China

Not Free
9
100
A Obstacles to Access 7 25
B Limits on Content 2 35
C Violations of User Rights 0 40
Last Year's Score & Status
10 100 Not Free
Scores are based on a scale of 0 (least free) to 100 (most free). See the research methodology and report acknowledgements.

header1 Overview

Conditions for internet users in China remained profoundly oppressive and confirmed the country’s status as the world’s worst abuser of internet freedom for the ninth consecutive year. Unprecedented protests were mobilized against the government’s COVID-19 policy during the coverage period. Though authorities responded with swift censorship, protesters successfully pressured the government into withdrawing its draconian zero-COVID policy. Ordinary users continued to face severe legal and extralegal repercussions for activities like sharing news stories, talking about their religious beliefs, or communicating with family members and others overseas. Separately, authorities wielded immense power over the technology industry through new legislation, regulatory investigations, and app-store removals. Authorities also imposed new restrictions on generative artificial-intelligence (AI) tools during the coverage period.

China’s authoritarian regime has become increasingly repressive in recent years. The ruling Chinese Communist Party (CCP) continues to tighten control over all aspects of life and governance, including the state bureaucracy, the media, online speech, religious practice, universities, businesses, and civil society associations. The CCP leader and state president, Xi Jinping, secured a third term as party leader in October 2022, further consolidating personal power to a degree not seen in China for decades. Following a multiyear crackdown on political dissent, independent nongovernmental organizations (NGOs), and human rights defenders, China’s civil society has been largely decimated.

Editor’s Note: Tibet and Hong Kong are not covered in this report. Certain territories that are assessed separately in Freedom House’s Freedom in the World report are excluded from the relevant country reports in Freedom on the Net, as conditions in such territories differ significantly from those in the rest of the country.

header2 Key Developments, June 1, 2022 - May 31, 2023

  • People in China held unprecedented protests over the restrictive zero-COVID policy and other grievances in late 2022. The protests were triggered by a fire in an Urumqi apartment building that killed 10 people in November, amid an ongoing COVID-19 lockdown. Authorities removed social media posts about the protests, ordered platforms to increase their capacity for censorship, and arrested protesters, but ultimately ended the zero-COVID policy (see B2, B3, B8, and C3).
  • Censors removed online criticism of President Xi’s third term and of the economy during the coverage period (see B2).
  • Regulators enforced content controls of large language model–based chatbots. Firms in China were ordered to discontinue access to ChatGPT and to alert authorities of their own plans to develop AI-driven chatbots (see B3).
  • In November 2022, the Cyberspace Administration of China (CAC), a telecommunications regulator, required social media operators to register users with their real names and to roll out new moderation tools, as part of the government's ongoing efforts to control online speech (see B6).
  • Authorities detained prominent human rights lawyers, journalists, activists, and members of religious and ethnic minority groups for their online activities. In April 2023, a court in Shandong Province sentenced prominent human rights lawyer and blogger Xu Zhiyong to 14 years in prison for “subversion of state power” (see C3).
  • In December 2022, the CAC released new rules requiring internet service providers (ISPs) to discontinue service to anyone who had not authenticated their real identity (see C4).

A Obstacles to Access

A1 1.00-6.00 pts0-6 pts
Do infrastructural limitations restrict access to the internet or the speed and quality of internet connections? 5.005 6.006

According to the government’s China Internet Network Information Center (CNNIC), there were 1.07 billion internet users in China—representing 75.6 percent of the population—as of December 2022. That figure represents an increase of 35.5 million since December 2021. Some 99.8 percent of users access the internet via mobile devices.1

Chinese internet users can access high-speed services, though connection speeds are slowed by the country’s blocking and filtering apparatus (see B1). According to Ookla, the median mobile download speed stood at 105.17 megabits per second (Mbps) in June 2023. The median fixed-line broadband download speed was 209.4 Mbps.2 Connection speeds in China are slowed by the government’s extensive censorship apparatus (see B1), which filters all cross-border traffic and makes the loading of content from foreign-hosted websites sluggish.3 Internet speeds vary significantly in different parts of the country. According to government data, the fastest available connections were in Shanghai, and the slowest were in less prosperous and more heavily censored regions, such as Xinjiang.4

Chinese companies have been at the forefront of building and deploying fifth-generation (5G) mobile networks. In November 2022, the state-run Xinhua News Agency reported that China had 2.2 million 5G base stations, while mobile service provider China Mobile said it had 556.8 million 5G subscribers in October 2022.5

A2 1.00-3.00 pts0-3 pts
Is access to the internet prohibitively expensive or beyond the reach of certain segments of the population for geographical, social, or other reasons? 1.001 3.003

Internet access is relatively affordable for the average user, but other digital divides limit access for certain populations.

According to the China Academy of Information and Communication Technology, fixed-line broadband services cost an average of 36.1 yuan ($5.08) in the first three quarters of 2022, while mobile users paid an average of 2.61 yuan ($0.37) per gigabyte (GB).1

The urban-rural digital divide narrowed during the coverage period, according to government figures. Internet penetration in cities was 82.9 percent as of June 2022, compared to 58.8 percent in rural areas.2 Some 362 million people did not have internet access as of that month, 41.2 percent of them in rural areas.3

A3 1.00-6.00 pts0-6 pts
Does the government exercise technical or legal control over internet infrastructure for the purposes of restricting connectivity? 1.001 6.006

The government maintains control over China’s gateways to the global internet, giving authorities the ability to restrict connectivity or access to content hosted on servers outside the country.1 This arrangement is the foundation for the “Great Firewall,” the informal name for the government’s comprehensive internet censorship system. All service providers must subscribe via the gateway operators, which are overseen by the Ministry of Industry and Information Technology (MIIT).

Human rights activists and their families are subject to targeted disconnections. In February 2023, a month ahead of the annual “Two Sessions” meetings of the National People’s Congress (NPC) and the Chinese People’s Political Consultative Conference (CPPCC),2 authorities disabled internet service for Wuhan-based activist Xu Wu and his family.3

The government has cut internet access in response to specific events, though authorities have relied more on other censorship tactics in recent years. The most dramatic example occurred in 2009, when authorities imposed a 10-month internet disruption in Xinjiang—home to 25.9 million people according to the 2020 census—after ethnic violence in the regional capital, Urumqi.4

Network shutdowns are often explained as national security precautions. The cybersecurity law and Article 84 of a 2015 antiterrorism law introduced fines and detentions of up to 15 days for telecommunications firms and ISPs, as well as relevant personnel, who fail to restrict certain forms of content including “shut[ting] down related services” (see B3 and C2).5 Under a cybersecurity rule implemented in June 2020,6 a government agency must conduct a national security review of the purchases of network products and services made by “critical information infrastructure operators.”7

A4 1.00-6.00 pts0-6 pts
Are there legal, regulatory, or economic obstacles that restrict the diversity of service providers? 0.000 6.006

Score Change: The score declined from 1 to 0 to reflect the state monopoly over the telecommunications sector and the extreme obstacles faced by private companies seeking to enter the market.

State-owned China Mobile, China Telecom, and China Unicom dominate the mobile market,1 though the government has occasionally authorized new players to enter the market.2 China Mobile dominates the mobile and fixed-line broadband markets, with 975 million and 272 million subscribers, respectively, as of December 2022.3 China Telecom reported 391 million mobile subscribers and 181 million fixed-line broadband subscribers as of December 2022,4 while China Unicom reported 320 million mobile subscribers and 100 million fixed-line broadband subscribers.5

Authorities exercise tight control over cybercafés and other public access points, which are licensed by the Ministry of Culture in cooperation with other state entities.6 Video gamers continue to gather at cybercafés regularly. Some 125,000 cybercafés were active as of February 2021.7

A5 1.00-4.00 pts0-4 pts
Do national regulatory bodies that oversee service providers and digital technology fail to operate in a free, fair, and independent manner? 0.000 4.004

Several government agencies and CCP bodies are responsible for internet regulation at the local and national levels, but the system has been consolidated during Xi Jinping’s tenure.

The CAC1 and the CCP’s Central Cyberspace Affairs Commission (CCAC) oversee the telecommunications sector and regulate internet content.2 The 2017 cybersecurity law identifies the CAC as the principle agency responsible for implementing many of its provisions.3 The CAC reports to the CCAC, which is headed directly by Xi.4 Since 2014, the CCAC has also overseen the CNNIC, an administrative agency under the MIIT that issues digital certificates to websites.5

The CCP has exerted greater control over the press, film, radio, and television industries in recent years, including online video and streaming services. Since an administrative restructuring occurred in March 2018, more agencies involved in media regulation, including online content, have been directly subordinated to the CCP’s Central Propaganda Department, although several continue to report to the State Council.6

Zhuang Rongwen has served as CAC director since mid-20187 and continued in that position during the coverage period, while simultaneously serving as deputy director of the Central Propaganda Department.8

B Limits on Content

B1 1.00-6.00 pts0-6 pts
Does the state block or filter, or compel service providers to block or filter, internet content, particularly material that is protected by international human rights standards? 0.000 6.006

The Great Firewall is the world’s most sophisticated internet censorship apparatus. Content that contains criticism of individuals, policies, or events that are considered integral to the one-party system is blocked. The breadth of censorship is constantly growing, leaving Chinese users with access to a highly controlled, monitored, and manipulated version of the internet.

According to GreatFire.org—an anticensorship group that tracks filtering in China—as of May 2023, at least 175 of the 1,000 most heavily visited websites and social media platforms in the world were blocked in China.1 A localized version of Duolingo became accessible in June 2022, after the app was blocked in August 2021.2 Many international news outlets and their Chinese-language websites are blocked, such as those of the New York Times, Reuters, the Wall Street Journal, the Australian Broadcasting Corporation, and the British Broadcasting Corporation. The websites of independent Chinese-language news services from Taiwan, Hong Kong, and the Chinese diaspora—such as the Liberty Times in Taiwan, Initium in Singapore, and the China Digital Times in the United States—remained blocked during the coverage period.

A wide range of other websites that might provide information critical of the Chinese government are also blocked, including those of human rights groups such as Amnesty International, Human Rights Watch (HRW), and Freedom House.

Most international social media and messaging platforms are blocked, including Facebook, WhatsApp, Twitter, Instagram, Signal, Clubhouse, YouTube, Telegram, Snapchat, Line, and Pinterest.3 The popular discussion forum Reddit was blocked in August 2018,4 while all languages of Wikipedia were blocked in April 2019.5 A number of Google services—including Maps, Translate, Calendar, Docs, Drive, Scholar, and Analytics—remained blocked as of the end of the coverage period.

The long-standing blocks on international communications platforms have helped to enable the exponential growth of local products, such as messaging service WeChat and microblogging platform Sina Weibo, which are subject to the government’s strict censorship demands (see B2).

Blocks on global search engines severely limit the content available on the Chinese internet. Google’s search engine has been blocked since 2012,6 while the Yahoo search function was blocked in 2018.7

A minority of Chinese internet users, though they number in the tens of millions, access blocked websites with circumvention tools like virtual private networks (VPNs) (see B7). However, the government has intensified its restrictions on these tools since 2017, when the MIIT banned the use of unlicensed VPNs.8 Service providers are barred from setting up VPNs without government approval, and illegal VPN operations have been increasingly targeted for closure or blocking (see C3).9 Blocks on VPNs typically escalate ahead of high-profile events, such as annual plenary sessions of the Chinese legislature.10 VPN providers have noted that a growing technical sophistication of Chinese authorities has been reflected in VPN blocking incidents.11 In November 2021, the CAC released a draft regulation, titled Network Data Security Management Regulations, that would punish individuals and institutions for helping users circumvent internet censorship. Presumably targeting app stores and hosting sites, the regulations would provide for penalties of up to 500,000 yuan ($70,300).12

In October 2022, Chinese censors imposed more restrictions on circumvention tools, mainly affecting transport layer security (TLS)–based programs, according to the Great Firewall Report. It was reportedly the largest-scale block of TLS-based tools, which are widely used by Chinese internet users, to date.13

B2 1.00-4.00 pts0-4 pts
Do state or nonstate actors employ legal, administrative, or other means to force publishers, content hosts, or digital platforms to delete content, particularly material that is protected by international human rights standards? 0.000 4.004

The government requires locally hosted websites, social media platforms, and other technology companies to proactively monitor and remove significant amounts of banned content and accounts. They can face severe punishment for failure to comply.

The scale of content removals, website closures, and social media account deletions continued to expand during the coverage period, reaching new types of platforms and extending to topics that were previously uncensored. Censored topics often involve news, commentary, or criticism related to the CCP, its officials, and foreign affairs, as well as content related to health, safety, and civil society.1 Content that violates long-standing taboos is consistently and systematically censored, including content related to the June 4, 1989, Tiananmen Square massacre; Taiwanese independence; and the government’s repression of marginalized communities like ethnic minorities in Xinjiang and Tibet.

The CAC regularly launches “rectification” and “clean-up” campaigns to pressure websites and social media platforms to more effectively police content. In August 2022, the CAC announced that it “dealt with” 1.34 billion social media accounts, “cleaned up” 22 million illegal messages, and closed 3,200 websites.2

In September 2022, ahead of the 20th Congress of the CCP, the CAC launched a three-month campaign to “crack down on internet rumors and false information.”3 It ordered platforms to “strengthen accountability-seeking” and “improve rumor-countering mechanism[s].”4 Ahead of the March 2023 annual “Two Sessions” meetings, online comments on the approval of a third presidential term for Xi Jinping were systematically deleted.5 Freedom House research released in August 2023 found that almost 20 percent of a sample of 4,170 Sina Weibo posts with dissent-related language were removed.6

The authorities pressure Chinese internet companies to tightly enforce censorship regulations or risk suspensions, fines, blacklisting, closure, or even criminal prosecution of relevant personnel. This has intensified under the cybersecurity law that took effect in 2017. The CCP’s Central Propaganda Department and its local subsidiaries issue regular instructions to news sites and social media platforms on what to restrict.7 In July 2022, Chinese video-sharing platform Bilibili cracked down on “malicious behavior and remarks” by virtual live streamers.8 In July 2022, people discovered that access to private files on domestic cloud-storage applications such as WPS Office could be restricted, or that the files could be removed, as providers worked to comply with existing regulations.9

Censors increasingly target “self-media,” a category that includes independent writers, bloggers, and social media celebrities (see B6). Tens of thousands of self-media accounts have been shut down.10 In August 2022, censors suspended two nationalistic Sina Weibo accounts with tens of millions of followers.11

International companies also respond to censorship demands or pressure from the authorities to restrict online content. In October 2022, Google closed its Translate service in China.12 Grindr, an LGBT+ dating app, removed itself from Chinese app stores in February 2022, citing a new privacy law.13 Similarly, LinkedIn shut down its service in China in October 2021.14

Apple has removed or otherwise restricted apps due to regulatory and political pressure in recent years. Since 2017, it has blocked websites that appeared on a Tencent-generated denylist on its Safari browser, restricting websites deemed politically sensitive by the CCP.15 In May 2021, the New York Times reported that since 2017, roughly 55,000 active apps have disappeared from Apple’s app store in China, including hundreds of VPN services.16 Applecensorship.com counted 11,783 apps that were unavailable on Apple’s app store in China as of late May 2023.17 Apple has similarly removed or restricted iOS features; in June 2021, for example, Apple announced that it would not roll out its new privacy measure, Private Relay, in China, citing regulatory concerns.18 In November 2022, Apple restricted the use in China of AirDrop, a file-sharing feature on iPhones,19 after Shanghai subway passengers used it to spread messages about a lone protester on a bridge in Beijing.20 AirDrop, which relies on direct connections between phones, has been a crucial communication tool for protesters to evade censorship in many authoritarian countries.21

Security officials have increasingly resorted to harassing and coercing users to delete content, particularly from the platform formerly known as Twitter, which is blocked in China. A small but savvy community of internet users access Twitter via circumvention tools, enabling participation in conversations that are heavily censored within the Great Firewall, including on protests (see B8). Over the past several years, numerous users faced reprisals for their Twitter activities, including prison time, with many forced to delete their posts en masse (see C3 and C7).22

Content that criticizes the government’s repression of online speech was censored, with particularly strict measures taken against posts that are widely shared or associated with a popular hashtag. In November 2022, censors removed an episode by popular podcaster and film critic Bo Mi that criticized censorship of the film industry.23

The COVID-19 pandemic, which first emerged in the city of Wuhan in Hubei Province in late 2019, continued to be one of the most censored topics during the coverage period. For instance, in August 2022, authorities suspended social media accounts belonging to Dingxiangyuan (“Lilac Garden”), a popular online outlet that discussed medical issues, after it challenged the effectiveness of traditional medicine in treating COVID-19.24 In September 2022, after a bus accident killed at least 27 people being transferred to a COVID-19 quarantine facility in Guizhou Province, the government censored and supressed online discussions of the incident.25 In October 2022, a man in Beijing lowered two banners over a city bridge, demanding the end of the zero-COVID policy and the introduction of democratic rule in China. The act drew widespread attention on social media; censors removed photos of the protest, discussion of the man’s identity, and associated hashtags.26

In November 2022, during the unprecedented protests over the zero-COVID policy, censors removed social media posts about the protests, ordered restrictions on search functions, and removed posts calling for the release of protesters (see B8). The CAC ordered social media companies to hire more censors and to scrub references to circumvention technology, presumably to restrict access to discussion of the protests on the uncensored internet.27 Shortly after the protests, the government abruptly rescinded COVID-19-related lockdown restrictions, causing widespread infections and deaths. Sina Weibo censored search results for the topic “pandemic in Beijing,” preventing real-time discussions on the reopening’s impact.28 In January 2023, authorities announced an online crackdown to ensure there were no “gloomy sentiments” caused by pandemic “rumors.”29 Articles and social media comments discussing case numbers and deaths were quickly removed.30

Censorship of content discussing zero-COVID was particularly stringent in minority areas. Lockdowns lasting approximately one month in some areas in Xinjiang and Tibet in August 2022 led Uyghurs, Tibetans, and members of other minority groups to post their frustrations online, resulting in censorship and arrests.31

In May and June 2023, authorities censored posts related to the June 4, 1989, Tiananmen Square massacre.32

Content related to marginalized ethnic, religious, and linguistic groups is also restricted. Keywords related to the banned Falun Gong spiritual group consistently appear on leaked lists of prohibited terms. References to the banned Church of the Almighty God are also reportedly marked as politically sensitive, with mobile users facing account deactivation for sharing religious information.33 Many Cantonese live streamers using Douyin, the Chinese version of TikTok,34 have been banned for not broadcasting in Mandarin.35

LGBT+ content has been increasingly censored on China’s internet since 2017, when the China Netcasting Services Association labeled homosexuality as “abnormal sexual behavior.”36 37 In September 2021, the National Radio and Television Administration (NRTA) ordered broadcasters and the entertainment industry to ban “sissy men,” prompting a wave of content removals on social media platforms.38 Throughout 2022, censors removed online discussions about the bullying of young LGBT+ people, after several people died by suicide because of bullying.39 In August 2023, Sina Weibo took down several prominent LGBT+ accounts.40

Women’s rights content, including content related to the #MeToo movement, continues to be censored. Information about Chinese tennis star Peng Shuai—who posted on Weibo in November 2021 alleging that she had been sexually assaulted and forced into a sexual relationship with former CCP politburo standing committee member Zhang Gaoli—continued to be strictly censored on Chinese social media during the coverage period.41 In June 2022, a surveillance video in which a group of men assaulted a group of women after they rejected their sexual advances in a Tangshan restaurant went viral and caused outrage online. In response, Weibo removed accounts that it deemed “incited gender confrontation.”42 A blogger reporting on the case was arrested (see C3).

Since the Chinese government declared victory in eradicating absolute poverty at the end of 2020, online content that depicted poverty was frequently censored. In October 2022, streaming sites removed Return to Dust, a film on poverty. The name of the film was also censored on Sina Weibo.43 Content that otherwise hints on the state of the Chinese economy were also censored. The social media accounts of influential financial writer Wu Xiaobo were removed after he compared the Chinese and US technology sectors.44

Content that mocked Chinese leaders, particularly President Xi, was strictly censored. In October 2022, the image of former president Hu Jintao being removed from the 20th CCP Congress was strictly censored online.45 In November 2022, an account that posted a satirical reference to the zero-COVID policy—“Beijing man causes 1.4 billion people to be quarantined long-term”—was removed.46 Censors on Sina Weibo, WeChat, Douyin, and other platforms deleted a Buick commercial that could have been interpreted as a comment on Xi staying in power. The commercial invited viewers to sit in a vehicle’s seats and said they were “so comfortable” the occupants may “never want to leave.”47 According to a leaked list, 35,467 different phrases linked to President Xi were blocked online.48

In August 2022, during then US House of Representatives speaker Nancy Pelosi’s visit to Taiwan, and as Chinese internet users criticized Beijing for failing to stop her visit, social media commentary was increasingly censored.49 In February 2023, American authorities shot down an alleged spy balloon from China, which had flown over the United States. Discussions on the state of Beijing-Washington relations after the incident were censored, while searches for the phrase “spy balloon” were restricted.50

Foreign governments’ official accounts were also censored. In July 2022, UK diplomats’ Sina Weibo posts on the Sino-British Joint Declaration on Hong Kong was erased 30 minutes after they were published.51 In September 2022, censors on WeChat and Sina Weibo targeted posts from the US embassy on a UN report discussing human rights abuses in Xinjiang.52

Large language model (LLM)–based chatbots were also subjected to censorship during the coverage period. In February 2023, Chinese regulators told technology firms in the country to discontinue access to ChatGPT and to disclose their own plans to develop AI-driven chatbots.53 A March 2023 Wall Street Journal report found Chinese chatbots refused to answer questions related to President Xi.54 The founder of ChatYuan, an AI chatbot, said that the chatbot would “filter certain keywords” with more layers of review than might be expected overseas.55

Automation is playing an increasingly important role in censorship. In August 2019, Citizen Lab revealed the existence image-filtering capabilities on WeChat, which targeted users’ creative efforts to circumvent text-based censorship through image-based commentary.56 Alibaba, Tencent, ByteDance, and the People’s Daily have emerged as industry leaders in content moderation and censorship technologies that intentionally target political content, selling the systems to other Chinese companies as well as foreign clients.57

B3 1.00-4.00 pts0-4 pts
Do restrictions on the internet and digital content lack transparency, proportionality to the stated aims, or an independent appeals process? 0.000 4.004

Censorship decisions are arbitrary, opaque, and inconsistent because the nation’s rule of law is weak and because of the number of actors and processes involved. Regulations issued by government and CCP agencies establish censorship guidelines. The impact of content restrictions may vary depending on factors like timing, technology, and geographic region. ISPs reportedly install filtering devices differently, including in the internet backbone or even in provincial-level internal networks.1 Lists of prohibited websites and sweeping censorship directives are closely held secrets but are periodically leaked. There are no formal avenues for appeal, and directives cannot be challenged in the courts. Criticism of censorship is itself censored.2 There is no transparency surrounding private companies’ day-to-day censorship in China, and users similarly lack avenues for appeal.

Article 84 of a 2015 antiterrorism law introduced fines and detentions of up to 15 days for telecommunications firms and ISPs, as well as relevant personnel, who fail to “stop transmission” of terrorist or extremist content; “shut down related services”; or implement “network security” measures to prevent the transmission of such content (see C2).3 The cybersecurity law, in effect since 2017, also provides legal grounds for officials to instruct network operators to stop the transmission of certain content to protect public security, among other restrictions (see A3).

The CAC and other bodies routinely introduce new rules and guidelines to further refine online restrictions, with an increasing focus on user-generated content. In December 2022, the CAC introduced new rules that require technology companies to review all social media posts before they are published and filter out “illegal and harmful” information.4 In December 2022, the CAC updated the 2017 Regulations on the Administration of Internet Post Comment Services, clarifying the responsibilities of ISPs and other operators to filter out “harmful” content.5 Under CAC regulations which took effect in March 2022, platforms’ recommendation algorithms must remove “illegal and undesirable content,” adhere to “mainstream values,” and promote “positive energy.” They also impose algorithmic transparency requirements on companies and require them to permit users to decide whether to enable automated content recommendation systems.6

CAC regulations, released in draft form in April 2023 and as interim measures in July, require content generated by LLMs to embody “core socialist values” and avoid “information that is violent, obscene, or fake.” The rules designate generative AI providers as online information content producers, subjecting them to CAC censorship regulations.7 In January 2023, new rules targeting deepfake technology went into effect. Deepfake providers must explicitly label and make traceable any doctored content; must abide by local laws, including national security requirements; and maintain the “correct political direction and correct public opinion orientation.”8

In March 2021, the NRTA published draft amendments to the Radio and Television Law expanding its coverage to include online video broadcasters and platforms.9 The amendments specify nine types of banned content, including content that “endangers security,” “slanders Chinese culture,” or does not help youth “establish the correct world view.”10

In June 2022, the NRTA and the Ministry of Culture jointly issued a new code of conduct for live streamers, podcasters, and other online content producers. The document banned any content that “weakens, distorts or denies the leadership of the CCP.”11 Previously, the NRTA released rules for the country’s massive live-streaming industry in November 2020, requiring that platforms notify authorities of celebrity and foreigner appearances ahead of time, and that they promote accounts that embody the “core socialist values.”12 In November 2021, the NRTA released a five-year blueprint specifying new rules designed to regulate content on live-streaming and short-video platforms, including a mechanism that will punish producers of what the CCP considers “toxic culture.”13

B4 1.00-4.00 pts0-4 pts
Do online journalists, commentators, and ordinary users practice self-censorship? 0.000 4.004

Self-censorship among ordinary users and journalists is common and takes place amid an increasing risk of account closures, real-world reprisals, and legal penalties for online commentary (see B2, C3, and C7). Self-censorship is exacerbated by nationalistic netizens’ intimidation and online harassment of those who they perceive as harming the reputation of China.

There is evidence that WeChat users are increasingly self-censoring to preempt the closure of their accounts or other penalties, since WeChat is relied on for a wide variety of services that include messaging, banking, ride-hailing, ordering food, and booking travel (see B2).1 The app’s critical role in daily life, alongside platform moderators’ growing propensity to close accounts rather than delete objectionable posts, has increased pressure on users to self-censor.

Self-censorship is pervasive among members of persecuted groups, especially Uyghurs, whose WeChat activities are closely monitored. Many block relatives living abroad to avoid being detained for having foreign contacts.2

Despite these pressures, various examples emerged during the coverage period of internet users speaking out on sensitive topics. Since the death of Dr. Li Wenliang, a COVID-19 whistleblower, millions of people have left comments under his final Sina Weibo post. It has become known as China’s “wailing wall” and as a place to deliver messages critical of Beijing. Many users left comments commemorating Li on February 6, 2023, the third anniversary of his death.3

In Guangzhou, residents used vernacular Cantonese terms, instead of Mandarin, to express their frustrations with the government’s draconian COVID-19 policy.4 Similarly, ethnic Mongolians used Mongolian script to post criticism of the authorities online.5

B5 1.00-4.00 pts0-4 pts
Are online sources of information controlled or manipulated by the government or other powerful actors to advance a particular political interest? 0.000 4.004

The government has significant control over digital news media and other information sources. Online discussion is subject to extensive manipulation. Websites and social media accounts, other than those operated by official news outlets, are not legally allowed to produce news content, though the definition of what constitutes “news” is unclear. Propaganda officials systematically instruct internet outlets to amplify content from state media and downplay news that might generate public criticism of the government.1

The Provisions on the Governance of the Online Information Content Ecosystem, implemented in March 2020,2 categorize online content as either encouraged positive content, discouraged negative content, or illegal content. Encouraged material includes “spreading party doctrine,” while negative categories include “excessive celebrity gossip” and sensationalist headlines. According to the rules, the encouraged content must be actively promoted in prominent online locations such as on web portal home pages, pop-up windows, hot topic lists, and default search results.3 They also call for online content providers to promote officially approved “mainstream values” via algorithms. Most of these actions had been occurring for years, but the provisions consolidated them into a single set of rules.

Paid commentators, known informally as the 50 Cent Party, are employed to post progovernment remarks and influence online discussions.4 Such commentators are known for reporting users who post offending statements, deliberately muddying the facts of particular incidents, and coordinating smear campaigns against government critics.5 According to a report released in December 2020, a robust government-funded industry of automated online commentating floods social media platforms with desired posts, even under the direction of small local agencies.6

A March 2022 report by German Marshall Fund showed that state media outlets promoted pro-Russia narratives on official social media accounts, including the false claim that the United States is funding biological weapons laboratories in Ukraine.7 In December 2022, the Intercept reported that Chinese and Russian officials signed an agreement to collaborate on the dissemination of favorable narratives in July 2021.8

In December 2022, a leaked propaganda directive showed the government ordered state media to portray the abrupt end of the zero-COVID policy as a well-organized, scientifically backed decision.9 The decision followed unprecedented mass protests against the policy (see B8).

Since 2017, content manipulation and disinformation campaigns have increasingly extended even to platforms that are blocked in China, including Facebook, Twitter, and YouTube,10 demonstrating an ability to influence online discourse internationally.11 In August 2022, intelligence firm Mandiant uncovered 72 disinformation sites in 11 different languages run by a China-based company.12 The sites attacked critics of Beijing’s crackdowns in Hong Kong and Xinjiang and promoted pro-CCP content.13 In August 2022, a wave of pro-CCP propaganda books promoting China’s claim over Taiwan were released on Amazon’s Kindle platform.14

In September 2022, Meta announced that it had taken down a China-linked network that disseminated disinformation about the November 2022 US election, the Czech Republic’s policy on China, and the Russian regime’s war against Ukraine for violating the platform’s policy against coordinated inauthentic behavior.15 In September 2022, an inauthentic campaign on Twitter targeted the NGO Safeguard Defenders, after it published a report on Chinese “overseas police stations.”16 An October 2022 report by the Australian Strategic Policy Institute (ASPI) showed that the government worked with international influencer-management agencies and popular YouTube influencers to spread pro-CCP propaganda on Uyghurs.17 A February 2023 ASPI report uncovered a Chinese government-linked social media campaign to spread disinformation about Australian politics.18 A February 2023 report by cybersecurity firm Graphika revealed the employment of AI-generated avatars in a disinformation campaign linked to China-based actors.19

In addition to paid commentators, local authorities have mobilized ziganwu—volunteer commentators motivated by ideology rather than money—to promote the government’s narratives online.20 A document leaked in 2015 revealed hundreds of thousands of “youth league online commentators” in China’s higher education institutions, tasked with turning students against supposed “Western” democratic values.21 Throughout 2022, nationalistic internet users attacked science blogs and feminist and LGBT+ accounts for their being “anti-China,” leading to those accounts and websites shutting down.22 A December 2022 Nikkei Asia article reported that nationalist users gained significant traction on Chinese social media in the past decade, amplified by official accounts and government censorship of more moderate content.23

B6 1.00-3.00 pts0-3 pts
Are there economic or regulatory constraints that negatively affect users’ ability to publish content online? 0.000 3.003

Growing censorship demands, licensing requirements, and data-localization mandates have made it more expensive to run internet-specific companies, including online news outlets, in China. While large companies have been able to absorb rising operational costs, new and smaller players operate with increasing difficulty. Arbitrary regulatory decisions have also contributed to an unstable investment climate.1

Under CAC regulations on managing internet news and information services that came into effect in 2017,2 only traditional media or state-controlled enterprises may obtain a license to gather or disseminate news online.3 Regulations from 2016 restrict foreign investment in online publishing and require at least eight full-time editorial or publishing staff members.4 In addition, CAC rules have made it harder for both independent media and individual bloggers, journalists, and writers to sustain themselves financially.5 Commercial media outlets such as Caixin, known for relatively aggressive and investigative reporting, have suffered from falling profits due to censorship and ad hoc directives requiring major news portals and other aggregators to favor state media. In January 2021, the CAC further restricted online self-publishing (individuals or companies that only publish through social media), requiring independently operated accounts to obtain a permit and prohibiting them from commenting on a list of topics.6

In October 2021, the CAC released an updated version of the Internet News Information Source List, containing over 1,300 authorized news outlets that can be republished by other news services—four times as many as the CAC’s 2016 list. The list removed Caixin and added the social media accounts of state media and government agencies.7 In August 2021, China’s top economic regulator released a draft regulation that would ban private investment in the media, including for the establishment or operation of news sites.8 The draft regulation would add further economic barriers for online outlets.

In November 2022, the CAC released an updated version of Regulations on Comments to Social Media Posts Services, requiring platforms to enforce real-name registration and roll out new content-moderation controls on comment threads (see C4).9

B7 1.00-4.00 pts0-4 pts
Does the online information landscape lack diversity and reliability? 1.001 4.004

China’s online information landscape is significantly less diverse than it had been before Xi Jinping came to power due to increasing censorship. This strict censorship of critical viewpoints and foreign news sources has also empowered nationalist and conservative voices, which are disproportionately represented in Chinese cyberspace.1 Nevertheless, the internet still provides narrow avenues for critical discussion and enables the sharing of information on some important social and political issues, particularly when users devise creative workarounds to disguise discussion.2

The stringent penalization of groups perceived as a political threat has effectively diminished the online space for civil society in recent years.3 While groups that work on less sensitive issues have a vigorous online presence, organizations and topics that were once tolerated, such as feminist group, have faced increasing scrutiny.4

Many members of marginalized ethnic, religious, and linguistic groups attempt to use the internet to disseminate banned content, but these views remain underrepresented online. For example, Uyghur-language content and relevant news reporting have been heavily censored, and many ordinary Uyghur users have been detained (see B2 and C3); Islamophobic commentary is permitted to circulate widely.5

Tens of millions of internet users bypass censorship with circumvention technology or creative workarounds. Although hundreds of VPN services are inaccessible,6 various options remain available.7 In late 2022, many Chinese internet users turned to VPNs to circumvent censorship in order to share and access information related to zero-COVID protests (see B8).8

Within the Great Firewall, netizens deploy neologisms, homonyms, and cryptic allusions to substitute for banned keywords, forcing censors to filter seemingly innocuous terms like “rice bunny,” which is pronounced “mi tu” in Mandarin and has become a well-known alternative for the intermittently censored #MeToo hashtag.9 For the past several years, the word “Xinjiang” and the human rights abuses documented there had been taboo on the Chinese internet. Netizens thus used “XJ” and “new jiang” (“xin” means new in Chinese) to try to circumvent censorship.10 In a similar vein, “JC” was used to reference the police (“jin cha” means police), and “zf” for government (“zhengfu” means government).11 In July 2022, Sina Weibo announced new rules that sought to restrict the use of homonyms to “spread harmful information.”12

B8 1.00-6.00 pts0-6 pts
Do conditions impede users’ ability to mobilize, form communities, and campaign, particularly on political and social issues? 1.001 6.006

The role of social media in providing a vibrant space for activism in China has waned significantly due to stricter internet controls under Xi Jinping.1 Growing censorship of popular apolitical platforms such as dating, video-sharing, live-streaming, and blockchain applications in recent years has effectively closed avenues users had used to disseminate information and mobilize. However, people mobilized unprecedented protests during the coverage period. Data collected for the China Dissent Monitor, which is operated by Freedom House, identified over 50 distinct instances of online mobilization during the coverage period, spanning issues like COVID-19 measures, land rights, and corruption.2

In November 2022, thousands of people across China held protests against Beijing’s zero-COVID policy and CCP rule more generally. The protest was spurred by a fire in Urumqi that killed 10 people that month, as an ongoing coronavirus-related lockdown persisted amid rising popular discontent over the zero-COVID policy. People took to social media platforms to mourn the victims and share information about protests. Censors responded by removing social media posts and accounts sharing news about the protests, and search functions on social media platforms made protest-related information hard to find (see B2).3 Chinese protesters also used circumvention tools to discuss the protests on global platforms like Facebook and Instagram.4 Authorities detained scores of protesters across the country (see C3). Online posts discussing their cases or calling for their release were censored. The CAC also ordered social media companies to hire more censors to scrub any references to VPNs.5

While overt activism on political issues is exceedingly rare, China’s #MeToo movement continues to demonstrate that citizens can organize around select social causes, and that their efforts sometimes have a real-world impact. Despite heavy censorship of the global campaign, #MeToo encouraged several Chinese students to speak out online about their experiences with sexual harassment, leading to a few instances of positive change.6 The civil code, released in June 2020, bans sexual harassment,7 a move seen as a response to numerous cases of professors harassing female students on college campuses.8

C Violations of User Rights

C1 1.00-6.00 pts0-6 pts
Do the constitution or other laws fail to protect rights such as freedom of expression, access to information, and press freedom, including on the internet, and are they enforced by a judiciary that lacks independence? 0.000 6.006

Article 35 of the constitution guarantees freedoms of speech and publication, but such rights are subordinated to the CCP’s status as the ruling party. The constitution cannot be invoked in courts as a legal basis for asserting rights. The judiciary is not independent and closely follows party directives, particularly in politically sensitive cases involving freedom of expression.

C2 1.00-4.00 pts0-4 pts
Are there laws that assign criminal penalties or civil liability for online activities, particularly those that are protected under international human rights standards? 0.000 4.004

Numerous laws and regulations limit online activities, and prosecutors exploit vague provisions to imprison people for their online speech. Trials and hearings typically lack due process. It can take years for cases to move through the court system; the accused are routinely denied bail and face lengthy pretrial detention.

Laws prohibiting offenses including defamation, creating disturbances, illegal commercial activities, and extortion have implications for online speech.1 Defamation has been interpreted to include “online rumors,” content deemed false, or online expression that “seriously harms” public order or state interests.2 It carries a possible three-year prison sentence under “serious” circumstances, which apply when the content in question receives more than 5,000 views or is reposted more than 500 times.3 Online messages deemed to incite unrest or protests are subject to criminal penalties under provisions punishing citizens for “picking quarrels and provoking trouble.”

Crimes such as “subversion” and “separatism,” as well as the incitement of such actions, can draw sentences as severe as life in prison.4 Article 300 of the criminal code punishes “using heterodox religion to undermine implementation of the law” and is often invoked against members of banned religious groups.5 A 2015 amendment to the criminal code increased the maximum penalties for these crimes from 15 years to life imprisonment6 and introduced penalties of up to seven years in prison for disseminating misinformation on social media.7

A March 2021 amendment to the criminal code stipulated that those who “insult, slander, or infringe the reputation and honor of heroes and martyrs” can be imprisoned for up to three years.8 A law in was introduced in June 2021 to ban “slander” of members of the armed forces.9

The 2015 antiterrorism law bars social media users from sharing information about acts of terrorism or spreading “inhuman” images that could encourage copycat attacks (see B3 and C5).10

In February 2020, amid the coronavirus outbreak, top judicial and law enforcement agencies released new guidelines for judges, prosecutors, and others working in the legal system urging strong action against crimes seen as weakening disease-control efforts and undermining the CCP’s authority. Among the 10 categories of crimes listed for tighter enforcement were “spreading false information and rumors online” and “disrupting social order, especially maliciously attacking the party and government, taking the opportunity to incite subversion of state power, or overthrow of the socialist system.”11

Some detentions can occur without court approval,12 and individuals can be detained without trial under poor conditions in drug rehabilitation centers.13 Chinese law also allows a form of criminal detention termed “residential surveillance at a designated location”14 where police may hold individuals in secret locations without informing their families or legal counsel for up to six months.15

C3 1.00-6.00 pts0-6 pts
Are individuals penalized for online activities, particularly those that are protected under international human rights standards? 0.000 6.006

Chinese citizens are regularly jailed for their online activities, and the risk of being detained or imprisoned has increased considerably in recent years. Rapid advances in surveillance technology and growing police access to user data have helped facilitate the rise in prosecutions (see C5 and C6).1

Zero-COVID protesters (see B8) were detained during the coverage period, including for their online activity. A Shanghai resident was detained in December 2022 after alleging on Twitter that police had tortured people arrested in the protests; the resident was reportedly released by January 2023. A Uyghur student residing in Henan Province who posted videos relating to the protests on WeChat was detained in December 2022, when she had returned to Xinjiang,2 and was sentenced to three years’ imprisonment in March 2023.3 Authorities also continued to detain or prosecute people for their online speech related to the COVID-19 pandemic. In September 2022, Xinjiang police detained four people for “spreading rumors” and “disrupting the order of anti-pandemic measures.”4 Data collected for the China Dissent Monitor, which is operated by Freedom House, identified over a dozen cases where online dissent resulted in detention or arrest during the coverage period (see B8).5

Journalists in China are frequently imprisoned for their work, online writing, or video posts. According to the Committee to Protect Journalists (CPJ), at least 43 journalists were jailed in China as of December 2022, second only to Iran in the number of imprisoned journalists.6 In February 2023, Ruan Xiaohuan, who had anonymously operated the blog Program Think, was sentenced to seven years’ imprisonment for “inciting subversion of state power.”7 In September 2021, Huang Xueqin, a prominent blogger and #MeToo activist, was detained by Guangzhou police on the same charge;8 she remained detained as of the end of the coverage period. In February 2023, advocates warned that Huang’s physical condition was deteriorating.9 In August 2022, blogger Mao Huibin was arrested and charged with “picking quarrels and provoking trouble” after reporting on police obstruction of investigations into a gender-based assault in Tangshan.10

Activists and human rights lawyers have been prosecuted for advocating for democratic rights and governance, exposing police abuses, unionizing efforts, and other online activities. In April 2023, a court in Shandong Province sentenced prominent human rights lawyer and blogger Xu Zhiyong to 14 years in prison for “subversion of state power.”11 In June 2023, a Shaanxi court sentenced human rights lawyer Chang Weiping to three and half years in prison under the same charge.12

In August 2022, the human rights website Weiquanwang reported that dissident writer Sun Daluo, formerly known as Sun Zhiming, had been tried in secret in December 2021 on charges of “picking quarrels and provoking trouble,” but the details of the sentence remained unknown.13 In December 2022, human rights activist Ou Biaofeng received a three-and-a-half-year prison sentence for writing articles for Hong Kong newspapers and for his social media activity.14 In December 2022, poet Wang Zang and his wife, Wang Liqin, received four-year and two-and-a-half-year prison sentences, respectively, in Sichuan. Wang Zang’s sentence was in relation to his public online comments and media interviews. Wang Liqin appeared to have been convicted for supporting Wang Zang.15

Members of persecuted religious and ethnic minority groups face particularly harsh treatment for their online activities. In Xinjiang, an estimated one million people are being held in political reeducation camps or forced labor facilities.16 HRW reported in September 2022 that some 500,000 people have received formal prison sentences.17 Some were targeted for their online activities that included communication with relatives living abroad via WeChat.18 A police officer confirmed in June 2023 that a university student was sentenced to 13 years in prison in 2017 after the student used a VPN and viewed “illegal information.”19

Tibetans living outside the Tibetan Autonomous Region have also been targeted for sharing information on Chinese social media or overseas websites. Tibetan activist and writer Thupten Lodoe was arrested in October 2021 in Kardze Tibetan Autonomous Prefecture in Sichuan Province and sentenced to four-and-a-half years’ imprisonment in June 2022 for “inciting separatism” in social media posts.20 In July 2022, authorities arrested a Tibetan man in Sichuan for creating an “unlawful” WeChat group to honor Tibetan religious leaders.21

People in the Tibetan Autonomous Region, a region which is not factored into this report’s scores (see Overview), also face arrest and prison terms for their online speech. In August 2022, a teacher in Lhasa was arrested for posts on WeChat and Sina Weibo documenting the harmful implementation of the city’s harsh COVID-19 lockdown.22 In March 2023, a Tibetan woman in the region was detained for sending photos to people outside of it.23

Many Falun Gong practitioners have been jailed in recent years for posting messages about the spiritual group or human rights abuses on social media, accessing banned websites, and possessing or sharing prohibited VPN technology.24 In June 2022, according to the Falun Gong information site Minghui, Falun Gong practitioner Guo Xiuhai was sentenced to nearly four years in prison for sharing a video about the religious practice on WeChat.25

Vague provisions barring online speech have been applied to people using generative AI tools. In May 2023, police in Gansu detained a man on charges of “picking quarrels and provoking trouble” after he allegedly used ChatGPT to generate and post a series of false news reports about a train crash.26

C4 1.00-4.00 pts0-4 pts
Does the government place restrictions on anonymous communication or encryption? 0.000 4.004

Anonymous communication is severely limited. Under 2012 data-privacy legislation and additional 2017 rules,1 web service companies are required to register users under their real names and national ID numbers.2

Draft guidelines released by the CAC in October 2021 direct services to use real-name registration to prevent users banned for legal or platform policy violations from reregistering and to disclose the geographic locations of users.3 In March and April 2022, social media platforms like Sina Weibo and Douyin began to display the cities or provinces of China-based users underneath posts. Users outside of China have their country displayed.4 In December 2022, the CAC released new rules that require ISPs to discontinue service to anyone who had not authenticated their real identity.5

Authorities also enforce SIM card registration,6 and in December 2019 regulations took effect that require users to have their faces scanned when registering for mobile services.7 All online gamers are also required to register using their name and phone number,8 and real-name registration for online literature platforms became mandatory in June 2020, limiting online spaces where many novelists have turned to discuss sensitive subjects in recent years.9 The CAC released draft guidelines in January 2022 that extend real-name registration requirements to companies that provide image-manipulation services based on machine learning, commonly known as deepfakes.10

Authorities in some areas have instructed public Wi-Fi providers to comply with user registration requirements.11 Cybercafés check photo identification, record user activities, and at times require facial scans, sometimes in cooperation with law enforcement.

Measures that erode anonymity disproportionately target groups that are perceived as threats to the regime. In Xinjiang, Uyghurs have been required since 2015 to register with their real names when purchasing electronic devices with storage, communication, and broadcast features. Stores selling such equipment are also required to install software that provides police with real-time electronic records on transactions.12

The use of encryption is also severely restricted. The 2015 antiterrorism law requires companies to offer technical support to decrypt information at the request of law enforcement agencies, among other provisions.13 Regulations for the Administration of Commercial Encryption dating to 1999, and related rules from 2006, separately require a government regulator to approve encryption products used by foreign and domestic companies.14 In January 2020, a law took effect that requires critical information infrastructure providers to apply for a review by the CAC if their use of encryption technologies is viewed as potentially impacting national security.15

In May 2021, the New York Times reported that Apple abandoned the encryption technology it typically uses when storing user data in China after the Chinese government prohibited the technology (see C6).16 In February 2023, only two days after it became available, Apple removed the new Damus app— a decentralized platform that allows users to create anonymous accounts and send encrypted messages—from its app store in China at the request of the CAC (see B2).17

C5 1.00-6.00 pts0-6 pts
Does state surveillance of internet activities infringe on users’ right to privacy? 0.000 6.006

Online surveillance is pervasive and highly sophisticated, and privacy protections from government infringement under Chinese law are minimal. In recent years, the Chinese government has increasingly moved toward big-data integration with the help of private companies, essentially consolidating in various databases a wide array of information on individuals, including their online activities.

According to a study by ChinaFile, one-third of Chinese counties purchased surveillance equipment— including facial recognition–enabled cameras, databases for storing citizen information and images, and Wi-Fi sniffers to monitor internet traffic—in 2019 alone. ChinaFile had examined over 76,000 local files published between 2004 and May 2020.1 Much of the equipment is connected to Project Sharp Eyes, which aims to cover all key public spaces in China by video surveillance. An October 2022 report by Kaspersky Labs uncovered spyware was bundled inside an inauthentic version of Tor advertised in China. Tor, a popular anonymizing browser, is blocked in China.2

When conducting investigations, the authorities have unfettered access to user communications and data on certain popular platforms, as indicated by reports of users being punished for their presumably private conversations, particularly on WeChat. A June 2022 New York Times investigation into bids for government surveillance contracts compiled by ChinaFile found that police sought to use international mobile subscriber identity (IMSI)–catchers and Wi-Fi sniffers to extract information about people’s mobile phone usage, like social media handles and whether they have downloaded apps deemed problematic.3 Following the late-2022 zero-COVID-related protests, authorities deployed China’s extensive surveillance system, including by using mobile phone data and surveillance cameras, to identify and then arrest protesters (see B8).4

Residents of Xinjiang are subject to severely invasive surveillance tactics. A leaked list of Uyghur detainees examined in a December 2020 report by HRW showed how Chinese authorities used big-data technology to arbitrarily detain Muslims in Xinjiang. Dated from late 2018, the list from Xinjiang’s Aksu Prefecture names 2,000 people flagged by the Integrated Joint Operations Platform for surveillance and detention. The research showed that the vast majority of the people on the list were targeted by police for nonviolent and mundane behavior like owning multiple mobile phones, receiving phone calls from relatives in foreign “sensitive countries,” wearing religious clothing, using Skype, or studying the Quran. Some of the punished actions appear to have taken place years or even decades ago—such as studying the Quran in the mid-1980s or staying overnight in another part of Xinjiang in 2013.5

Surveillance technologies and policies deployed in one part of China are often later expanded to other parts of the country, with Xinjiang emerging as a particularly important testing ground.6 Chinese border police are reported to have installed surveillance apps on the phones of tourists traveling in Xinjiang, which grant authorities the ability to extract user data and identify politically and culturally sensitive material stored on targeted devices.7 In 2019, reports emerged that Chinese border officers were beginning to check the photographs, messages, and apps on the mobile phones of anyone arriving in China from Hong Kong for evidence of support for the prodemocracy protest movement in that city.8

Existing, albeit geographically disparate, data sets assembled through surveillance efforts could feed into “social credit” systems that create an assessment of individuals’ online activities and other personal data; a few municipalities have tested such systems,9 which have been subject to criticism within and outside China due to privacy and other concerns. A fully integrated national social credit system does not yet exist, and questions have been raised about the scale and effectiveness of local pilot efforts.10 Nevertheless, an individual listed as problematic by municipal or provincial authorities under such systems could face restrictions on movement, education, and financial transactions.11

Overbroad surveillance and poor data security practices have facilitated the sale of people’s private information on the open market, as with the July 2022 Shanghai police database leaks (see C8). Chinese authorities were able to amass massively intrusive data during the COVID-19 pandemic through the mandatory tracking app Health Code and other applications (see C6).

C6 1.00-6.00 pts0-6 pts
Does monitoring and collection of user data by service providers and other technology companies infringe on users’ right to privacy? 0.000 6.006

Internet-specific companies are required under numerous laws and regulations to assist the government in monitoring users’ online activities. Recent regulations include a September 2018 rule that provides security authorities with broad powers to enter the premises of all internet service companies to inspect and copy any information deemed important to cybersecurity.1 The regulatory package complemented the 2017 cybersecurity law’s requirement that network operators assist police and security agencies with criminal investigations or national security operations.2

Other surveillance laws include a 2013 amendment to the criminal procedure code that lays out a vague review process for allowing police monitoring of people’s electronic communications, which the Ministry of Public Security permits in many types of criminal investigations.3 The State Secrets Law obliges telecommunications companies to cooperate with authorities investigating leaked state secrets, or risk losing their licenses.4

Individuals or entities that refuse to comply with government requests for surveillance assistance risk detention or criminal punishment. A National Intelligence Law adopted in 2017 states that those deemed to be obstructing national intelligence work can be subject to 15 days of administrative detention and further criminal charges.5

Various regulations outline requirements for companies to retain and make user data available to officials.6 CAC rules issued in 2016 oblige Chinese app providers to register users and keep user activity logs for 60 days.7 ISPs are required to retain user information for 60 days and submit it to the authorities upon request, without judicial oversight or transparency.8

Tencent, which operates the widely used WeChat and QQ social platforms, directly assists the Chinese government with surveillance. Research by Citizen Lab found in early 2020 that WeChat was monitoring messages and images sent by users registered outside of China, scanning them for politically sensitive keywords and retaining the relevant conversations in order to train the platform’s censorship system.9 In October 2021, after a backlash over apps accessing users’ photo libraries, WeChat pledged that it would stop accessing users’ photos.10

The e-commerce giant Alibaba also helps the government with surveillance. In 2019, research published by the Open Technology Fund revealed evidence that the government’s Study the Great Nation app, which was designed by Alibaba to promote the CCP’s ideology and encourage the study of “Xi Jinping Thought” in particular, was built with a backdoor that could grant the authorities access to users’ browsing histories, text messages, photos, and other personal information.11

In September 2021, a Lithuanian government security audit found that a Xiaomi phone included software modules specifically designed to leak data to Chinese authorities and censor information related to topics the Chinese government considered sensitive. The assessment also found suspicious activities on a Huawei phone.12

The 2017 cybersecurity law mandates that internet companies store the data of Chinese residents on servers based in the country, a practice that makes it easier for the government to access user information.13 In 2018, Apple’s iCloud began storing the data of its Chinese users in partnership with Guizhou-Cloud Big Data, a state-run company;14 a second data center was expected to open in Inner Mongolia.15 A May 2021 report from the New York Times further clarified how Chinese government workers operate Apple-owned data centers in the country, detailing their physical control of computers and administration of other technology.

COVID-19-related “health code” apps—which determine an individual’s risk of being exposed to COVID-19—developed by regional officials in partnership with Alibaba and Tencent, do not adhere to privacy-by-design standards. For example, the privacy policy of Beijing’s app did not state any time limit on the retention of data.16 A New York Times investigation showed that the Alipay Health Code app automatically shared data with the police.17 Authorities appear to have used the health code app system to arbitrarily restrict movements of citizens involved in activism. For example, in June 2022, people involved mobilizing to demand banks in Henan unfreeze their deposits had their health codes turned red before a planned protest.18

In December 2022, the government ended the use of its travel tracing app;19 it is unknown what will happen to user data collected by the app. In March 2023, authorities in the city of Wuxi said they deleted one billion pieces of personal data collected since the start of the pandemic;20 the claim could not be independently verified.

The government also seeks to proactively counter efforts to evade surveillance. In March 2021, the CAC announced that it started talks with technology firms, including ByteDance, Tencent, Alibaba, Xiaomi, and Kuaishou, to explore how to counteract deepfake technologies and voice-changing software, which are often used by activists to elude identification by government authorities.21

Several government efforts aim to regulate how Chinese tech companies collect, share, and store users’ data. The Personal Information Protection Law (PIPL) went into effect in November 2021.22 The law, which applies to all organizations and individuals in China seeking to access Chinese citizens’ data, is the country’s first comprehensive legislation about the protection of personal information. However, the law exempts government agencies from data protection obligations and requires certain companies to store sensitive data on servers located in China, which leaves them vulnerable to government access.23

In March 2021, the CAC issued the Provisions on the Scope of Necessary Personal Information for Common Used Mobile Internet Applications, defining what constitutes “necessary personal information” that mobile internet applications can require consumers to provide.24 Regulators subsequently alleged that over 100 apps—including those from Tencent and Baidu—violated the rules.25 Separately, in July 2021, authorities asserted that the ride-hailing app Didi illegally collected users’ personal information; the app was pulled from China-based app stores.26 In December 2021, authorities ordered app stores to remove over 100 apps, including the relatively liberal forum Douban, for data privacy and security violations among other offenses.27

In May 2020, the NPC passed the country’s first civil code, which included a provision that requires an individual’s consent for private companies to collect, share, or disclose their digital and biometric data. The provisions came amid growing public concern in China over data collection, hacking, and potential leaks by private companies.28

In September 2021, the Data Security Law29 took effect. The law imposes extensive data security obligations for businesses. The law regulates data processing and management activities within China, along with those outside of China that would purportedly harm China’s national security or the public interest of Chinese citizens or organizations. It requires companies to obtain approval from the state prior to sharing data with a foreign judicial or law enforcement entity.30

In March 2023, the Chinese government announced a sweeping plan to restructure government agencies. A new national data bureau will be created for coordinating the sharing and development of data resources. The bureau will be overseen by the National Development and Reform Commission.31

C7 1.00-5.00 pts0-5 pts
Are individuals subject to extralegal intimidation or physical violence by state authorities or any other actor in relation to their online activities? 0.000 5.005

Cases of extralegal intimidation and violence involving internet users are widespread, including in detention. People detained in ordinary criminal cases often experience torture, and political and religious prisoners experience especially severe treatment.1

Law enforcement officials frequently summon individuals for questioning in relation to online activity, an intimidation tactic referred to euphemistically as being “invited to tea.”2 For example, activists who expressed opposition to the Chinese government's attempts to exercise greater political control over Hong Kong have been summoned.3

Activists have also been experienced movement restrictions during sensitive political events, effectively keeping them away from their normal online activities.4 For example, Wang Liqin, the wife of the imprisoned poet Wang Zang, said she and their children were under around-the-clock surveillance by agents; that surveillance intensified when she posted about it on social media.5

In May and June 2023, ahead of the anniversary of the Tiananmen Square massacre, police restricted the movement and communications of families of Tiananmen victims and activists. Hunan authorities detained activist Chen Siming after he refused to delete his Twitter post commemorating the events at Tiananmen Square.6 In September 2023, after the coverage period, Chen disclosed that he had fled to Taiwan in July and said he would seek asylum in the United States or Canada.7

Journalists sometimes experienced physical violence for their work. In May 2023, a reporter with the online outlet Jimu News was beaten by police in Bijie while reporting on the flood-related deaths in the area.8

A June 2022 ASPI report documented the harassment targeted at female researchers, journalists, and activists of Asian descent across all major social media platforms, including women living in China. The tactics include surveillance, intimidation, and disinformation campaigns.9

Members of marginalized religious and ethnic minority groups are among the internet users most vulnerable to extralegal detention, torture, and killing. In Xinjiang, some of the estimated one million Uyghurs and other Muslims were taken to reeducation camps as a result of their online activities.10 Those detained or imprisoned typically suffer harsh conditions and often torture,11 resulting in long-term health problems and sometimes death.

People in Tibet, a region that is not covered by this report (see Overview), also face physical violence for their online activity. For example, HRW reported in July 2021 that Tibetan monk Choegyal Wangpo was arrested and severely beaten in 2019 after police found his phone at a café, which contained WeChat messages to other monks in Nepal. The police then raided his village and detained approximately 20 other monks.12

C8 1.00-3.00 pts0-3 pts
Are websites, governmental and private entities, service providers, or individual users subject to widespread hacking and other forms of cyberattack? 0.000 3.003

China is a significant origin point of global cyberattacks,1 though those with commercial targets have declined or become harder to trace.2 Attacks known to have originated in China can rarely be linked directly to the state, and much of the activity appears decentralized and uncoordinated. However, many attacks employ sophisticated technology, and experts believe that Chinese military and intelligence agencies either sponsor or condone technical attacks on political targets both within and outside China.

In December 2022, the Canadian branch of Amnesty International said it had been targeted by cyberattack believed to be sponsored by the Chinese government. The attackers reportedly attempted to gain information on the organization’s work on China.3

Mass surveillance also leaves people’s personal information vulnerable to breaches. In July 2022, hackers claimed to have accessed a Shanghai police database that contained records associated with one billion Chinese citizens. The database reportedly contained information including names, addresses, national ID numbers, and criminal records.4

Websites associated with Uyghurs have faced frequent cyberattacks, while China-based hackers have deployed various phishing campaigns to infiltrate these communities in exile and collect information about them. A November 2022 report by the cloud security firm Lookout revealed that a malware campaign linked to the Chinese government targeted Uyghurs inside and outside of China. Malware disguised as fully functional Android apps was designed to track people who use VPNs or share religious content.5

On China

See all data, scores & information on this country or territory.

See More
  • Global Freedom Score

    9 100 not free
  • Internet Freedom Score

    9 100 not free
  • Freedom in the World Status

    Not Free
  • Networks Restricted

    Yes
  • Websites Blocked

    Yes
  • Pro-government Commentators

    Yes
  • Users Arrested

    Yes