China

No complete nationwide political blackout was established during the assessment window. Cross-border connectivity nevertheless remained subject to continuous protocol interference at state-controlled gateways. On August 20, 2025, forged reset packets disrupted cross-border port 443 traffic for about 74 minutes; researchers could not determine whether the event resulted from testing, a configuration error, or another cause. [8][7][5]

Keeping domestic services online does not establish open connectivity when state-run gateways continuously interfere with access to the outside internet. The absence of a full blackout prevents a lower score, but persistent protocol disruption and opaque centralized control match the two-point anchor.

Network measurements show persistent nationwide blocking of major search, social media, messaging, news, human-rights, and circumvention services. Researchers also confirmed nationwide inspection and selective blocking of QUIC connections, while OONI's own site and measurement app remain blocked. [7][6][9]

This is persistent national filtering across large parts of the global internet, matching the zero-point anchor.

The Cybersecurity Law directs operators to stop transmission and remove prohibited information, including content arriving from abroad. CAC's 2025 campaigns instructed platforms to act against categories ranging from false information and harmful content to material said to stir group conflict or excessive pessimism. Individual restriction orders and complete totals are not routinely published. [11][13][14]

Major platforms are required to enforce state content priorities at scale, including political and broadly framed content. That fits the direct-control anchor.

Platforms must authenticate users, maintain moderation staff, inspect comments, withhold some news comments until approval, remove prohibited material, preserve records, and report violations. Failure can bring service suspension, license consequences, fines, and personal liability for responsible staff. [12][11][30]

The liability structure makes intermediaries operational participants in state censorship under serious regulatory threat.

Unapproved VPN and proxy services are routinely blocked, app distribution is restricted, and providers need state authorization for cross-border connectivity. Some technically capable users still reach the global internet through changing tools, but access is unstable and legal enforcement has included penalties for unauthorized services and, in some cases, users. [6][30][7]

Circumvention remains possible for a minority, preventing a zero. It is neither generally lawful nor reliably available in ordinary practice.

Article 35 of the Constitution provides for freedom of speech, press, assembly, and association. Those guarantees are not enforceable against national laws and administrative controls that authorize filtering, licensing, real-name rules, and removal of prohibited information. [10][11][27]

The zero anchor applies because the legal system establishes formal censorship and supplies no effective constitutional remedy against it. A written guarantee that cannot constrain those controls is not counted as practical protection.

Authorities continue to use offenses such as inciting subversion, picking quarrels and provoking trouble, defamation, and dissemination of false information in cases centered on reporting, criticism, or online advocacy. The terms leave wide discretion over what political and public-interest speech becomes criminal. [26][24][25]

The severe-liability anchor applies because peaceful reporting and criticism regularly support multi-year prison terms, rather than remaining a risk found only in statutory language.

CPJ counted 50 journalists imprisoned in China on December 1, 2025, the largest national total in its census. Citizen journalist Zhang Zhan received a second four-year sentence after a closed proceeding in September 2025, and anti-censorship blogger Ruan Xiaohuan remained in prison on a seven-year sentence. [23][24][25]

Repeated long imprisonment for journalism and online expression meets the methodology's most severe anchor.

Journalists, rights defenders, lawyers, and their families face device searches, account loss, police visits, travel restrictions, closed proceedings, and sustained digital and physical surveillance. Public documentation is incomplete because affected people and sources risk further punishment. [20][26][24]

State-linked pressure is pervasive among targeted groups, with weak independent investigation. The record is close to routine repression, though the available evidence does not show the same exposure for every ordinary user.

Residents use domestic platforms for consumer complaints, local grievances, mutual aid, and some policy discussion. Independent political organizing and journalism operate under much tighter limits; 2025 campaigns reached discussion of the economy, social conflict, and pessimistic views, encouraging broad self-censorship. [14][26][13]

Local complaints do not amount to an independent civic sphere. National political organizing and independent journalism can scarcely operate openly, matching the zero anchor.

The Personal Information Protection Law provides consent, access, correction, deletion, security, and sensitive-data protections, and authorities have enforced privacy rules against businesses. State organs receive separate treatment, supervision is divided among government departments, and independent research found extensive first-party tracking inside WeChat's integrated services. [17][16][18][19]

Consumer rights are meaningful but weakened by state exceptions, nonindependent supervision, and gaps between statutory rights and platform practice. This matches the partial-law anchor.

Public-security and national-security bodies have broad statutory access and technical-assistance powers. The available legal framework does not require independent prior judicial authorization comparable to a warrant for major intelligence collection, and most targeting, volume, and audit information remains secret. [11][20][30]

Internal executive procedures are not independent authorization or oversight. Expansive secret surveillance powers operate without a meaningful external legal check, matching the zero-point anchor.

SIM cards and major online services are tied to verified identity, and comment providers may not serve users who decline authentication. The national online identity service that took effect in July 2025 is formally voluntary and can reduce the amount of identity data given to individual platforms, but it operates within the existing real-name system and retains authentication records. [12][15][16]

The zero anchor applies because routine communications are comprehensively bound to verified identity and practical anonymity from the state is absent. Reduced disclosure to individual platforms does not alter that result.

Police systems combine communications, location, transactions, travel, biometrics, device forensics, and camera data, with especially intensive use against ethnic minorities and political targets. A 2025 investigation based on internal documents traced predictive-policing systems, while rights monitors documented phone inspections and arrests tied to online activity in Tibetan areas. [20][21][22][18]

The use of intrusion and integrated surveillance systems against political and ethnic targets is institutional and recurring, meeting the systematic-use anchor. New safeguards for civilian facial recognition do not govern away that record.

Commercial cryptography is lawful and widely used for banking, commerce, and network security. Products and services remain subject to licensing, security review, and lawful-assistance duties; popular domestic messaging does not provide users with the same practical anonymity as end-to-end encrypted services, many of which are blocked. [31][7][15]

Commercial encryption prevents a zero, but users lack dependable access to independent strong encryption and anonymous communication. The record falls between the prohibition and severe-restriction anchors.

Courts hear large volumes of civil, commercial, and administrative cases, and a 2025 judicial interpretation clarified government-information litigation. The Supreme People's Court also states that courts must uphold the Communist Party's absolute leadership; sensitive speech cases may be closed and access to counsel or observers restricted. [28][27][24]

For political censorship and surveillance, courts do not operate as an independent check or provide a meaningful remedy. That is the zero-point anchor even though ordinary adjudication functions in other fields.

CAC, MIIT, public-security organs, and propaganda authorities regulate infrastructure, data, platforms, news, and content under Party and State Council direction. They issue and enforce censorship duties rather than independently reviewing them. [11][13][27]

Regulators directly administer the control system, matching the zero-point anchor.

Authorities publish laws, draft rules, campaign priorities, enforcement examples, connectivity statistics, and some court decisions. They do not publish a complete list of blocked resources, censorship directives, removal totals by legal basis, surveillance volumes, or auditable data on challenges and remedies. [1][13][11][6]

Aggregate disclosure prevents a zero, but the decisions most relevant to political censorship and surveillance remain secret. The record falls below the one-point anchor for reviewable operational transparency.

Platforms maintain complaint channels, privacy rights can support civil claims, and government-information refusals may be challenged. Network blocks commonly have no public notice, political account restrictions often give only broad policy reasons, and neither CAC nor the courts supply an independent route to contest national censorship policy. [17][28][14]

Consumer remedies do not provide an appeal against national censorship or surveillance. For the restrictions measured by this index, secret decisions provide no effective independent remedy, matching the zero anchor.

Agencies release some draft digital rules for public comment and consult firms and technical bodies. The final online identity rules preserved voluntary use and data-minimization provisions. Final authority remains concentrated within Party-state institutions, and agencies do not routinely publish all comments or a point-by-point response. [15][16][26]

Consultation can affect safeguards, which warrants more than the formalistic anchor. Participation is still controlled and independent rights groups cannot engage on equal terms.