via a Comparison of Chinese Wikipedia with Hudong and Baidu Baike
So sad I wasnât able to attend Theorizing the Webâs annual conference this year, so I figured Iâd post my 2014 presentation about the many challenges of figuring out what exactly is censorship in China as well as how important companies are for mediating the negotiation over acceptable sensitive material online. Iâve previously uploaded it as a slidedeck and have a more detailed report posted on the Citizen Labâs blog (I also presented versions of this talk at AAS 2014 and WikiConference USA), but I figured it might be easiest to share it here in a single, readable blog post.Â
(Note: all findings accurate as of April 25, 2014, all data accurate as of August 2013 except where noted.)
Hi, my name is Jason Q. Ng and I primarily research Internet censorship and the effects it has on netizens and tech companies. Obviously, much of this sort of work centers on China, where a lot of the most âinterestingâ online censorship is taking place today.
So Iâll be sharing with you the results from a project on Chinese online encyclopedias that I worked on last summer. Hopefully itâll not only add to the data we have on Chinese censorship, but itâll also contribute to the conversation about how hard it can be to pinpoint when âcensorshipâ is taking place, especially when it is as decentralized as it is on these encyclopedias. This project will hopefully also touch on the ways which Chinese netizens are still able to create and share content despite restrictions imposed on them, as well as the fluid boundaries that authorities place on sensitive discussions.
So based on that introduction, you can safely assume that weâll be discussing how censorship isnât a black and white issue, but this project actually began in the tradition of attempting to identify clear-cut instances of censorship, which many researchers have successfully done. For example, folks like Xia Chu have worked out in great detail most of the websites which are blocked by Chinaâs so-called Great Firewall.
Others like Jed Crandall and The Citizen Lab have figured out what words on a webpage or URL might trigger your connection to a website to be dropped [âŠ]
Or your chat message to be unsendable.
Others look at social media, figuring out what keywords or images cause posts to disappear from timelines [âŠ]
Or from sites altogether.
What all these examples have are clear-cut indicators that censorship is being employed. There was a result we expected to receive, but due to some sort of interference, it didnât happen, and this we chalk up to censorship. Whether thatâs due to the government that mandated certain topics as off-limits or a company deciding to preemptively restrict certain content is less clear, but in these examples, the user is definitely not in control.
This can get muddled on some social media because the censorship can be distributed down to the users themselves. They can report their fellow users for content violations, like in this instance from WeChat, taking down posts that may have escaped the initial notice of authorities, or as in online encyclopedias, they can also delete and change the collective content directly.
But putting aside these thorny questions of attribution, how do we go about even identifying potential censorship to begin with? You could go the traditional route of figuring out what keywords you canât search for, what articles you canât create, and what articles you canât access from inside China [âŠ]
as others have already begun to do. But one of the neat things about Chinese online encyclopedias is that you can not only do traditional testing, but you can also do quasi-experiments due to the fact that unlike here in the US, where Wikipedia is the only major online encyclopedia, in China you have multiple ones.
So of course, we have the Chinese-language version of Wikipedia, where sensitive content and topics flourishâso long as they are notable, factual, and described in a neutral manner. We take for granted that Wikipedia does not intentionally censor the content on its website, and thus, it serves as a sort of baseline for comparing with the two much more popular online encyclopedias in China: [âŠ]
Hudong Baike [âŠ]
And Baidu Baike.
And in fact, Hudong and Baidu Baike are much more popular than the Chinese-language version of Wikipedia. Baidu and Hudongâs dominance can be credited to the fact that theyâve leveraged competitive advantages and built sites that mainland Chinese users prefer. Wikipedia China instead is mostly edited and viewed by users from Taiwan and Hong Kong.
So this project began with a question: everyone âknowsâ Hudong and Baidu Baike, like all Chinese websites, have to restrict certain kinds of content, but is it possible to empirically prove that censorship is taking place on the sites? If so, weâd have to think about how weâd even look for indicators of censorship in the first place.
The most obvious would be the lack of certain articles on topics that are definitely notable, for example, here with an article for the June 4th incident, which exists on Wikipedia China but not on the other two.
Second might be by identifying articles that are much shorter on Chinese baike than their counterpart on Wikipedia China. For example, the Wikipedia article for Incest is over 18,000 characters long, while on Baidu it is only 212.
Third, some especially sensitive or controversial articles are unable to be edited except by users with much higher privileges than ordinary members.
Being unable to change an article is not in and of itself a sign of censorship; for instance, Wikipedia âprotectsâ certain articles to prevent vandalism. Baidu and Hudong no doubt have similar intentions in mind as well, but what matters here is transparencyâwhile Wikipedia publishes a list of all protected pages, as far as I can tell, no such corresponding list exists for Hudong and Baidu Baike. Plus, who got to decide that certain articles be locked? Was there an open discussion about whether something should be protected, as is typically the case on Wikipedia, or did someone higher up unilaterally step in to shut it down? Again, what started as an attempt to find clear cut instances of censorship is already starting to wade into murky waters.
So, for the purposes of this project, I made a first pass at trying to identify potential censorship by looking at three indicators: if a Wikipedia article doesnât exist on either baike, a Chinese article is much shorter than its Wikipedia counterpart, or an article is locked. As for who might be responsible for these issues, we should keep in mind that there are numerous possible reasons. Judging whether or not these factors are actual instances of top-down censorship or due to explainable, organic reasons can be quite tricky. Because of the multiple layers of oversight, what may appear to be outright censorship may simply be a lack of relative interest, or perhaps a case of self-censorshipâor at this early stage in my project, a failure in my data collection.
Doing this sort of experiment required me to match up articles across each of the encyclopedias. I automated the process by writing a simple script, which took various keywords and ran them on all three services, extracting and comparing the pages returned from each. As for what keywords to use, in this initial test I chose a short list of what I thought might be very interesting and relevant keywords. The first three categories here contain keywords that allowed us to test for how each of the Chinese encyclopedias would handle sensitive topics. I added to that two other categories that covered a broad base of both popular and topical issues according to Wikipedia China.
There was some overlap in the lists, and in the end I had 5,144 distinct keywords to be used for testing, and roughly 2/3 were successfully matched to a corresponding article on either Hudong or Baidu Baike.
I gathered all sorts of data from each of the three encylopedias⊠And from Hudong and Baiduâs pages, I was also able to capture the number of likes and number of edits to the page, which are interesting metrics which Iâll talk about in a moment.
Anyway, so now the fun stuff. Hereâs a select list of Wikipedia articles that donât exist on either baike. Of course, we have to keep in mind that while some of these missing articles may be due to enforced censorship, others may simply indicate a lack of interest on the part of mainland Chinese users, who are the primary creators of content on Hudong and Baidu Baike. So in that context, itâs not surprising that a small Hong Kong bookstore may not have an entry on either baike. However, the absence of articles on the Falun Gong and Charter 08, even if they arenât topics of daily discussion amongst most Chinese Internet users, clearly indicates that thereâs an intentional effort to restrict certain sensitive topics from the sitesânot surprising.
This is going back to that table of articles that are much longer on Wikipedia than Baidu or Hudong. And actually, if you go through these one at a time, some of them would actually be re-classified into the first group of non-existent articles, for instance, the one on Tor on Baidu is actually just a dictionary entry for the English phrase âterms of referenceâ while on Hudong itâs about a tank. So in actuality, an article about Tor the software doesnât exist on Hudong or Baidu Baike [âŠ]
Though if you look at the version history for the âtorâ article, it may very well have included information in the past about using Tor to beat censorship, before it was itself censored.
But anyway, going back to the topics on the chart, it appears this list isnât purely random: itâs filled with sensitive topics, and thus we can assume some form of regulation appears to be in forceâthough I should point out that a Japanese manga having a 16,000 character article on Wikipedia says more about some Wikipedia users than it does about censorship in China. But anyway, itâs probably not surprising that the Wikipedia article for the Masanjia forced labor camp was roughly 25 times longer than on its China-based counterparts as of last August. But whatâs more interesting to me is that an article about this labor camp exists at all and that someone contributed to that article, and it was accessible to mainlanders despite the fact that itâs an obviously sensitive topic.
Interestingly enough, I went back to check on the article last month to see what had been written in it, and both China-based articles are actually longer than they were when I captured this data last summer. In fact, the Hudong one is now over 4,000 characters long, with specific sections detailing the types of torture that were alleged to have taken place there.
By contrast, the Baidu entry now stands at 1,200 characters, and it also mentions that torture has been alleged at the camp.
But the bulk of the entry is about the investigation that refuted those claims and how the conditions at the camp were actually of acceptable quality and that media shouldnât fall into the trap of spreading falsehoods. So again, we see the fluidity of the boundaries of acceptable discourse shifting, this time with different results from different sites. We canât simply push a narrative about authorities winning or netizens successfully fighting the censors.
Itâs not a zero-sum game of this sort, but rather itâs a process. Instead of thinking of it as a battle over free speech or a cat-and-mouse game, both of which are true to a certain extent, the better metaphor might be to think of the camps as engaging in a dialogue, with information about certain topics like a female labor camp being negotiated on. And even then, we shouldnât feel comfortable generalizing netizensâor even authoritiesâas this bloc, when in fact theyâre composed of numerous groups with different motivations.
Anyway, the third of our censorship indicators that I mentioned: protected articles. Hereâs the breakdown. Of the 5144 keywords tested, Baidu had 403 protected articles, nearly twice as many as Hudong, but if you look at the large overlap of protected articles between Hudong and Baidu, itâs apparent that the distribution isnât purely random.
Hereâs a sampling of the list of articles protected on Wikipedia, which unlike the two baike, is publicly available and posted on their site.
These are Baiduâs protected articles. Again, most of the data that was collected fits with what weâd expect based on our past work with sensitive topics on Chinese websites.
And finally Hudong. Of note is the sensitive topics that do exist on these Chinese baike: for instance children of high-ranking officials, which is a major issue in China, especially with stories of how these children are becoming incredibly wealthy due to their family connections. The Hudong article doesnât name names, that might be crossing the line, but it does note their negative effect on society. So itâs not like all critical information has been wiped clean from the Chinese Internet, itâs just being managed in ways that arenât totally transparent. There exists this in-between space where the topic is clearly sensitive, but netizens have the capacity to perhaps push up against the boundaries, like what took place with the Masanjia prison camp article. [âŠ]
Or this Baidu article on political prisoners. The article is only a brief stub that speaks in generalities, with no names or links to lists of political prisoners around the world like youâd expect. But the fact that itâs been viewed 54,251 times, this despite the fact that the article is only 300 some characters long, means that folks are actively seeking out this content, and some are even âlikingâ it.
And indeed, before I close, Iâd like to dwell a bit on the âlikesâ each article has, which was one of the variables I mentioned that I was able to collect for this project. In case youâre curious, the most liked articles as of last August on Baidu are mostly pop stars: [âŠ]
Jay Chou, Michael Jackson, Jacky Wu, and ⊠the that other famed Chinese Lothario, the Great Helsman himself [âŠ]
Mao Zedong, with over 249,000 thumbs upsâthough all of them were blown out of the water [âŠ]
by the South Korean pop group 2NE1, whose Baidu Baike article was liked nearly 2 million times. So what are users trying to say when they mark something as helpful by clicking the thumbs up icon on each baike service? For instance, what can we conclude from the fact that more people have classified disgraced politician Bo Xilaiâs article as âhelpfulâ than they have for President Xi Jinpingâs?
Or that 320 citizens âlikedâ the Baidu Baike entry on the Masanjia prison camp, the one which refuted that any torture took place. Are these folks patriots who support the governmentâs position? Or do these âlikesâ signal their endorsement of a topicâs existence even if the content is drastically different from what they believe in? I tend to think itâs the latter, and I certainly donât think we should look down on this sort of less active online engagement, what some might consider a kind of slacktivism, especially considering more active types of support for the wrong cause could land a user in trouble. As Internet scholar Guobin Yang has saidâand Iâm paraphrasing a bitâonline play of this sort isnât devoid of politics. These kinds of interactions may be one of the few ways users can safely engage with issues which are typically locked down or censored.
So by looking not only at what content and data doesnât exist on baike, but also at the content and indicators that do, we can think about what kind of knowledge and information is fit for public display. If articles are shorter on Hudong and Baidu, what information do they carry? Does this information reveal anything about the authorsâ intentions? By examining the topics and articles that are left visible and considering the motivations behind those who view, edit, and approve of these articles, we can seek a more nuanced view of the typical narratives about censorship in China. Trying to understand what sorts of expressions netizens are making via these online encyclopedias, despite whatever restrictions are placed on them, is as interesting as studying the censorship itself.
Censorship messages return plus various blocked Xi Jinping keywords
Last October, I noted that Weibo had removed their censorship notice for blocked searches on the site. At the time, I surmised that this might
be a temporary bellwether for enhanced censorship on the site due to a particularly sensitive period in Chinese politics;
mark a shift toward more obscured censorship; orÂ
simply be Weibo testing out new tactics.
Now that the censorship messages have returned to Weibo (apparently some time around Jan 24, 12pm Beijing time according to my logs), it's still uncertain why Weibo removed those notices nor why they reinstated them. If I had to guess, I'd go with c. Whatever the case might be, it makes tracking sensitive keywords easier for researchers, giving us another data point into what appears to be considered off-limits for online discussion and sharing by Chinese authorities--a judgment which sometimes changes day to day.
To celebrate the return of these censorship messages, below are 20 Xi Jinping (äč èżćčł)-related keywords, most of which have not been previously found to be censored and are currently confirmed to be blocked on Weibo. (äč èżćčł on its own is searchable and returns a set of highly filtered results, mostly from verified, official state media.)
How a Weibo post gets censored: what keywords trigger the automatic review filters
I wrote up on The Citizen Lab's blog a report into the various ways a post can be censored on Weibo (see image above), with a particular emphasis on the "automatic review" filters. Among the preliminary results were the identification of 66 keywords which cannot be posted to Weibo and 133 keywords which cause posts to be invisible and camouflaged.
Below is my summary and links to the data/screenshots (which include lists of the keywords which trigger various kinds of censorship). Hopefully more research will be done into these other paths of censorship.
(For more of my recent writing, see: Guernica interview with Evan Osnos, Technology Review essay on what memes can and can't do in the fight against censorship, World Policy Journal piece on the challenges Chinese tech companies face, Wall Street Journal article on WeChat, and LA Review of Books review of a book on Internet activism.
Summary:
Part 1: Weibo has removed their conventional censorship notice from searches on the site. This may be a bellwether for enhanced censorship on the site due to a particularly sensitive period in Chinese politics or it may mark a shift toward more obscured censorship. Or it may simply be Weibo testing out new tactics.
Part 2: Automatic review of content refers to moderating messages before they get circulated widely. A number of studies have described these mechanisms, which include keywords which trigger your post to become hidden or your inability to post in the first place. We sought to outline with more clarity the pathways a Weibo post might takeâfrom submission to potential censorship.
Part 3: We accomplished the above by posting sensitive keywords and tracking how they were censored or not based on a number of factors. This is only a preliminary set of tests and will hopefully serve as a basic methodology for others who are interested to generate more rigorous testing.
Data and screenshots: Weâve posted to Github the data used for this test, lists of suspected keywords which trigger automatic review, as well as the screenshots of the various censorship messages.
Data
Part 1:Â Weibo search data of CDT keywords, tested May 2014 and Nov 2014 (CSV)
Part 2 & 3:Â Weibo censorship testing data of probable automatic review keywords, tested Nov 8, 2014 and Nov 10, 2014
66 keywords which cannot be posted to Weibo according to our preliminary test (explicit filtering)
14 keywords which return a data error (implicit filtering)
133 keywords which cause posts to be invisible and camouflaged
Screenshots:
Explicit filtering message (box 1 in Figure 1): Chinese | English
Implicit filtering message (box 3 in Figure 1): Chinese | English
Comparison of missing messages in timeline due to âcamouflagedâ/invisible posts (box 2A1 in Figure 1): own timeline vs when viewed by another user
Message when visiting deleted weibo post (box 4a in Figure 1):Â screenshot
Post deletion message in inbox (box 4a in Figure 1):Â screenshot
Account warning, 48 hour ban (box 4C in Figure 1):Â screenshot
Account abnormal notice (box 4C in Figure 1):Â screenshot
List(s) of Chinese keywords for censorship testing and sensitive content collection
Last week, a researcher during The Citizen Lab's annual Connaught Summer Institute workshop raised an interesting problem. She wanted to test for censorship on a Chinese online service, and she had somewhat limited resources and time. What keywords should she use for her test?
In theory, this is a solved problem, what with the numerous lists of censored and sensitive Chinese keywords available on the web, including those shared by this site. However, sometimes the keyword list may be too broad for one's taste, or may simply have too many keywords to efficiently use. And plus, what if I only want to test the most sensitive of the keywords, e.g., Falun Gong, June 4, Xi Jinping, and so on? For those not experienced in Chinese or Internet censorship, this can be a daunting task to winnow down already existing lists to something more usable.
Thus, a few of us sat down at the workshop and we collected 8 known Chinese keywords lists (see below) and aggregated them together in a single, easily share-able and sortable file, which we've posted to Github. The CSV files contain not just the keywords, but all sorts of other info like translations and tags (though not all of them; it's an ongoing project which you are welcome to contribute to since it's an open-source project).
As of Aug 4, there are 8,087 sensitive keywords collected from 8 different lists. To get a sense of what data is included in these CSV files, you can view a spreadsheet of these 8,087 keywords sorted by the number of lists they appear on.
Creator Tested on/found from # of keywords Year Method + source The Citizen Lab Sina UC 1,818 2013 reverse engineered from the client; more analysis here; download link The Citizen Lab Tom-Skype 2,574 2013 reverse engineered from the client; more analysis here; download link The Citizen Lab LINE 673 2014 reverse engineered from the client; more analysis here; download link Jason Q. Ng (Blocked on Weibo) Sina Weibo 839 2013 running Wikipedia China article titles through Sina Weibo search; more analysis and book Xia Chu Great Firewall 669 2014 HTTP request scans of Wikipedia China articles to see if they'd trigger GFW block; more analysis here; download link (removed duplicates and keywords related to meta and user pages) China Digital Times Sina Weibo 2,448 ongoing crowdsourced testing of suspected sensitive keywords on Sina Weibo; more analysis on CDT and in CDT's Grass Mud Horse Lexicon e-book; download link GreatFire.org Wikipedia 488 2013 testing to see if Wikipedia pages are available in China; more info; download link Google/ATGFW.org Google/Great Firewall 456 2012 ATGFW.org and GreatFire.org reverse engineered the keywords Google was using to warn users of censorship while using their service in China; download link
To follow future changes to these lists, you can follow the Github repository. You are encouraged to adapt and update these lists as you see fit, however please do credit back to the Github repo if you do. Hopefully this is helpful to researchers who are searching for sensitive content in Chinese or testing for network interference.
Anya is live and ready to show you everything. Watch her strip, dance, and perform exclusive shows just for you. Interact in real-time and make your fantasies come true.
â Live Streamingâ Interactive Chatâ Private Showsâ HD Quality
Anya is LIVE right now
FREE
Free to watch âą No registration required âą HD streaming
64 Tiananmen-related words blocked today (June 4, 2014)
     (photo credit: CND.org?)
Today, on the 25th anniversary of troops being ordered into Tiananmen Square to clear student demonstrators, I tested several hundred June 4-related keywords on Weibo. I used the same set of keywords that I tested last year at The Citizen Lab, and I found relatively similar levels of censorship this year. Also like last year, the enhanced restrictions on keyword searches were apparently implemented specifically for the anniversary: for instance, ćŠć (tank) and ć ć (6-4, i.e., June 4) were free to be searched as recently as May 11.
After the jump are the 64 keywords that are currently blocked from searching on Weibo (though some will no doubt be unblocked once this sensitive period passes):
ć°éæ¶šä»·æčæĄ (subway price increase plan / dĂŹtiÄ zhÇngjiĂ Â fÄng'Ă n) refers to Beijing's plan to shift its subway fares from the current rate of 2 yuan per trip (roughly 30 cents) to a distance-based price system. Beijing has among the cheapest subway fares in China, and the system is heavily subsidized by the state, with the government claiming they lose 5 yuan on every trip.
Why it is blocked: The plan was floated in late 2013, igniting outcry among subway users online after forthcoming price adjustments were officially confirmed in March of this year. The story picked up steam again last week when a photograph of a supposed document outlining the specific price increases was circulated. The alleged document stated that short trips would rise to 3 yuan, and long trips would be capped at 5 yuan.
However, while previous discussion of the price increases show results that appear to have been mostly uncensored, with large numbers of posts criticizing the plan (though a fair number do support it, arguing that the increases may reduce the subway's notorious overcrowding), posts sharing this new document were censored according to Free Weibo. Authorities quickly denounced the document and stated that pricing plans were still being evaluated. Not all posts containing the photo of the document were deleted: for instance, this one from China Daily, a state newspaper, were allowed to stand as they contained a message refuting the document.
No doubt, Beijing authorities are very sensitive to the potential for increasing unrest surrounding the issue, which gets at the rising income inequality in the city. (More about this in my WSJ article.)
On March 1, an organized group of knife-wielding attackers indiscriminately stabbed passengers at a railway station in the Chinese city of Kunming. Over 140 were injured and 33 were reported killed. Suspicions on Weibo and state media turned to Uyghur separatists from the northwestern province of Xinjiang. China Digital Times reported that the government had issued the following directive to news organizations:
Media may publish a moderate amount of criticism and Internet commentary which oppose terrorism and violence and which condemn the killers. However, do not hype this incident.
The following word combinations are found to be currently blocked from searching on Weibo:
ææ + æ°ç (terrorist + Xinjiang / has been blocked in past)
ç æćżç«„ (children stabbed and killed / also connected to other past incidents)
Why it is blocked: And while none of the articles make direct mention of Zhou Yongkangâs name it appears that Zhou is being methodically prepared for a downfall. (Update: As of February 27, a Baidu Baike user edited the articleâs oblique reference about âa certain retired member of the standing committeeâ to Zhou directly; one wonders how quickly that version will stay.) The investigation of Zhou, who is already reportedly under house arrest, is drawing intense scrutiny from domestic and foreign observers as to how far Xi Jinping is willing to crack down on corruption at even the highest levels of the Partyâor use the charge as a fig leaf to take down someone who was once Bo Xilaiâs most ardent supporter.Â
While blocking a politician's name is often about protecting them from criticism, one might argue that in this case the government is less concerned with protecting Zhou than with controlling any sort of discussion that might spring up from an opening up of Zhouâs misdeeds for public discussion. The governmentâs crackdown last summer on online rumorsâwhich included the targeting of journalists and anti-corruption watchdogs who were once encouraged by authoritiesâshows that officials are still incredibly wary of the unpredictable nature of Internet discourse.
äžćźçœȘ (seven deadly sins / qÄ« zĆng zuĂŹ) are a category of vices that according to Catholic teachings threaten a person with eternal damnation. They are wrath, greed, sloth, pride, lust, envy, and gluttony. The phrase was blocked on Weibo until October 2012, at which point it was unblocked. It remains searchable to this day.
Why it is blocked: At the time I was writing my book in 2012, I theorized that the word might be related to religious sensitivity or some other moral issue, but otherwise was quite mystified as to the specific reason for the censorship. However, while reading Emily Parker's Now I Know Who My Comrades Are, I finally connected the dots: in the book, she interviews numerous Chinese bloggers and activists, including Michael Anti. She notes that in 2002, Anti, who was fed up with traditional Chinese news media, wrote a guide for aspiring reporters entitled Manual for New Journalists (æ°æ°é»äșșèȘćŠæć), which included exhortations like:
After weâve said âF**ckâ to the giant media system we can begin our individual journeys through the desert to become new journalists. Maybe we wonât successfully reach the Holy Land of freedom of the press, but at least we will leave the enslavement of truth. (translation from Parker)
Note: If you are interested in digital activism, I highly recommend you take a look at Emily Parker's Now I Know Who My Comrades Are, out next week. I hope to write more about it later, but suffice it to say it's an insightful look at how some bloggers and activists in China, Cuba, and Russia are using the Internet. Filled with interviews of these folks on the front lines, it was an especially good palate cleanser for me after reading Evgeny Morozov's To Save Everything, Click Here;Â yet she also does a very good job acknowledging the limitations of the Internet and those who utilize it. Clearly, as evidenced by this post, I learned a lot from it.
Anya is live and ready to show you everything. Watch her strip, dance, and perform exclusive shows just for you. Interact in real-time and make your fantasies come true.
â Live Streamingâ Interactive Chatâ Private Showsâ HD Quality
Anya is LIVE right now
FREE
Free to watch âą No registration required âą HD streaming
èèĄć (old hutong / lÇo hĂștĂČng) is an urban feature found in the historic districts of several Chinese cities, most notably in Beijing. Hutongs are literally the narrow streets or alleys in these old neighborhoods, some of which trace their roots back to nearly a thousand years ago, but hutongs now generally refer to these old neighborhoods themselves and the distinctive style of architecture and traditional culture held within.
Why it is blocked: Hutongs stand as a marked contrast to the new commercial and dense residential buildings found in cities across China. Not surprisingly, hutongs have been the source of numerous controversies, especially in recent years as urban development in China continues. The destruction of hutongs, which admittedly has been ongoing for centuries in China, received particular attention in the run-up to the 2008 Beijing Olympics, for which city officials razed numerous old neighborhoods in order to build infrastructure and modern buildings. It was reported that citizens who were ordered to vacate their homes were undercompensated, and the loss of history was mourned by local residents, preservationists, netizens, and tourists alike. It is this combination of citizens protesting the loss of their homes--and as with any story about land grabs in China, a whiff of corruption between city officials and the developers who stand to profit the most also hangs in the air--and foreign media attention that causes èèĄć to be a sensitive term.
Updated Jan 30: The peerless Brendan O'Kane smartly points out that èèĄć could simply be a mocking nickname for Hu Jintao, in which case èèĄć would translate to Old Comrade Hu. I wasn't aware if this nickname was popularly used, but even if not, this would not be the first time an incredibly obscure reference to an official was blocked because it was insulting. However, a little sleuthing reveals that it has been used, although apparently not always in a mocking fashion. Some of the references appear genuine (though how much irony is being lost on me, I don't know since one would have to be part of the community to really get if it's an in-joke or not.)
Another theory (though as @bokane notes, it's not quite grammatical): It could also be an abbreviation for èĄéŠæ¶èććż, that is Hu Jintao's old comrades (ć might also be short for ććŠ, classmate). In that case, èèĄć would be a criticism of the Communist Party patronage system, wherein top officials promote and appoint their longtime friends, business partners, and classmates. Hu Jintao wasn't quite as notorious as some top leaders for bringing his old-boy's network with him to the top (or perhaps he wasn't as successful at it as Jiang Zemin, whose Shanghai Clique ruled much of Chinese government and business throughout his time in power), but the so-called Youth League Faction was seen as Hu's base of support. Though Hu Jintao and numerous other top officials are now technically unblocked from searching on Weibo, many combinations of the surnames of Hu, Wen (Jiabao), and Xi (Jinping) with other words are blocked still, and èèĄć would fit that pattern.
Here's a long, wide-ranging interview I did with VICE magazine about Internet censorship in China. Thank you Reihan for the thoughtful questions and for putting up with my rambling.
Also, if any folks are in DC, I'll be giving a lunchtime talk at Georgetown's School of Foreign Service on Friday, Jan 24. It's free and open to the public, so please share with any who are in the area and might be interested. Much appreciated.
Comments and takeaways from Xia Chu's "Complete GFW Rulebook for Wikipedia"
Note: Update to "64-byte search string limitation indicates Weibo and GFW" section added Jan 11
If you are interested in Chinese Internet censorship, I highly recommend you flip through Xia Chuâs latest update to his* research project âComplete GFW Rulebook for Wikipedia.â This latest revision of a document originally released last October, it identifies a massive list of actual trigger words (which Xia calls ârulesâ because they are often attached to specific conditions) which cause a Chinese Internet userâs connection to specific sites like Wikipedia to be disrupted by the Great Firewall (GFW). Not only that, it also includes a list of over 3,600 websites that he has currently confirmed to be unreachable from within China due to the GFW. The conclusions in the paper donât necessarily upend anything that we thought about the GFW, but if you want a peek behind the curtain of how the GFW works (big takeaway: ITâS REALLY HAPHAZARD), this is as close as we can currently get.
The key to efficient testing on GFW or Weibo
Xiaâs paper primarily looks at Wikipedia and the specific pages on the site that are unreachable from within China, utilizing Wikipediaâs open-source sharing of article titlesâthe same corpus I used back in 2012 for uncovering sensitive words on Sina Weibo. Suffice to say, he tested the 700,000+ Chinese and 4,000,000+ English keywords (in addition to 1,000,000 URLs from Alexa, 700,000+ IP strings, and 600,000+ Apple AppStore titles) a lot faster and efficiently than I did with my script back in 2012. However, Iâve since re-wrote my testing script last year and was struck by how similar his method for identifying sensitive keywords is to mine these days (see his section 3.3).
Basically, one of the keys to efficiently identifying words which trigger censorship on the Great Firewall or Sina Weibo is to pack together multiple keywords together and test them at the same time. This relies on the knowledge that for the sake of simplicity and speed, GFW and Weiboâs censorship rules are not sophisticated and will block an entire URL/phrase if any part of the URL/phrase contains a blocked keyword or sets of keywordsâfor the most part, one doesnât have to worry about whitelists and other more complex rules. As Xia notes:
In all the cases Iâve examined, GFW never uses regex features like metacharacter, character classes, or Boolean âorâ (boolean âorâ can be viewed as equivalent to multiple rules). This is not unexpected, because GFW devices need to scan the huge internet trafïŹc between China and the world in real time, for this purpose, plain string matching combined with boolean âandâ is the most cost-effective approach.
Thus, if you know that the corpus you are working with mostly contains non-sensitive terms and you are merely trying to winnow down and identify those that are censored, combining terms and searching all at once may allow you to quickly rule out those that are non-sensitive.
However, if the sentence is unsearchable due to filtering, we must now identify the specific terms which caused the censorship. In this, Xia and I are again mostly in sync. Our primary method is to move from one end of the censored string, removing one piece at a time until the string is no longer blocked, at which point we know that that piece was crucial to the blocking. Once you figure out where the sensitive string begins, you go from the other end, figuring out where the sensitive string ends. (We differ in that it looks like Xia went character-by-character whereas I went by tokenized word-by-tokenized word, and Xia does some extra checking by concatenating different strings.)
So a summary: say you want to test the words AA, BB, and CC, and BB is the censored term. Rather than make three separate tests, you concatenate all the terms and test AABBCC. AABBCC is banned, so now you remove CC and test AABB. Still banned. Now try AA. Whoa, no longer banned, so you put BB back in and go the other way. You drop AA and test BB. Indeed, itâs banned and since there are no more words, you can stop searching. You may notice that this requires 4 tests, one more than if youâd tested the words one-by-one. However, if the majority of your words are clean, youâll more than make up for it by ruling out so many non-sensitive terms all at once in other tests. Iâm sure Xia went through the same process as I did trying to figure out the right balance between efficiency in ruling out long strings of websites/words and labor in narrowing down to the trigger words after a phrase has been initially identified as containing a banned word. But basically, this is the technique we both use to efficiently identify keywords which elicit censorship responses.
64-byte search string limitation indicates Weibo and GFW use the same censorship platform
[Updated Jan 11: After some correspondence with Xia, I realize now that I was confused on some of the distinctions he made in his paper. While indeed there does exist a 64-byte limit in both Weibo and GFW censorship (the second paragraph of this section is still true), these limitations are in fact applied to different things. The GFW limitation applies to the rule which is used for detecting censorship. So for instance, if the page that censors want to block on Wikipedia is "zh.wikipedia.org/wiki/2013ćčŽćæčćšæ«æ°ćčŽçźèŻèą«ć æčäșä»¶", then so long as a part of a HTTP request contains those first 64 bytes (the above string in UTF-8 is 68 bytes long), it will be blocked. It is the rule itself that is truncated. However, the 64-byte limitation I detail on Weibo is to the HTTP request itself, which is very different. In my surprise at identifying the coincidence that both GFW and Weibo exhibit 64-byte limitations, I momentarily confused the two for the same thing, and jumped to the conclusion that perhaps the two were using similar censorship platforms. In fact, as Xia reminded me and to which I agree, GFW censorship is by all reported accounts done within the backbone of the Chinese Internet and primarily performed with various networking hardware, while Weibo and other censorship by online content-providers is likely largely done with server-side software. I'll leave my conclusion below as a warning to myself to consult folks before rushing things out. So to reiterate, the 64-byte limit on Weibo is not the same as the one shown in Xia's GFW paper, and it is not a smoking gun for the two using the same censorship systems. However, it is a REALLY strange coincidence for the two to arbitrarily do truncating at that same specified limit, and I'd love to hear theories about what's going on here and whether the two are at all connected in some other way. /update]
One of the most notable things in the paper which I donât think Iâve seen reported before is that Xia reveals that there is a 64-byte limit on GFWâs string matching rules. Basically, a GFW censor might enter a very specific long string into their system to trigger a drop in the userâs connection to a server. However, if that string is longer than 64-bytes (each Chinese character takes up between two and three bytes depending on the character encoding), the system doesn't bother checking to see if all the characters are included, and instead only checks to see if the first 64-bytes are included. As Xia concludes, which I concur with, âThis is a [sic] supporting evidence that GFW devices are custom made hardware, built to run string matching highly efïŹciently.â
Having read this, I tested Weibo to see if it similarly truncated long searches, and loâ and behold, it did. I had always assumed that it didnât matter how long a search string you entered in Weibo was (via the URL that is; the actual input box on the website is restricted to 40 characters) and that so long as it contained a sensitive word, it would be blocked. However, when tested, Xiaâs conclusion based on his Wikipedia experience held true on Weibo. A search for âaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaćœć瀟â is blocked (ćœć瀟 being âBloombergâ) while adding one more âaâ renders it unblocked. Perhaps itâs merely a coincidence, but if we count bytes for the former search request, s.weibo.com/weibo/aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa%E5%BD%AD%E5%8D%9A%E7%A4%BE is exactly 64 bytes long. Thus, adding another character would make the whole string 65 bytes long, and thus if the system was only looking at the first 64 bytes, it would no longer see the sensitive phrase ćœć瀟 in full.
[Jan 11: This paragraph should now be read with the update above, which rejects these conclusions. - JNG] Essentially, this appears to indicate that the same hardware/software solution that the GFW is using to restrict Wikipedia and other foreign websites is being used by Sina on Weibo--Weibo just happens to serve up a custom blocked keyword page as opposed to dropping a user's connection as GFW does after detecting a sensitive keyword. One could argue perhaps that it is GFW censors who are in fact pulling the strings on what individual keywords are blocked from searching on Sina Weibo and other content-providerâs sites, but I think thatâs fairly implausible considering what we know about the censorship teams in place within Chinese Internet companies. More than likely, we can conclude that the similarity in the way censorship is restricted to checking the first 64-bytes of a URL on both external sites like Wikipedia via the GFW as well as domestic sites like Weibo means that Sina just happens to be using the same censorship platform as the GFW censors. Really interesting, especially when combined with King, Pan, and Roberts recent revelation of how Chinese hosting companies and online forum companies give start-up content providers advice on how to setup their censorship procedures (section 2.1).
Bumbling censors?
I guess Iâll conclude with this final takeaway from Xiaâs paper: âThe revealed rulebook demonstrates that the GFW operation is haphazard and ill-maintained. The GFW ïŹltering rules are like a cesspool.â Indeed, at times the paper becomes a hilarious list of the ineptness or sloppiness of the censors. Pages get blocked on one section of Wikipedia but accidentally left alone on the main site, etc. It sort of reminds me of how Knockel/Crandall/ The Citizen Labâs China Chats project uncovered instances of the censors failing to encrypt censorship lists correctly or deleting lists by mistake and then having to quickly fix things. Itâs a reminder that as frustratingâand for the most part, effectiveâas The Great Firewall appears to be, it still is manned by humans, who are more than fallible. Thus, as Xia writes, ââGFW does not operate at 100%. Certain level of randomness does exist.â
I look forward to the next update and highly encourage folks to follow Xia on Twitter at @summeragony in order to get the next release.
*Or her; according to Xia, due to the sensitivity of his/her position, Xia isnât willing to divulge his/her affiliation and the name is likely a pseudonym. ^
**Sort of; itâs a little more complicated than that but Iâm generalizing here. I further save myself time by pre-screening the sentence and removing any already known sensitive keywords, that way I donât waste time. ^
Looking back on 2013: Five Blocked on Weibo posts I particularly liked from last year
2013 has personally been an incredibly fun year. I finished grad school, my book was published, and I started working for this neat research lab. Chinese Weibo users though, especially prominent ones, had a particularly rougher time, with increased harassment and censorship by authorities inducing an unfortunate chill on discussion of sensitive topics on the site. Here's hoping the next year brings a relaxation of such policies: I couldn't be happier if I had nothing to write about on this blog.
So before we move on to 2014, a look back at five Blocked on Weibo keywords and posts that I particularly enjoyed uncovering and writing about in the past year:
1) Jan 23:Â ćźȘæłæłéą (constitutional court) is blocked during the Southern Weekend censorship controversy.
2)Â Mar 9: Weibo censors delete post of masked Mao portrait criticizing Beijing air pollution.
3) Jun 4: âThe Flower of Freedomâ (èȘç±è±) is a Cantonese song written by Hong Kong lyricist Thomas Chow to commemorate the victims of the 1989 Tienanmen crackdown.
4) Oct 4:Â æé» + è (smash the black + Bo Xilai's surname) refers to Bo's controversial and heavy-handed campaign to rid the city of Chongqing of organized crime.
5) And lastly, my final post of 2013: "I Have No Enemies" (ææČææ”äșș) is a speech written by jailed dissident and Nobel Peace Prize winner Liu Xiaobo. The speech is part-manifesto, part-reflection on Liuâs past, and part-love letter to his wife. Here's a short passage featuring the latter:
Throughout all these years that I have lived without freedom, our love was full of bitterness imposed by outside circumstances, but as I savor its aftertaste, it remains boundless. I am serving my sentence in a tangible prison, while you wait in the intangible prison of the heart. Your love is the sunlight that leaps over high walls and penetrates the iron bars of my prison window, stroking every inch of my skin, warming every cell of my body, allowing me to always keep peace, openness, and brightness in my heart, and filling every minute of my time in prison with meaning. My love for you, on the other hand, is so full of remorse and regret that it at times makes me stagger under its weight. I am an insensate stone in the wilderness, whipped by fierce wind and torrential rain, so cold that no one dares touch me. But my love is solid and sharp, capable of piercing through any obstacle. Even if I were crushed into powder, I would still use my ashes to embrace you.
Why it is blocked: "I Have No Enemies: My Final Statement" was a prepared speech Liu read to the court during his trial. However, after 14 minutes, the judge cut him off, saying Liu had used up his allotted time. The full speech was published and widely circulated online in Chinese in January 2010 and gained even more prominence when it was read aloud in English by actress Liv Ullmann during the 2010 Nobel Peace Prize Ceremony (video: part 1 | part 2).
Though Liu decries the supposed crimes for which he has committed and maintains his innocence, it is not an angry rant. Liu primarily takes on a martyr's role: sticking to his ideals, accepting his fate as a victim, thanking his prosecutors and judges for their decency during the trial, and noting with optimism that change is on the horizon. Liu then addresses his wife, Liu Xia:
Throughout all these years that I have lived without freedom, our love was full of bitterness imposed by outside circumstances, but as I savor its aftertaste, it remains boundless. I am serving my sentence in a tangible prison, while you wait in the intangible prison of the heart. Your love is the sunlight that leaps over high walls and penetrates the iron bars of my prison window, stroking every inch of my skin, warming every cell of my body, allowing me to always keep peace, openness, and brightness in my heart, and filling every minute of my time in prison with meaning. My love for you, on the other hand, is so full of remorse and regret that it at times makes me stagger under its weight. I am an insensate stone in the wilderness, whipped by fierce wind and torrential rain, so cold that no one dares touch me. But my love is solid and sharp, capable of piercing through any obstacle. Even if I were crushed into powder, I would still use my ashes to embrace you.
The speech is part-love letter, part-reflection on Liu's past, part-manifesto. I'll cite a few notable passages, but it should be read in full (HRIC translation | David Kelly translation):
When I think about it, my most dramatic experiences after June Fourth have been, surprisingly, associated with courts: My two opportunities to address the public have both been provided by trial sessions at the Beijing Municipal Intermediate People's Court, once in January 1991, and again today. Although the crimes I have been charged with on the two occasions are different in name, their real substance is basically the same - both are speech crimes. [...]
But I still want to say to this regime, which is depriving me of my freedom, that I stand by the convictions I expressed in my "June Second Hunger Strike Declaration" twenty years ago â I have no enemies and no hatred. None of the police who monitored, arrested, and interrogated me, none of the prosecutors who indicted me, and none of the judges who judged me are my enemies. Although there is no way I can accept your monitoring, arrests, indictments, and verdicts, I respect your professions and your integrity, including those of the two prosecutors, Zhang Rongge and Pan Xueqing, who are now bringing charges against me on behalf of the prosecution. During interrogation on December 3, I could sense your respect and your good faith.
Hatred can rot away at a person's intelligence and conscience. Enemy mentality will poison the spirit of a nation, incite cruel mortal struggles, destroy a society's tolerance and humanity, and hinder a nation's progress toward freedom and democracy. That is why I hope to be able to transcend my personal experiences as i look upon our nation's development and social change, to counter the regime's hostility with utmost goodwill, and to dispel hatred with love. [...]
I hope that I will be the last victim of China's endless literary inquisitions and that from now on no one will be incriminated because of speech. Freedom of expression is the foundation of human rights, the source of humanity, and the mother of truth. To strangle freedom of speech is to trample on human rights, stifle humanity, and suppress truth. In order to exercise the right to freedom of speech conferred by the Constitution, one should fulfill the social responsibility of a Chinese citizen. There is nothing criminal in anything I have done. [But] if charges are brought against me because of this, I have no complaints. Thank you, everyone.
Anya is live and ready to show you everything. Watch her strip, dance, and perform exclusive shows just for you. Interact in real-time and make your fantasies come true.
â Live Streamingâ Interactive Chatâ Private Showsâ HD Quality
Anya is LIVE right now
FREE
Free to watch âą No registration required âą HD streaming
Infographic showing Weibo censorship being linked to offline events in Bo Xilai scandal
In the infographic below, we have collected data from a number of sources, including GreatFire.org, China Digital Times, Blocked on Weibo, and Twitter users to chart the moments when Boâs name became blocked or unblocked on Weibo. The speculation is that the authorities blocked his name when online conversations got too unpredictable to control and unblocked it when they sought to give netizens the space to criticize Bo. We have lined up those moments with what was taking place offline at the same time, presenting a connection between how real-life political turmoil was often reflected in changes in censorship online. Click to launch the interactive infographic.
Concept, Research, and Authorship: JASON Q. NG
Design: JANE GOWAN and ANDREW HILTS
Development: ANDREW HILTS
Update: The Chinese keywords on messaging app LINEâs âbad wordsâ list and why they are âbadâ
Last week, the research lab I pitch in at published the first in a series of posts investigating censorship and privacy concerns in three chat applications: WeChat, LINE, and KakaoTalk. These instant messaging programs, which often replace text messages on smartphones, are expanding rapidly across the world. While WeChat has garnered most of the foreign press, LINE, a Japanese subsidiary of the Korean Internet giant Naver, is no pushover: it has over 200 million registered users, generated $130 million in revenue last year, and is poised for a $10 billion market cap value when it goes public next year.
I've already written a number of blog posts translating and describing some of the 150 words that were initially revealed to be on LINE's "bad words" list. This list, uncovered by Twitter users @hirakujira, was thought to be a precursor to future censorship by the LINE application, but The Citizen Lab's recent reports uncovered a second set of 370 keywords which do trigger censorship--but only for users who have registered with a Chinese phone number. Thus, LINE users in China would receive error messages when sending messages that contain any of these keywords and asterisked-out text when receiving them.Â
In addition to the series of 21 blog posts I did on the first chunk of the original list of uncovered "bad words" in LINE, I have translated the remainder of the 150 keywords on the original list as well as translated the majority of the 370 keywords on the recently decrypted list in the following spreadsheets:
Translation of Line "bad words" list extracted by @hirakujira and confirmed by Citizen Lab (150 words)
Many of the keywords are those that you would expect to find on any list of censored Chinese topics: Tiananmen/June 4, family wealth and corruption by top political officials, supposed infighting within the CCP, and even a whole section of keywords devoted to Bo Xilai. However, others referred to incredibly obscure events that garnered little to no attention from Western or mainstream Chinese media--sometimes the only references to these keywords were in lightly-trafficked message boards and Falun Gong-supported media (which is known for being critical of the Chinese Communist Party). Of course, that might speak to the effective job authorities have done censoring and curtailing the spread of such stories, but it could also be a reflection of a decision by whoever developed these lists to take preemptive measures to prevent the spread of rumors or slander.
Unfortunately (if you're a government censor), the creation of censorship lists containing rumors that you hope to contain has backfired on them with the revelation of these keywords. Instead of hiding these events forever via censorship or even letting them get forgotten in the sands of time as they likely would have on their own, the publication of the secret keywords gives prominence to previously heretofore overlooked incidents, creating a sort of Streisand Effect. In many ways, the revelation of this censorship keyword list (not the first time this sort of thing has happened: see QQ in 2004 or TOM-Skype in 2013) is a cautionary tale for those responsible for implementing censorship into software or websites.
Some of these more surprising and atypical words on these lists include:
æ”æ±çŸćć„ (Zhejiang's receipt-signing Brother): As mentioned in this post, Zhejiang's Vice minister of propaganda was accused in by netizens of charging over 54 million yuan in expenses to his public office and illegally embezzling hundreds of millions in other corrupt activities. Netizens posted images of his receipts, which contained his signature, thus meriting him the name of "Zhejiang's receipt-signing-Brother."  However, it is unclear whether or not this is merely a fabricated rumor or contains a kernel of truth since no reliable sources have corroborated the few unofficial user forums that mention this supposed scandal.
ææșæć€«äșș (Li Yuanchao's wife): Li is the Vice President of China and his wife, Gao Jianjin, is a music professor; I wasn't able to find anything negative about her, and in fact there was a very flattering profile of her in Hong Kong media, but the fact that her name appears on this list alongside corrupt officials and others embroiled in scandal makes me curious.
Like most companies who hope to successfully enter the Chinese market, the Japanese company LINE has partnered with a local company, Qihoo 360--best known for its controversial business practices against competitors like Baidu, Microsoft, and Tencent. Of course, this raises more questions: who initiated the implementation of the censorship "features" into LINE? Who is responsible for updating the keyword lists? How much say does LINE have over what gets added to these lists? It's unlikely LINE or Qihoo would be able to reveal such answers without suffering repercussions from authorities in China, and based on LINE's response to these censorship revelations, they appear to have justified the censorship of their Chinese users as a necessary step. I guess it will be up to other companies other than LINE to either reveal more about how the censorship system works for foreign content providers or to challenge the system.
Hmm, I have an idea: who wants to start a chat app company with me?