History of Internet Censorship: From the CDA to Global Shutdowns

The history of internet censorship runs from early-1990s fights over obscenity on U.S. bulletin boards to a global system in which dozens of governments filter, throttle, or shut down networks each year, courts and regulators argue over what platforms may remove, and users route around the whole apparatus with encrypted tools. What follows traces that arc: the founding American cases and statutes, the authoritarian architectures built in China, North Korea, Russia, and Iran, the shutdown playbook popularized after the Arab Spring, and the newer disputes over platform moderation, Section 230, and the EU’s Digital Services Act.

The First Fight: The CDA and Reno v. ACLU

The debate began before the web went mainstream. Bulletin Board Systems in the 1980s and early 1990s were run by individual operators who set their own content rules, and by the mid-1990s a media panic over “cyberporn” was pushing Congress toward federal regulation.¹ Two contradictory rulings sharpened the pressure. In Cubby, Inc. v. CompuServe (1991), a court treated CompuServe as a passive distributor not liable for user posts. In Stratton Oakmont, Inc. v. Prodigy Services Co. (1995), a court held Prodigy liable precisely because it moderated its forums, creating a perverse incentive against cleanup.²

Congress responded with the Communications Decency Act, passed as part of the Telecommunications Act of 1996. The CDA made it a crime to transmit “obscene or indecent” material to anyone under 18 and outlawed the “knowing display” of “patently offensive” content in a manner accessible to minors, with fines and prison time attached.²

In Reno v. American Civil Liberties Union (1997), the Supreme Court struck down the indecency and “patently offensive” provisions as unconstitutionally overbroad. Justice John Paul Stevens wrote that the law’s terms were so sweeping they could criminalize protected discussions of health care or AIDS. The decision extended full First Amendment protection to online speech and rejected the argument that protecting children justified such broad suppression of adult expression.³²

One part of the CDA survived and reshaped the internet: Section 230, drafted to resolve the CompuServe–Prodigy paradox. It shields online services from liability for content posted by third parties, even when the service chooses to moderate. Every large social platform that followed was built on that foundation.²

How the U.S. Framework Filled In

The years after Reno brought a series of narrower laws that shaped what users could see and what governments could see about users.

The Children’s Online Privacy Protection Act of 1998 required websites directed at children under 13 to obtain verifiable parental consent before collecting personal information, which pushed many platforms toward blanket age restrictions.⁴ The Digital Millennium Copyright Act, also passed in 1998, criminalized circumvention of digital copyright protection and created the “notice and takedown” system that rights holders still use to strip content off the web. The Child Online Protection Act, a second attempt at restricting minors’ access to sexual material, was struck down as overbroad in 2004.³

The Children’s Internet Protection Act of 2000 conditioned federal funding for schools and public libraries on installing internet filters, and the Supreme Court upheld it in 2003 in U.S. v. American Library Association. A 2012 ruling in PFLAG v. Camdenton R-III School District later held that districts must monitor those filters to prevent viewpoint discrimination; the case involved software that selectively blocked LGBT content.³

After the September 11 attacks, the USA PATRIOT Act vastly expanded electronic surveillance powers, and the 2007 Protect America Act enabled the NSA’s PRISM program to pull internet communications from major technology companies.³ Under Section 702 of the Foreign Intelligence Surveillance Act, the government captures communications flowing in and out of the United States in bulk, and a “backdoor search loophole” allows agencies to query the resulting database for Americans’ communications without a warrant. The ACLU has argued in federal court that this mass surveillance chills speech, particularly among journalists, activists, and religious minorities.⁵

China’s Great Firewall

China operates the most sophisticated censorship infrastructure on earth. The system traces to 1998, when the Ministry of Public Security launched the “Golden Shield Project” to control information flow and monitor citizens’ internet activity. The result, known colloquially as the Great Firewall, functions as a virtual boundary that selectively separates Chinese cyberspace from the global internet.⁶

The technical methods stack together. Authorities use DNS poisoning to redirect queries for banned sites, deep packet inspection to analyze traffic, TCP reset attacks to sever specific connections, and keyword filtering to block searches on sensitive topics. Companies operating in China must self-censor, removing prohibited topics, obscenity, and anything that might incite political resistance or reveal state secrets.⁶

The blocklist is broad. Facebook, Instagram, Twitter, Discord, YouTube, Netflix, and major Western news outlets including the New York Times, the Washington Post, and the Economist are all inaccessible. In their place, domestic platforms — WeChat, Weibo, Baidu, Bilibili, Youku — serve more than a billion users.⁶ Unlike countries that filter at the platform level, China filters at the infrastructure level. All global traffic enters and exits through a limited number of terrestrial gateways, where it is inspected before reaching users.⁷

North Korea: Replacing the Internet

Where China filters the global internet, North Korea replaces it. The regime operates a domestic-only network called Kwangmyong (“Bright Star”), developed in 1997 by the Central Scientific and Technological Information Agency. It is entirely disconnected from the worldwide internet and limited to government-approved sites, email, and forums.⁸

Access to the actual global internet is reserved for a small number of trusted scientists, engineers, and security officials, routed through China and a Thai satellite partnership, and closely supervised.⁸ Ordinary citizens are denied access even to Kwangmyong for the most part; most have no home or mobile connection.⁹

Enforcement is layered. Bureau 27 of the State Security Department monitors phone calls and internet activity with specialized detection equipment. An inter-agency unit called Group 109 conducts physical inspections of digital media to seize foreign content on CDs, DVDs, and USB drives.⁸ Around 2014, the state embedded software in domestically used Android devices that automatically identifies and deletes any file lacking an official state signature. Even so, roughly one-third of surveyed defectors reported consuming foreign media — South Korean dramas, international news, K-pop — despite the risk of imprisonment.⁹

Russia’s Sovereign Internet

Russia’s censorship buildout has been gradual and continuous. The 2016 “Yarovaya Amendments” required telecom companies to retain communications metadata for three years and to hand the FSB encryption keys to decrypt user messages without a court order. Separate 2017 laws required companies processing Russian citizens’ data to store it on Russian soil, and LinkedIn was blocked in 2016 for noncompliance. VPNs and proxy services were prohibited from providing access to banned websites.¹⁰

The cornerstone is the 2019 “Sovereign Internet” law. It requires ISPs to install deep packet inspection equipment, officially called “Technical Measures to Combat Threats” (TSPU), that lets the state regulator Roskomnadzor block sites and filter traffic unilaterally. By November 2022, roughly 6,000 such devices were active on Russian networks.¹¹ The law also mandated a national DNS independent of global infrastructure, giving the government theoretical ability to cut Russia off from the worldwide internet entirely.¹⁰

The full-scale invasion of Ukraine in February 2022 accelerated the crackdown. Russia blocked Facebook, Twitter, and Instagram within weeks, and a Moscow court designated Meta an “extremist organization” in March 2022. A 2024 law banned information about and advertising for VPNs, and Roskomnadzor has systematically blocked VPN protocols. YouTube was reportedly throttled in August 2024, and Signal was blocked the same month. As of May 2024, at least 935 criminal cases had been opened for “discrediting” or spreading “false information” about the Russian military.¹¹

Iran and the National Intranet

Iran began exploring a domestically controlled network under President Ahmadinejad in 2005. After the 2009 Green Movement protests, the state required ISPs to receive government approval and filter content, and by 2013 had begun building a National Infrastructure Network to host essential services domestically, easing monitoring and filtering.¹²

The government has repeatedly used total shutdowns during unrest. A five-day nationwide shutdown accompanied the November 2019 fuel-price protests.¹³¹²

A “User Protection Bill” ratified by parliament in early 2022 consolidated control of internet infrastructure under the armed forces and security agencies. The paramilitary arm of the Islamic Revolutionary Guard Corps owns at least half the shares in the Telecommunications Company of Iran, and the state has expanded its ability to block encrypted DNS traffic and criminalize VPN use.¹³ In January 2026, Iran imposed a near-total nationwide blackout during renewed protests.¹⁴

The Arab Spring and the Shutdown Playbook

The 2010–2011 uprisings marked a turning point. In Tunisia, the Ben Ali government ran a filtering regime using DNS tampering, URL blocking, IP filtering, and keyword censorship.¹⁵¹⁶

The blackout didn’t stop the protests. Activists routed data through proxies abroad, a group called “We Rebuild” connected via dial-up modems in Sweden and published instructions for maintaining connectivity, and global Twitter activity about Egypt surged from 2,300 tweets per day to 230,000 in the week before Mubarak’s resignation. Some analysts have argued the shutdown itself pushed people who could no longer follow events online into the streets.¹⁷ The Electronic Frontier Foundation has said the period transformed how governments respond to unrest, prompting heavy investment in surveillance, legal control mechanisms, and automated censorship, and turning internet blackouts into a “normalized tool of crisis response.”¹⁵

Shutdowns Go Global

The normalization shows up in the counts. According to the Access Now and #KeepItOn coalition, at least 313 internet shutdowns were imposed across 52 countries in 2025, the highest total since tracking began in 2016.¹⁸ The Internet Society documented 133 shutdowns in 2024 and 124 in 2023, with India recording the highest cumulative total since 2019, followed by the Middle East and Sub-Saharan Africa.¹⁹

Governments cite national security, fake news, exam integrity, and public order, but the pattern is consistent: shutdowns cluster around elections, protests, and armed conflicts. Iraq alone has recorded 140 shutdown incidents since 2018, including a countrywide outage in June 2015 to prevent cheating on national exams.¹⁹²⁰

Ethiopia and Sudan

Ethiopia’s government imposed a near-total communications blackout across the Tigray region beginning in November 2020, cutting broadband, mobile internet, and phone lines. It lasted more than two years in parts of the region, one of the longest shutdowns on record. International investigations attributed it to the federal government despite denials and concluded it was intended to suppress information about the conflict and alleged war crimes.²¹

In Sudan, the military imposed a 25-day internet blackout after the October 2021 coup, cutting phone and SMS services before a planned nationwide protest. When a Khartoum court ordered the three main telecom providers to restore service, the Telecommunications and Post Regulatory Authority ignored the order, citing national security. Researchers documented a pattern of strategic shutdowns targeting pro-democracy protests while leaving the internet operational during a pro-military demonstration.²¹

Myanmar After the 2021 Coup

Myanmar’s February 1, 2021 coup came with an immediate internet shutdown; soldiers entered telecom offices to enforce it. In the days that followed, the junta blocked Facebook, WhatsApp, Twitter, and Instagram in sequence and imposed recurring nightly shutdowns between roughly 1:00 AM and 9:00 AM.²² The military later shifted from nationwide blocks to targeted shutdowns in areas of armed resistance, doubled data prices in 2021, imposed new SIM taxes, and pressured carriers (two of four have direct military ties) to activate surveillance technology and surrender user data. By 2024, the military was arresting individuals found with VPN apps on their phones.²³

Pakistan’s Escalation

In 2024, Pakistan deployed a new internet firewall using Chinese-sourced deep packet inspection technology at a reported cost of 20 to 30 billion rupees ($72 million to $107 million). The firewall’s “in-line” architecture lets authorities block or throttle specific features within apps, such as WhatsApp’s multimedia functions, rather than shut down entire platforms.²⁴ The platform X was restricted from February 2024 through May 2025, and a near-total internet blackout occurred in November 2024 ahead of opposition protests. The Pakistan Software Houses Association estimated the disruptions could cost the IT sector $300 million.²⁵

India: Blocking at Scale

India is a democracy that leads the world in localized shutdowns and maintains a largely secret system for ordering websites blocked. The legal hook is Section 69A of the Information Technology Act of 2000, which lets the central government block content in the interest of sovereignty, defense, state security, public order, or friendly relations. Between 2015 and 2022, more than 55,600 websites, URLs, apps, and social media posts were blocked, roughly split between executive orders from the Ministry of Electronics and Information Technology and court orders for copyright infringement.²⁶

Under the 2009 Blocking Rules, blocking requests and actions are subject to strict confidentiality, and the Ministry has invoked that provision to refuse even to tell content creators their material has been blocked.²⁷ In Shreya Singhal v. Union of India (2015), the Supreme Court struck down Section 66A of the IT Act, which had criminalized “grossly offensive” or “annoying” online messages, as unconstitutionally vague and open to “arbitrary and whimsical” enforcement. The court emphasized that “mere discussion or even advocacy of a particular cause howsoever unpopular” is protected.²⁸ Prosecuting agencies have continued to file cases under the scrapped law anyway, prompting the People’s Union for Civil Liberties to petition the Supreme Court in 2021 to stop further arrests under the defunct provision.²⁹

SOPA, PIPA, and the 2012 Blackout

Democratic pushback against censorship-adjacent legislation has its own landmark. In 2011, Congress considered the Stop Online Piracy Act in the House and the PROTECT IP Act in the Senate, framed as anti-piracy but written broadly enough that critics argued they would let copyright holders and courts order ISPs and search engines to block entire websites accused of facilitating infringement.³⁰

On January 18, 2012, the internet staged what has been called the largest online protest in history. Wikipedia went dark for 24 hours. Reddit, Boing Boing, Mozilla, WordPress, the Internet Archive, and more than 115,000 other sites published protest content or shut down entirely. Google used its homepage to highlight the issue.³¹³⁰ Members of Congress abandoned the bills in droves, and by January 20, votes on both were shelved. The Obama administration signaled it would not support legislation that reduced free expression or undermined innovation.

Is Platform Moderation Censorship?

As social media absorbed public discourse, the argument shifted. When Facebook removes a post, YouTube demonetizes a video, or X suspends an account, is that censorship?

Two states tested the question directly. Florida’s SB 7072 and Texas’s HB 20 both sought to prevent large platforms from removing or deprioritizing content based on viewpoint. A circuit split followed: the Eleventh Circuit found most of Florida’s law likely unconstitutional, holding that platforms exercise First Amendment-protected editorial judgment when they curate. A divided Fifth Circuit panel upheld the Texas law, with Judge Andrew Oldham characterizing platform moderation as unprotected “censorship” and suggesting platforms could be regulated as common carriers.³²

The Supreme Court took up the cases as Moody v. NetChoice and NetChoice v. Paxton and issued a decision on July 1, 2024. Justice Kagan wrote for the Court, which vacated both lower-court rulings and sent the cases back on the ground that neither circuit had properly analyzed the full scope of the laws. The Court affirmed that platforms engage in “expressive activity” when they curate and filter content and that the government cannot justify these laws by asserting an interest in “correcting” the mix of viewpoints on a platform.³³³⁴ The lower courts must still perform a full facial analysis, so the fate of the state laws is not yet settled.

Section 230 in the Crosshairs

Section 230 remains central to the American debate. Critics on the right argue it lets platforms suppress conservative viewpoints; critics on the left argue it shields platforms from accountability for harmful content. Both push for reform.

In Gonzalez v. Google (2023), the Supreme Court had a chance to decide whether Section 230 protects platforms’ algorithmic recommendations. It didn’t take it, issuing a brief per curiam opinion that resolved the case on narrower grounds under the Anti-Terrorism Act without reaching the Section 230 question.³⁵

Legislation keeps coming. In December 2025, Senator Lindsey Graham introduced the “Sunset Section 230 Act” (S.3546), which would repeal Section 230 entirely two years after enactment. The bill attracted a bipartisan group of cosponsors including Senators Durbin, Grassley, Hawley, Klobuchar, Blackburn, and Blumenthal, and was referred to the Senate Commerce Committee.³⁶ A companion House bill, the “Sunset To Reform Section 230 Act” (H.R.6746), was introduced in the 119th Congress.³⁷

The EU’s Digital Services Act

The European Union has staked out a third path between the U.S. free-speech framework and authoritarian filtering. The Digital Services Act, enacted in 2022, requires platforms to minimize exposure to illegal and harmful content while being transparent about how they moderate.³⁸

“Very large online platforms” with more than 45 million monthly EU users, a category that captures Google, Facebook, Amazon, X, and YouTube, face additional obligations to identify and mitigate systemic risks including threats to civic discourse, electoral integrity, and freedom of expression. Noncompliance can bring fines of up to 6% of global turnover.³⁹

American officials and tech executives have pushed back sharply. Vice President J.D. Vance has called EU content moderation “authoritarian censorship,” FCC Commissioner Brendan Carr has called it “incompatible with America’s free speech tradition,” and Meta CEO Mark Zuckerberg has accused the EU of “institutionalizing censorship.”³⁹ Legal scholars have flagged concerns about “collateral censorship,” the risk that platforms over-remove lawful speech to avoid fines, and a “de facto Brussels effect” in which platforms apply EU standards globally rather than run region-specific moderation. Proponents counter that the law does not require changes to what American users see, and that any global application is a business decision by the platforms.⁴⁰

Circumvention: The Arms Race

For every censorship technology, there’s a circumvention tool. The most widely used include VPNs, which mask a user’s IP address and encrypt traffic; the Tor network, which routes traffic through multiple volunteer relays; and specialized obfuscation protocols like obfs4, which randomizes packet characteristics to evade deep packet inspection, and Meek, which disguises traffic as communication with permitted cloud services such as Amazon or Google.⁴¹

Censors have responded with matching sophistication. Deep packet inspection can identify and block traffic patterns tied to known circumvention tools. “Active probing” attempts to unmask circumvention servers by pretending to be ordinary users. Some governments, like Russia during its 2018 attempt to block Telegram, have shown willingness to block millions of IP addresses belonging to major cloud providers, accepting widespread collateral damage. Belarus has officially banned Tor and proxy servers since 2015 and used DPI to disrupt access during political crises, yet during the 2020 presidential election the circumvention tool Psiphon reported 1.76 million users in the country, with protesters distributing access instructions via USB drives and QR codes.⁴¹ Russia, China, Iran, and Myanmar have all moved to restrict or criminalize VPN use with varying enforcement, and the pattern is consistent: legal prohibition slows circumvention but doesn’t end it.

Where Things Stand

The current landscape is defined by scale and precision. UNESCO has warned that 2024 was the worst year on record for internet shutdowns, with the trend continuing into 2026, particularly around demonstrations and elections.¹⁴ Authoritarian states have moved beyond crude network shutdowns to granular, feature-level throttling that is harder for users to detect and harder for outside observers to document. Democracies remain locked in debate over where content regulation ends and censorship begins, a question neither the Supreme Court nor EU enforcement has settled.

The economic costs keep rising. A single day of internet blackout in Bangladesh during the 2024 constitutional crisis cost an estimated $411,000, and a hypothetical one-day shutdown in Italy would cost an estimated $5 million. Because networks are interconnected, shutdowns spill across borders; a 2013 shutdown in Sudan knocked landlocked Chad offline.¹⁹ Nearly three decades after the Supreme Court first extended full First Amendment protection to the internet, the fight over who controls online speech is still open.

1
Issues in Science and Technology. Prehistory of Social Media
2
First Amendment Encyclopedia. Communications Decency Act and Section 230
3
NCAC. A Selective Timeline of the Internet and Censorship
4
ITIF. How to Address Children’s Online Safety in the United States
5
ACLU. Warrantless Surveillance Under Section 702 of FISA
6
Encyclopædia Britannica. Great Firewall
7
Internet Society. The Chinese Firewall
8
Reporters Without Borders. Enemies of the Internet Report – North Korea
9
CSIS. North Koreans Want External Information, Kim Jong-un Seeks to Limit Access
10
Human Rights Watch. Russia: Growing Internet Isolation, Control, Censorship
11
Freedom House. Russia: Freedom on the Net
12
Cloudflare. Internet Use in Iran During the Mahsa Amini Protests
13
Middle East Institute. Mahsa Amini and the Future of Internet Repression in Iran
14
UN Geneva. UN Warns of Rising Internet Shutdowns
15
Electronic Frontier Foundation. Digital Hopes, Real Power: Reflecting on the Legacy of the Arab Spring
16
Brookings Institution. The Economic Costs of Internet Shutdowns
17
Al Jazeera. Arab Spring Anniversary: When Egypt Cut the Internet
18
Access Now. #KeepItOn Campaign
19
Internet Society. Internet Shutdowns Policy Brief
20
ThousandEyes. Internet Censorship Around the World
21
Access Now. Evading Accountability Through Internet Shutdowns
22
OONI. Myanmar Internet Blocks and Outages
23
Freedom House. Myanmar: Freedom on the Net
24
Al Jazeera. Pakistan Tests China-Like Digital Firewall
25
Index on Censorship. Pakistan Faces Increasing Internet Censorship
26
SFLC.in. Finding 404: Report on Website Blocking in India
27
Internet Freedom Foundation. Publish Blocking Orders
28
Indian Kanoon. Shreya Singhal v. Union of India
29
Supreme Court Observer. Section 66A: Must-Reads
30
The Guardian. SOPA Blackout Protest Makes History
31
ABC News. Wikipedia Blackout: Websites Go Dark in Protest
32
FIRE. Free Speech and Social Media
33
SCOTUSblog. Moody v. NetChoice, LLC
34
Supreme Court of the United States. Moody v. NetChoice, LLC, Opinion
35
First Amendment Encyclopedia. Gonzalez v. Google
36
U.S. Senate. Graham Leads Bill to Sunset Section 230 Immunity
37
U.S. Congress. H.R.6746 – Sunset To Reform Section 230 Act
38
European Commission. Digital Services Act
39
CSIS. Does the EU’s Digital Services Act Violate Freedom of Speech
40
Stanford Law School. A Primer on Cross-Border Speech Regulation and the EU’s Digital Services Act
41
Georgetown Law Technology Review. Censorship Circumvention Tools