China harvests masses of data on Western targets, documents show
China is turning a major part of its internal Internet data surveillance network outward, mining Western social media, including Facebook and Twitter, to equip its government agencies, military and police with information on foreign targets, according to a Washington Post review of hundreds of Chinese bidding documents, contracts and company filings.
China maintains a countrywide network of government data surveillance services - called public opinion analysis software - that were developed over the past decade and are used domestically to warn officials of politically sensitive information online.
The software primarily targets China’s domestic Internet users and media, but a Washington Post review of bidding documents and contracts for over 300 Chinese government projects since the beginning of 2020 include orders for software designed to collect data on foreign targets from sources such as Twitter, Facebook and other Western social media.
The documents, publicly accessible through domestic government bidding platforms, also show that agencies including state media, propaganda departments, police, military and cyber regulators are purchasing new or more sophisticated systems to gather data.
These include a $320,000 Chinese state media software program that mines Twitter and Facebook to create a database of foreign journalists and academics; a $216,000 Beijing police intelligence program that analyses Western chatter on Hong Kong and Taiwan; and a Xinjiang cybercenter cataloguing Uyghur language content abroad.
“Now we can better understand the underground network of anti-China personnel,” said a Beijing-based analyst who works for a unit reporting to China’s Central Propaganda Department. The person, who spoke on the condition of anonymity to discuss their work, said they were once tasked with producing a data report on how negative content relating to Beijing’s senior leadership is spread on Twitter, including profiles of individual academics, politicians and journalists.
These surveillance dragnets are part of a wider drive by Beijing to refine its foreign propaganda efforts through big data and artificial intelligence.
They also form a network of warning systems designed to sound real-time alarms for trends that undermine Beijing’s interests.
“They are now reorienting part of that effort outward, and I think that’s frankly terrifying, looking at the sheer numbers and sheer scale that this has taken inside China,” said Mareike Ohlberg, a senior fellow at the German Marshall Fund who has conducted extensive research on China’s domestic public opinion network.
“It really shows that they now feel it’s their responsibility to defend China overseas and fight the public opinion war overseas,” she said.
Some of the Chinese government’s budgeting includes buying and maintaining foreign social media accounts on behalf of police and propaganda departments. Yet others describe using the targeted analysis to refine Beijing’s state media coverage abroad.
The purchases range in size from small, automated programs to projects costing hundreds of thousands of dollars that are staffed 24 hours a day by teams including English speakers and foreign policy specialists.
The documents describe highly customizable programs that can collect real-time social media data from individual social media users. Some describe tracking broad trends on issues including U.S. elections.
The Post was not able to review data collected by the systems but spoke to four people based in Beijing who are directly involved in government public opinion analysis and described separate software systems that automatically collect and store Facebook and Twitter data in real time on domestic Chinese servers for analysis.
Twitter and Facebook both ban automated collection of data on their services without prior authorization. Twitter’s policy also expressly bars developers from gathering data used to infer a user’s political affiliation or ethnic and racial origin.
“Our API provides real-time access to public data and tweets only, not private information. We prohibit use of our API for surveillance purposes, as per our developer policy and terms,” said Katie Rosborough, a Twitter spokesperson, referring to the company’s Application Programming Interface (API), which allows developers to retrieve public data from the platform among other functions.
Facebook did not respond to requests for comment about whether it is aware of the monitoring or whether several companies, universities and state media firms listed as supplying the software were authorized to collect data on its platform.
China’s Foreign Ministry did not respond to a request for comment.
‘Public opinion guidance’
China’s systems for analyzing domestic public opinion online are a powerful but largely unseen pillar of President Xi Jinping’s program to modernize China’s propaganda apparatus and maintain control over the Internet.
The vast data collection and monitoring efforts give officials insight into public opinion, a challenge in a country that does not hold public elections or permit independent media.
The services also provide increasingly technical surveillance for China’s censorship apparatus. And most systems include alarm functions designed to alert officials and police to negative content in real time.
These operations are an important function of what Beijing calls “public opinion guidance work” - a policy of molding public sentiment in favor of the government through targeted propaganda and censorship.
The phrase first came to prominence in policymaking after the 1989 Tiananmen Square pro-democracy demonstrations, when officials began exploring new ways to preempt popular challenges to the Communist Party’s power, and has since become integral to the underlying architecture of China’s Internet, where users are linked by real name ID, and Internet services are required by law to maintain an internal censorship apparatus.
The exact scope of China’s government public opinion monitoring industry is unclear, but there have been some indications about its size in Chinese state media. In 2014, the state-backed newspaper China Daily said more than 2 million people were working as public opinion analysts. In 2018, the People’s Daily, another official organ, said the government’s online opinion analysis industry was worth “tens of billions of yuan,” equivalent to billions of dollars, and was growing at a rate of 50% a year.
That surveillance network system is expanding to include foreign social media at a time when global perceptions of Beijing are at their lowest in recent history.
A Pew Research survey released in June showed that perceptions of China among 17 advanced economies had dipped to near historic lows for a second year in a row in the aftermath of the U.S. trade war, the Xinjiang human rights crisis, Hong Kong and the coronavirus pandemic.
In May this year, Xi called on senior officials to portray a more “trustworthy, lovable and reliable” image of China abroad, calling for the “effective development of international public opinion guidance.”
His comments reflect Beijing’s growing anxieties over how to control China’s image abroad.
“On the back of the Sino-US trade talks and the Hong Kong rioting incident, it’s becoming clearer day by day that the public opinion news war is arduous and necessary,” China Daily said in a July 2020 bidding document for a $300,000 “foreign personnel analysis platform.”
The invitation to tender lays out specifications for a program that mines Twitter, Facebook and YouTube for data on “well known Western media journalists” and other “key personnel from political, business and media circles.”
“We are competing with the US and Western media, the battle for the right to speak has begun,” it said.
The software should run 24 hours a day, according to the specifications, and map the relationships between target personnel and uncover “factions” between personnel, measuring their “China tendencies” and building an alarm system that automatically flags “false statements and reports on China.”
Warning systems like the one outlined in the China Daily document are described in over 90% of tenders that list technical specifications, The Post’s review of the documents show.
Two people who work as analysts in public opinion analysis units contracted by government agencies in Beijing told The Post that they receive automated alarms via SMS, email and on dedicated computer monitors when “sensitive” content was detected. Both of the people spoke on the condition of anonymity because they were not authorized to speak to foreign media.
“Having responsibility for [the monitoring] is a lot of pressure,” said one of the people. “If we do our work poorly, there are severe repercussions.”
Highly sensitive viral trends online are reported to a 24-hour hotline maintained by the Cybersecurity Administration of China (CAC), the body that oversees the country’s censorship apparatus, the person said of their unit.
The person added that most of the alarms were related to domestic social media but that foreign social media had also been included in the units’ monitoring since the middle of 2019.
The person’s account is supported by four bidding documents for unrelated systems that mention direct hotlines to the CAC.
“In case of major public opinion, directly contact the staff on duty of the CAC by telephone to ensure that notifications are in place through various communication tools,” said one December 2020 tender for a $236,000 system purchased by the municipal propaganda department in eastern China’s Fuzhou city for monitoring Facebook and Twitter alongside domestic social media.
It specifies that reports to the CAC should include the details of individual social media users.
State media-led data mining
Suppliers of the systems vary. The China Daily awarded its contract to Beijing’s Communications University, one of a half dozen Chinese universities that have launched specialized departments to develop public opinion analysis technology.
However, some of the most prolific public opinion monitoring services are provided to police and government agencies by state media themselves.
The documents provide insight into the scope of foreign social media data collection done by China’s major state media outlets, which maintain offices and servers abroad, and their key role in providing Beijing with publicity guidance based on increasingly sophisticated data mining analysis.
The growing clout of Beijing’s propaganda efforts abroad, spearheaded by state media, has triggered alarms in Washington.
In 2020, the State Department reclassified the U.S.-based operations of China’s top state media outlets as foreign missions, increasing reporting requirements and restricting their visa allocations, angering Beijing.
The People’s Daily Online, a unit of the state newspaper the People’s Daily, which provides one of the country’s largest contract public opinion analysis services, won dozens of projects that include overseas social media data collection services for police, judicial authorities, Communist Party organizations and other clients.
The unit, which recorded $330 million in operating income in 2020, up 50% from 2018, says it serves over 200 government agencies, although it is not clear how many request foreign social media data.
In one tender won by the People’s Daily Online, the Beijing Police Intelligence Command Unit purchased a $30,570 service to trawl foreign social media and produce reports on unspecified “key personnel and organizations,” gathering information on their “basic circumstances, background and relationships.”
It also calls for weekly data reports on Hong Kong, Taiwan and U.S. relations. Issued shortly before the 2020 U.S. presidential election results were ratified on Jan. 6, it also called for “special reports” on “netizens’ main views” related to the election.
“The international balance of power has been profoundly adjusted,” said the request for tenders. “Through the collection of public Internet information we can keep a close eye on the international community, analyze sensitivities and hot spots, and maintain the stability of Chinese society.”
In an April 2020 article, the chief analyst at the People’s Daily Online Public Opinion Data Center, Liao Canliang, laid out the ultimate goal of public opinion analysis.
“The ultimate purpose of analysis and prediction is to guide and intervene in public opinion,” Canliang wrote. “. . . Public data from social network users can be used to analyze the characteristics and preferences of users, and then guide them in a targeted manner.”
In the article, Liao points to Cambridge Analytica’s impact on the 2016 U.S. election as evidence of social media’s ability to mold public opinion.
“The West uses big data to analyze, research and judge public opinion to influence political activities. ... As long as there is a correct grasp on the situation, public opinion can also be guided and interfered with,” he wrote.
People’s Daily subsidiary Global Times, a firebrand newspaper known for its biting coverage of China’s critics, also has a unit gathering foreign social media data for China’s Foreign Ministry, Beijing’s Foreign Affairs Office and other government agencies.
In late 2019, the Global Times Online won a three-year contract worth $531,000 to provide a “China-related foreign media and journalist opinion monitoring system” that monitors overseas social media on behalf of China’s Foreign Ministry and produces comprehensive regular reports, as well as special briefings in “urgent circumstances.”
Documentation accompanying the project says that close to 40% of the Global Times monitoring unit’s staffers are senior Global Times reporters and that the publication maintains large overseas social media monitoring platforms.
A description on the website of the Global Times’s public opinion research center says the group conducts “overseas monitoring and overseas investigation services” and provides “comprehensive response plans” to government and private clients.
Both the People’s Daily and the Global Times were among the outlets designated as foreign missions in the United States.
The increase in China’s monitoring of foreign public opinion on social media coincides with efforts by Beijing to boost its influence on Twitter and other U.S. social media platforms.
In June 2020, Twitter suspended 23,000 accounts that it said were linked to the Chinese Communist Party and covertly spreading propaganda to undermine pro-democracy protests in Hong Kong. This month, Twitter said it removed a further 2,048 accounts linked to Beijing and producing coordinated content undermining accusations of rights abuses in Xinjiang.
Experts say those accounts represent a small fraction of China’s efforts to boost pro-Beijing messaging on foreign social media.
‘Extreme chilling effect’
Just under a third of the public opinion analysis systems reviewed by The Post were procured by Chinese police.
In 14 instances, the analysis systems included a feature requested by the police that would automatically flag “sensitive” content related to Uyghurs and other Chinese ethnic minorities. An additional 12 analysis systems included the police-requested capability of monitoring individual content authors over time.
“It must support information monitoring of overseas social media . . . and provide for targeted collection of designated sites and authors,” said one invitation to tender released by the Fuzhou city police in October that lists coverage of Facebook and Twitter as a requirement.
The monitoring of social media abroad by local police throughout China could be used in investigating Chinese citizens locally and abroad, as well as in flagging trends that stir domestic dissent, experts say.
“The public security monitoring is very much about stability maintenance, tracking people down and finding people’s identity, and when they monitor overseas social media, it’s also often with an eye to monitoring what news could cause trouble at home in China,” said the German Marshall Fund’s Ohlberg.
Companies providing overseas public opinion monitoring to police include a mix of private and state-owned firms, including the People’s Daily Online.
Six police contracts awarded since 2020 stated that the People’s Daily was chosen to conduct monitoring on the basis of its technical ability to gather data abroad.
“It’s the only one in the industry that deploys overseas servers. It is a public opinion service organization that can monitor and collect more than 8,000 overseas media without ‘overturning the wall,’ “ said the Guangdong Police Department in a $26,200 contract offer posted in July 2020, referring to the ability of the People’s Daily unit to collect overseas data outside China’s Great Firewall.
Experts say the increasingly advanced social media surveillance technology available to Chinese police could worsen the targeted harassment of Beijing’s critics.
“The Chinese government is one of the worst offenders when it comes to targeting individuals outside of the country,” said Adrian Shahbaz, the director for technology and democracy at the think tank Freedom House.
“It has an extreme chilling effect on how Chinese citizens outside of China are using social media tools, because they know that back home, their information is very easily monitored by Chinese authorities,” he said.
The China Public Security Bureau did not respond to a request for comment.
A police bureau in southern China’s Nanping city purchased a $42,000 system that “supports collection, discovery, and warning functions for ... Twitter and Facebook social media data according to different classifications and keyword groups, as well as overseas information lists,” according to bidding documents released in July 2020.
Other procurements for public opinion services outline programs purchased by Chinese police and Xinjiang government bodies to track “sensitive” ethnic language content abroad. (China’s mainly-Muslim Uyghurs are concentrated in Xinjiang.)
A $43,000 system purchased by police in central China’s Shangnan county included a “foreign sensitive information” collection system that requested Uyghur and Tibetan staff translators, according to the contracts.
Military procurement documents - less detailed than other types - did not offer much detail on the purpose of the foreign data collection but alluded to vague categories of data including “key personnel.”
One heavily redacted June 2020 contract issued by the People’s Liberation Army described a system that would trawl foreign sites and categorize data on the basis of affiliation, geography and country.
Source Data Technology, the Shanghai-based company that won the contract, says on its website that it uses “advanced big data mining and artificial intelligence analysis technology” to cover more than 90% of social media in the United States, Europe and China’s neighboring countries.