The explosive popularity of short video platforms like Douyin and Kuaishou has brought correspondingly intense scrutiny from China’s censors. New content rules in January were so detailed that they even included a “specific prohibition on wearing apparel with leaders faces on it, then wiggling or folding it to make the faces have strange expressions.” The global success of Douyin’s international sibling service TikTok has fueled concern that the platform might provide a channel for Beijing’s information controls to reach overseas. On Wednesday, The Guardian reported on leaked content guidelines for TikTok which suggest that these suspicions were well founded. From Alex Hern:
The guidelines divide banned material into two categories: some content is marked as a “violation”, which sees it deleted from the site entirely, and can lead to a user being banned from the service. But lesser infringements are marked as “visible to self”, which leaves the content up but limits its distribution through TikTok’s algorithmically-curated feed.
[…] In every case, [guidelines covering China] are placed in a context designed to make the rules seem general purpose, rather than specific exceptions. A ban on criticism of China’s socialist system, for instance, comes under a general ban of “criticism/attack towards policies, social rules of any country, such as constitutional monarchy, monarchy, parliamentary system, separation of powers, socialism system, etc”.
Another ban covers “demonisation or distortion of local or other countries’ history such as May 1998 riots of Indonesia, Cambodian genocide, Tiananmen Square incidents”.
[…] “In TikTok’s early days we took a blunt approach to minimising conflict on the platform, and our moderation guidelines allowed penalties to be given for things like content that promoted conflict, such as between religious sects or ethnic groups, spanning a number of regions around the world,” the company said. “As TikTok began to take off globally last year, we recognised that this was not the correct approach, and began working to empower local teams that have a nuanced understanding of each market. As we’ve grown we’ve implemented this localised approach across everything from product, to team, to policy development.” [Source]
[Updated at 9:15:48 AM PDT on Sep 27, 2019: a follow-up report on Thursday described bans in some countries on “any content that could be seen as positive to gay people or gay rights, down to same-sex couples holding hands, even in countries where homosexuality has never been illegal,” among other region-specific content rules. TikTok again claimed that the guidelines cites “are no longer in use.”]
Questionable content handling is hardly an issue unique to TikTok. Facebook’s announcement this week of a “newsworthiness” exemption from its content policies for posts by politicians has stirred strident criticism. Hern reported earlier this month on leaked guidelines for Apple’s voice assistant Siri which stated that it “should be guarded when dealing with potentially controversial content,” and that “care must be taken here to be neutral” on issues like feminism.
However, the scale of TikTok’s global popularity, the youth of its audience, and its ownership by a company operating under a government notorious for political and moral censorship have all heightened concern still further. In May, Buzzfeed’s Ryan Broderick described TikTok as “China’s most important export right now,” potentially heralding “the arrival of a subtler form of algorithmic influence, with sophisticated Chinese AI controlling what becomes viral content potentially shared among millions of young Americans.” The Washington Post’s Drew Harwell and Tony Romm, reporting earlier this month on the conspicuous absence from TikTok of content related to protests in Hong Kong, described it as “the first Chinese app to truly pierce the global Internet mainstream”:
[R]esearchers have grown worried that the app could also prove to be one of China’s most effective weapons in the global information war, bringing Chinese-style censorship to mainstream U.S. audiences and shaping how they understand real-world events. Compounding researchers’ concerns are TikTok’s limited public comments about the content it removes and its purported independence from censors in Beijing.
TikTok’s parent company ByteDance said in a statement that U.S. user data is stored domestically and that the app’s content and moderation policies in the U.S. are led by a U.S.-based team not influenced by the Chinese government. ByteDance repeatedly declined to make executives available for on-the-record interviews.
In its statement, the company defended TikTok as a place for entertainment, not politics, and said its audience gravitates there for positive and joyful content as a possible explanation for why so few videos relate to sensitive topics such as the protests in Hong Kong.
[…] It’s impossible to know what videos are censored on TikTok: ByteDance’s decisions about the content it surfaces or censors are largely opaque. The company provides no information about the videos it removes for violating its prohibitions against hate speech or extremism, and it does not offer the kinds of tools that would make the platform accessible to outside research. It’s also possible that users in Hong Kong could be self-censoring by not posting politically fraught content onto an app closely scrutinized by Chinese censors.
But popular hashtags used by Hong Kong protester that have spread widely across other social media barely exist on TikTok. [Source]
Bytedance’s appeal to the separation between the Chinese Douyin and international TikTok in its response to the Post highlights its strategy for navigating the extreme wariness that has dogged other Chinese tech firms overseas, most notably Huawei. Assessing the relationship between the two apps in January at What’s On Weibo, Gabi Verberg wrote that “although Tiktok and Douyin have the same functions, layout, and logos, its users in China and overseas are kept completely separate and are not able to interact.”
In an in-depth piece on Bytedance and TikTok in this week’s tech-focused edition of The New Yorker, Jia Tolentino described the platform as “the last sunny corner on the Internet,” which “in five minutes, […] had sandblasted my cognitive matter with twenty TikToks that had the legibility and logic of a narcoleptic dream.” But she also explored the company’s claims about its international platform’s independence, its relationship to Chinese authorities, and the potentially dark applications of the opaque AI technology powering many of its products:
TikTok employees in Los Angeles declined to talk in any detail about their relationship to ByteDance headquarters, in Beijing, and everyone I spoke to emphasized that the U.S. operation was fairly independent. But one former employee, who left the company in 2018, described this as a “total fabrication.” (A ByteDance spokesperson, in response, said that the markets were becoming more independent and that much of that process had happened within the past year.) TikTok’s technology was developed in China, and it is refined in China. Another ex-employee, who had worked in the Shanghai office, said that nearly all product features are shipped out from Shanghai and Beijing, where most of ByteDance’s engineers are based. “At a tech company, where the engineers are is what matters,” the writer and former Facebook product manager Antonio Garcia-Martinez told me. “Everyone else is a puppet paid to lie to you.”
[…] Chinese tech companies are often partly funded by the government, and they openly defer to its requests, turning over user messages and purchase data, for instance. Tencent, which owns WeChat, has a “Follow Our Party” sign on a statue in front of its headquarters. The Wall Street Journal has reported that a ByteDance office in Beijing includes a room for a cybersecurity team of the Chinese police, which the company informs when it “finds criminal content like terrorism or pedophilia” on its apps. […]
[…] Dinesh Raman, an A.I.-alignment researcher in Tokyo, who has studied ByteDance as a consultant for some of its investors, spoke with a mixture of alarm and admiration about the company’s A.I. capabilities. “The system is doing billions of calculations per second,” he said. “It’s data being transmitted at a scale I’ve never seen before.” Raman insisted that TikTok had kept its platform tightly policed in part through its algorithm, which, he said, is able to identify videos with dangerous content. (TikTok’s moderators are trained to apply different standards to every market, the company told me.) He pointed me to the “Gaga Dance” challenge, a meme on Indonesian TikTok that asked users to mirror the poses of cheerful yellow stick figures that floated across the screen. The A.I., he suggested, was training itself in pose estimation, a deep-learning capability with major surveillance implications. […] [Source]
While the wall between TikTok and Douyin may be impassable to most users, The Guardian’s report this week suggests that it is much more porous to censorship pressures. At Quartz in May, David Carroll—a Parsons School of Design professor whose role in uncovering the Cambridge Analytica/Facebook scandal was prominently featured in recent Netflix documentary The Great Hack—investigated whether the same was true for users’ personal data:
I wanted to apply Cambridge Analytica like a litmus test to TikTok. The result illustrated a perplexing new economic and political web of data trade routes across regulatory boundaries.
[…] The company responded to my concerns with an inscrutable combination of yes, no, and maybe. This is the nature of the emerging “splinternet,” where democratic and authoritarian states are disconnecting and fracturing from the global vision of the hyperconnected internet. That liminal feeling is the new uncanny data sovereignty in the age of surveillance capitalism.
Does their answer mean that ByteDance entities in China are now accessing US-based servers and processing the data here? Is this how they firewall TikTok from the Chinese government? Are teams sitting side-by-side in Beijing building both TikTok and Douyin, with one international team processing data on servers abroad beyond the reach of the government, and the domestic team running on servers in China subject to censors and security forces?
[…] The Cambridge Analytica story was just the canary in the coal mine. It woke us up to our reckless attitude about our personal data. We may roll our eyes at the seemingly innocuous, but the risk of regulatory penalties for data-privacy violations is real. My experience with TikTok—a company on probation with the feds—informs our understanding of data trade routes, as China learns how to navigate the data-protection regimes and norms of democratic states. [Source]
At The Financial Times last year, Emily Feng described the Douyin/TikTok division as “a potential road map for how mainland tech companies can go global without falling foul of Beijing’s content restrictions.” TikTok, she noted, is unavailable from Chinese app stores, and requires registration through other platforms that are blocked in China, while users with Chinese SIM cards are unable to use the app even if they clear these other hurdles.
At Coda this week, Isobel Cockerell described efforts to cross the gap in the opposite direction, as diasporic Uyghur activists scour Douyin for insight into the situation in Xinjiang and the fate of its people amid an ongoing security crackdown and mass detention campaign.
[… E]very so often, Erkin comes across a video that reveals something about the realities of China’s mass surveillance crackdown and brainwashing policy in Xinjiang: a video of a propaganda rally, with Uyghurs singing songs praising the Communist Party of China; footage from inside a Uyghur orphanage for children with parents in detention; crowds of Uyghurs chanting in Mandarin rather than their native Uyghur language; a mosque being demolished. Often he’ll find shots of the deserted streets of once-bustling Kashgar, his now empty home city.
[…] Thousands of videos from Xinjiang, filmed by both Uyghur and Han Chinese users, are uploaded every day to Chinese TikTok and other copycat video sharing apps. The sheer volume of videos makes it difficult for the authorities to censor everything. “They’re plugging the gaps,” said Darren Byler, a scholar who has extensively researched Uyghurs and technology. “But it’s done in a piecemeal way. The internet is a big place and it’s hard to police it.” And sometimes, compromising content slips through the net.
[…] China’s firewall — originally designed to keep Chinese people from accessing foreign websites — now appears to be also stopping foreigners from seeing in. “It looks like they’re creating a reverse great firewall, and Douyin is a perfect example. They want to keep TikTok outside and Douyin inside; there’s an intentionality there that has an element of censorship about it,” said James Leibold, associate professor in politics and Asian studies at La Trobe University, Melbourne. Day by day, he said, it’s becoming more difficult to access online content from Xinjiang. The solution, he believes, is to be ever more innovative and methodical.
Once they’ve got around the firewall and accessed TikTok, the international Uyghur activists then have to “teach” the app’s algorithm to show them the videos they want to see. “You have to train it in a certain way,” Yasin said. “You can’t really search, because they cleaned up all the location-based search results. Anything that uses Xinjiang keywords is censored.” [Source]
Others have previously noted “what appear to be precautionary performances of loyalty to the government, [in which] Uighur users of Douyin have recorded themselves singing pro-party songs, pointedly in Mandarin,” and “images of missing people, with a photograph or video of the person posting the clip superimposed over the top. Many of those posting the videos are crying.” Cockerell’s account provides an update on the silencing of many of the latter accounts.
Such videos are exceptional, though some other content, like a recent trend of complaints and satires about soaring pork prices, does venture into somewhat sensitive territory. Meanwhile, as part of a broad push to engage young Chinese through new platforms, official bodies have been adopting the platform for their own purposes. From Masha Borak at Abacus in August:
Douyin still has the silly dance moves, comical skits and stunts. But over the last few years, dozens of state-owned media outlets like People’s Daily and China Daily have created their own accounts. So have government agencies, including police and military outposts, giving them an outlet to directly share their own points of view.
Among short videos of teenage lip-syncers and farmers doing the robot dance, you’re now likely to stumble across statements from Chinese government supporters or speeches from Chinese Ministry of Foreign Affairs spokeswoman Hua Chunying accusing the US of playing a role in the Hong Kong anti-extradition protests.
And as it usually goes with social media, sharing and liking certain types of content only surfaces more of the same. A few swipes may lead not only to videos of officials explaining the latest government directives, but also to Chinese military officials claiming jurisdiction over disputed islands in the South China Sea.
It could even bring up promo videos for the Chinese army and police, such as the one posted by the PLA’s garrison in Hong Kong. Videos of soldiers exercising, running military drills or just generally looking tough are a big part of patriotic Douyin. It’s the part of the social network people turn to for nationalist content, like military videos that play like modern renditions of Rambo movies. [Source]
Bytedance itself, along with many other compatriot tech companies, has been recruited to support the Party and state. Quartz’s Jane Li reported last week on the formation of an “Intelligent Media Research Institute” and AI Media Lab by Party mouthpiece People’s Daily:
The People’s Daily announcement comes after draft rules last week from the country’s internet regulator called on internet firms to tinker with their algorithms to direct people to content about party ideology and away from stuff like celebrity gossip. China has long relied on tech giants to follow its directions on “cleaning up” cyberspace since Chinese president Xi Jinping became the party’s top official in 2012, censoring content deemed politically incorrect or sensitive by the party, as well as obscene or violent material.
Other Chinese tech majors, such as Alibaba, Tencent, JD.com, Bytedance—whose AI driven news app has attracted censure from authorities for spreading “useless information“—Xiaomi, app delivery giant Meituan Dianping and e-commerce site Pinduoduo are also members of the new research institute, according to People’s Daily.
The uneasy relationship between the tech giants and the Party—which at times has seen companies lose billions in market value because of a critical editorial in the People’s Daily—is likely to be an increasing headache for companies. Huawei, for instance, has repeatedly denied that it is state-owned, or has a deep relationship with China’s military, but was nevertheless put on a trade blacklist by the US in May for national security reasons.
Initiatives like this, which appear to outside observers to be further drafting tech firms into China’s efforts to control public opinion, won’t help with the perception in other nations that China’s private firms are too close for comfort to the government. [Source]
In another sign of deepening ties this week, Hangzhou, an important tech hub and home to ecommerce giant Alibaba, announced plans to assign officials to “function as a bridge” (from an Alibaba statement) between the local government and the private sector. Commenting on the news to The Financial Times, “Fraser Howie, author of Red Capitalism, said the move highlighted how ostensibly private companies are […] ‘effectively either state owned enterprises or state overseen enterprises[….] And there seems to be no move to get away from that and indeed more and more effort to make it very clear the private sector are beholden to the Party.'”
Bytedance has attracted increasingly intense oversight as its various audiences have exploded. In January 2018, the company’s news, blog, and entertainment aggregator Jinri Toutiao was accused of “spreading pornographic and vulgar information,” “causing a negative impact on public opinion online,” and operating without the required license for an online news service. Various channels and accounts were subsequently suspended in order, the company said, to “to promote the spirit of the Communist Party congress.”
That April, the company’s Neihan Duanzi joke-sharing platform was shut down completely, prompting performative soul-searching by founder Zhang Yiming over content “that did not accord with core socialist values and was not a good guide for public opinion.” (Many speculated that the app’s downfall was related to its strong community, which had started to spill offline, and may therefore have been perceived as a potential threat.)
After these setbacks, the chastened company drew attention by apparently trying to pre-empt censors’ wishes with a crackdown on subversive Peppa Pig remixes. In October, it joined other platforms in deleting videos by massively popular streamer Yang Kaili, who was detained for five days for showing alleged disrespect to the national anthem.