Censored word lists are 'proprietary assets' for Chinese big tech

A Q&A with a former Weibo censor on how many censors ByteDance hires, where the new 'red lines' lie and what's different under Xi Jinping.

A view from outside ByteDance's headquarters in Beijing.
A view from outside ByteDance's headquarters in Beijing.
Emmanuel Wong / Contributor via Getty Images

Each year, around the anniversary of the 1989 Tiananmen pro-democracy protests and ensuing crackdown, censorship tightens in China. This year, web users found it to be business as usual. Social media users and even video gamers discovered they were not able to change profile images. Internet companies prevented users from sharing content "due to maintenance." The candle emoji disappeared from WeChat's default emoji collection. Weibo users reported account deletions or suspensions after they shared images of candles, even when they didn't mention a specific event or cause.

The moves felt familiar, but censorship has been tightening nevertheless. As Protocol | China has reported, several major hate campaigns led by ultra-nationalist influencers, or "Red Vs," have triggered widespread censorship in recent months. In the weeks leading up to the June 4 anniversary this year, analysts who track censorship in China have observed an uptick in censorship activities that seemed rather random, as well as an elevated level of punishment for speech infractions. Last weekend, Chinese web users reported that "lie flat," a popular term referring to the retreat from the rat race in protest against cutthroat competition, had been censored.

Protocol spoke about these evolving censorship measures with Liu Lipeng, who worked as an internet censor and a content quality manager at several Chinese tech companies, including Weibo, for nearly 10 years. Liu is currently an editor at China Digital Times, a U.S.-based publication tracking censorship in China. Liu has recently made public over 1,000 internal memos that he saved while working on Weibo's content moderation team, as well as government directives and lists of censored content he received while working at, a Beijing-based video streaming company.

The below transcript has been lightly edited for clarity and sequencing.

Protocol: Can social media platforms achieve censorship largely through technology, or is it still a labor-intensive undertaking?

Liu: Human censors are still a critical part of it. One reason why ByteDance has become so successful is that it hires the most censors among all social media companies. Though they boast their AI capabilities, they spend the most on manpower moderating content; they have at least 10,000 content moderators in Tianjin alone. I think Weibo's content moderation apparatus is just 10% of ByteDance. When I worked for Weibo, they only hired about 200 censors.

Why aren't other tech companies hiring more censors, then?

It's costly. If the profitability of your product isn't there yet, you won't be able to expand your censorship staff drastically. To some degree, a Chinese tech company's censorship mechanism determines the kind of social media product they can offer. ByteDance is able to achieve a host of features in their products because of the size of their content moderation team.

Are companies outsourcing censorship to third-party companies?

Startups do that, but big tech companies have in-house censorship teams because they wouldn't risk leaking their own data. The bottom line is that every user-generated content platform needs to be closely monitored, because if you don't do that, you'll be forced to close shop. The companies' censorship mechanisms now have to be inspected. Starting in 2018, social media companies have had to carry out "security evaluations," which have to be reviewed and approved by authorities.

How valuable is the censored list to each company?

It's their proprietary asset. Why? No one will hand it to you. You can't communicate openly about what needs to be censored. Authorities definitely won't give you a specific list. So you have to come up with your own list. And if you do it well, that will give you a leg up in the competition.

Are you saying the companies often have to guess what's sensitive?

It's a mix of explicit directives from censorship authorities and self-initiation. For example, Douban recently suspended groups that are related to "lie flat" and the term is censored on other platforms as well. But in this case, they probably were just preemptively deleting content, instead of receiving any explicit directive to do so, because the degree to which the term is gutted is different across platforms.

Who are the censorship authorities, exactly?

There are many at the central and local levels. It's a shifting list, and infighting often occurs among them over jurisdictions. The main one is the Cyberspace Administration of China, which has various local branches. Then the public security organs have their internet police and security apparatus. Each local government has set up an Internet Culture Management Office. The propaganda organs can also direct companies to censor content.

These agencies can distribute so-called "harmful samples" that contain text, images, videos or links that need to be censored. Many of the banned words are distilled from those "harmful samples."

Why did you decide to keep the internal memos from Weibo? And what made you come forward and release them?

Before I joined Weibo, I had hoped to work for a social media company and provide value to users through my work. But I couldn't pursue my professional goal, so I left. I felt it was my responsibility to preserve history. Every day, things disappeared from the internet. But I got to keep the photo negatives, which I believe have their historical value. It's personally gratifying to publish those records and now work against censorship. In some way, I am making up for my past engagement in censorship.

Many Chinese web users now feel anything they post online can trigger censorship. Is there still a red line?

The censors' strategy is to make you feel that the red line no longer exists, scaring you into complete self-censorship. It's always a cat-and-mouse game. Once censors realize users have tested a red line, they move it. The red line has become a moving target.

How has censorship evolved over the years?

It's tightened over time, of course. From [censorship of] Hong Kong and Xinjiang to COVID-19, the space for discussion is shrinking by the day. Political discussions and social issues have always been sensitive, but the scale of censorship is far bigger these days. Before Xi Jinping came to power, the focus of censorship was on collective action — whether any news or discussion could stir public outrage and lead to potential protests. These are still closely monitored, but what feels like a bigger target today is anything that doesn't align with mainstream political ideology. Anyone who's deemed unpatriotic, disrespectful of state leaders, state-designated heroes or whose politics are considered to deviate from so-called "socialist core values" are damned. And then, of course, you can't joke about Xi. Before, corporate censors also monitored mockery of leaders, but the intensity of censorship then had nothing on how they treat Xi-related comments today.


Judge Zia Faruqui is trying to teach you crypto, one ‘SNL’ reference at a time

His decisions on major cryptocurrency cases have quoted "The Big Lebowski," "SNL," and "Dr. Strangelove." That’s because he wants you — yes, you — to read them.

The ways Zia Faruqui (right) has weighed on cases that have come before him can give lawyers clues as to what legal frameworks will pass muster.

Photo: Carolyn Van Houten/The Washington Post via Getty Images

“Cryptocurrency and related software analytics tools are ‘The wave of the future, Dude. One hundred percent electronic.’”

That’s not a quote from "The Big Lebowski" — at least, not directly. It’s a quote from a Washington, D.C., district court memorandum opinion on the role cryptocurrency analytics tools can play in government investigations. The author is Magistrate Judge Zia Faruqui.

Keep ReadingShow less
Veronica Irwin

Veronica Irwin (@vronirwin) is a San Francisco-based reporter at Protocol covering fintech. Previously she was at the San Francisco Examiner, covering tech from a hyper-local angle. Before that, her byline was featured in SF Weekly, The Nation, Techworker, Ms. Magazine and The Frisc.

The financial technology transformation is driving competition, creating consumer choice, and shaping the future of finance. Hear from seven fintech leaders who are reshaping the future of finance, and join the inaugural Financial Technology Association Fintech Summit to learn more.

Keep ReadingShow less
The Financial Technology Association (FTA) represents industry leaders shaping the future of finance. We champion the power of technology-centered financial services and advocate for the modernization of financial regulation to support inclusion and responsible innovation.

AWS CEO: The cloud isn’t just about technology

As AWS preps for its annual re:Invent conference, Adam Selipsky talks product strategy, support for hybrid environments, and the value of the cloud in uncertain economic times.

Photo: Noah Berger/Getty Images for Amazon Web Services

AWS is gearing up for re:Invent, its annual cloud computing conference where announcements this year are expected to focus on its end-to-end data strategy and delivering new industry-specific services.

It will be the second re:Invent with CEO Adam Selipsky as leader of the industry’s largest cloud provider after his return last year to AWS from data visualization company Tableau Software.

Keep ReadingShow less
Donna Goodison

Donna Goodison (@dgoodison) is Protocol's senior reporter focusing on enterprise infrastructure technology, from the 'Big 3' cloud computing providers to data centers. She previously covered the public cloud at CRN after 15 years as a business reporter for the Boston Herald. Based in Massachusetts, she also has worked as a Boston Globe freelancer, business reporter at the Boston Business Journal and real estate reporter at Banker & Tradesman after toiling at weekly newspapers.

Image: Protocol

We launched Protocol in February 2020 to cover the evolving power center of tech. It is with deep sadness that just under three years later, we are winding down the publication.

As of today, we will not publish any more stories. All of our newsletters, apart from our flagship, Source Code, will no longer be sent. Source Code will be published and sent for the next few weeks, but it will also close down in December.

Keep ReadingShow less
Bennett Richardson

Bennett Richardson ( @bennettrich) is the president of Protocol. Prior to joining Protocol in 2019, Bennett was executive director of global strategic partnerships at POLITICO, where he led strategic growth efforts including POLITICO's European expansion in Brussels and POLITICO's creative agency POLITICO Focus during his six years with the company. Prior to POLITICO, Bennett was co-founder and CMO of Hinge, the mobile dating company recently acquired by Match Group. Bennett began his career in digital and social brand marketing working with major brands across tech, energy, and health care at leading marketing and communications agencies including Edelman and GMMB. Bennett is originally from Portland, Maine, and received his bachelor's degree from Colgate University.


Why large enterprises struggle to find suitable platforms for MLops

As companies expand their use of AI beyond running just a few machine learning models, and as larger enterprises go from deploying hundreds of models to thousands and even millions of models, ML practitioners say that they have yet to find what they need from prepackaged MLops systems.

As companies expand their use of AI beyond running just a few machine learning models, ML practitioners say that they have yet to find what they need from prepackaged MLops systems.

Photo: artpartner-images via Getty Images

On any given day, Lily AI runs hundreds of machine learning models using computer vision and natural language processing that are customized for its retail and ecommerce clients to make website product recommendations, forecast demand, and plan merchandising. But this spring when the company was in the market for a machine learning operations platform to manage its expanding model roster, it wasn’t easy to find a suitable off-the-shelf system that could handle such a large number of models in deployment while also meeting other criteria.

Some MLops platforms are not well-suited for maintaining even more than 10 machine learning models when it comes to keeping track of data, navigating their user interfaces, or reporting capabilities, Matthew Nokleby, machine learning manager for Lily AI’s product intelligence team, told Protocol earlier this year. “The duct tape starts to show,” he said.

Keep ReadingShow less
Kate Kaye

Kate Kaye is an award-winning multimedia reporter digging deep and telling print, digital and audio stories. She covers AI and data for Protocol. Her reporting on AI and tech ethics issues has been published in OneZero, Fast Company, MIT Technology Review, CityLab, Ad Age and Digiday and heard on NPR. Kate is the creator of and is the author of "Campaign '08: A Turning Point for Digital Media," a book about how the 2008 presidential campaigns used digital media and data.

Latest Stories