Free Speech and the Internet

The internet's early architects and advocates believed they were building a medium that was structurally hostile to censorship. The network's distributed architecture — designed for survivability in a nuclear attack — meant that information could route around any single point of failure. Anyone with a computer and a modem could become a publisher, reaching any other connected user anywhere in the world. The cost of publishing fell from the price of a printing press to the price of an internet connection. For the first time in history, the gatekeeper function that newspapers, broadcasters, and publishers performed — deciding what speech reached mass audiences — could be bypassed entirely.

EFF founder John Perry Barlow's 1996 'Declaration of the Independence of Cyberspace' articulated the utopian vision at its most extreme: 'Governments of the Industrial World, you weary giants of flesh and steel, I come from Cyberspace, the new home of Mind. On behalf of the future, I ask you of the past to leave us alone.' The declaration was extravagant rhetoric rather than legal analysis, but it captured the genuine sense among early internet enthusiasts that the network's architecture rendered traditional content regulation obsolete.

This early promise was partially realized. The internet did dramatically democratize publishing and enabled forms of political organization, citizen journalism, and cross-border communication that were previously impossible. Dissident voices in authoritarian countries, whistleblowers inside powerful institutions, and marginalized communities that lacked access to mainstream media all benefited from the internet's low barriers to publication. The Arab Spring, the early Black Lives Matter organizing, and countless other social movements were shaped by the internet's capacity for rapid, distributed coordination.

Reno v. American Civil Liberties Union (1997) is the Supreme Court's foundational statement on internet speech and the First Amendment. The case challenged the Communications Decency Act's provisions making it a crime to post 'indecent' or 'patently offensive' content accessible to minors on the internet. The Clinton administration argued that the internet, like broadcast television, could be regulated for indecency to protect children. The ACLU and a coalition of internet and civil liberties organizations argued that the internet was more like print media — which receives full First Amendment protection — than like broadcasting.

Justice Stevens's majority opinion sided with the challengers. The internet, the Court held, 'constitutes a vast platform from which to address and hear from a worldwide audience of millions of readers, viewers, researchers, and buyers' and deserves full First Amendment protection rather than the reduced protection applicable to broadcasting. The Court distinguished broadcast: the internet is not subject to spectrum scarcity, does not 'invade' the home in the way that broadcast does, and does not have the history of government regulation that justified reduced protection for broadcasting under Pacifica.

Reno established the constitutional baseline that has governed internet regulation in the United States ever since: internet speech receives the same protection as print media, and content-based restrictions on internet speech must survive strict scrutiny. Combined with Section 230's immunity provisions enacted the year before, Reno created the legal environment that enabled the development of social media, online journalism, and user-generated content at the scale that defines the contemporary internet. Whether Reno's framework remains adequate for an internet characterized by platform concentration rather than the distributed publishing the Court imagined is an active debate.

The early internet's distributed architecture was followed by a period of extreme concentration. By the mid-2010s, a small number of platform companies — Facebook, Google, Twitter, Amazon — had become dominant intermediaries through which most internet users experienced the web. Search, social media, e-commerce, and cloud computing were each controlled by one or two companies with near-monopoly positions. The distributed publishing vision of the early internet had given way to a centralized platform economy in which most user expression, most content discovery, and most communication occurred within the walled gardens of a handful of private companies.

This concentration had significant consequences for free expression. The 1990s and early 2000s internet featured thousands of competing forums, news sites, blogs, and social spaces with diverse content policies, governance models, and community cultures. The platform economy collapsed this diversity: by 2015, most public online discourse occurred on Facebook and Twitter, most video consumption on YouTube, and most web discovery through Google. Each of these platforms made content moderation, algorithmic curation, and speech governance decisions affecting billions of users, with limited competition to discipline those decisions through user choice.

Network effects — the dynamic by which platforms become more valuable as more users join — created barriers to entry that prevented meaningful competitive challenge to dominant platforms. This dynamic means that user 'choice' to use alternative platforms has significant practical costs: moving from Facebook to a smaller competitor means losing access to one's existing social network. The speech governance decisions of dominant platforms are thus less subject to market discipline than the analogy to other market choices suggests.

Section 230 of the Communications Decency Act is arguably the most important law governing internet speech in the United States, though it is largely unknown to the general public. Its two core provisions — immunity from liability for user-generated content, and immunity for good-faith content moderation — created the legal foundation for the development of social media, user review platforms, and most of the modern internet as it is experienced today.

Section 230's immunity for user-generated content means that platforms are not liable for what their users post — they cannot be sued for defamation, harassment, or other speech-based claims arising from user content. Without this immunity, the risk of liability for billions of daily user posts would make hosting user-generated content economically unviable except for the largest companies with the most legal resources. The immunity enabled the development of everything from Facebook and Twitter to Yelp reviews and Wikipedia, by making it economically feasible to host enormous amounts of user content without being exposed to the legal liability that would follow from treating those decisions as editorial choices.

Section 230's critics argue that the immunity has enabled platforms to profit from harmful content — disinformation, harassment, radicalization — while escaping the accountability that liability would create. The absence of liability, in this view, means platforms have insufficient incentive to invest in content moderation. Proposed reforms range from narrowing the immunity for specific categories of content (CSAM, terrorism, health misinformation, sex trafficking) to conditioning the immunity on platforms meeting certain due diligence standards. The debate about Section 230 reform is one of the central free speech policy debates of the current era.

The early internet's cross-border architecture was premised on the idea that geographic boundaries were largely irrelevant to digital communication. This premise has proven incorrect. Governments around the world have developed legal and technical tools to control internet content within their borders, and the combination of legal pressure, technical filtering, and corporate compliance has produced a global internet that is significantly less free and less uniform than its architects imagined.

China's Great Firewall is the most sophisticated national internet filtering system, blocking access to most Western platforms and requiring domestic platforms to comply with political content restrictions enforced by a massive bureaucracy. Russia has passed 'sovereign internet' legislation enabling the government to isolate the Russian internet from the global network and has increasingly enforced content restrictions through fines and blocking orders against foreign platforms. Iran, Saudi Arabia, and dozens of other countries maintain extensive internet filtering systems. Even democratic countries impose content requirements on platforms operating within their jurisdictions: Germany requires removal of hate speech, the UK imposes age verification requirements, and France has ordered removal of right-wing extremist content.

Platform compliance with local content restrictions creates direct tension with internet freedom values. When Google removes search results in Germany, when Apple removes apps from its Chinese App Store at government request, and when Twitter/X complies with government-ordered account suspensions in Turkey or India, global platforms are implementing domestic censorship policies that affect all users in those jurisdictions. The alternative — refusing to comply and being blocked — eliminates platform access for users in those countries entirely. There is no clean resolution: global platform operation requires either accepting local content restrictions or losing access to users in restrictive jurisdictions.

Artificial intelligence is transforming both the internet's information environment and the tools available for controlling it, in ways that have profound implications for free expression. AI recommendation and content curation systems determine what information users encounter online — shaping the effective information environment even when users are technically capable of accessing anything. AI moderation systems make billions of speech-related decisions daily. AI-generated content is becoming an increasing fraction of online information. And AI surveillance tools enable governments and corporations to monitor online communication at scales previously impossible.

The deployment of large language models as primary information access tools represents a significant structural change in how people find information online. As users shift from search engines to AI assistants for many information tasks, the design choices embedded in those AI systems — about what information to surface, what perspectives to represent, what topics to decline to address — become central determinants of what information is effectively accessible. This is a form of information architecture with no close historical precedent, and its implications for expressive freedom are only beginning to be understood.

The future internet's relationship to free expression will be shaped by decisions that are being made now about AI governance, platform regulation, antitrust enforcement against dominant platforms, and international governance of cross-border speech. The Reno framework — treating the internet as a maximally protected speech environment analogous to print media — was developed for an internet that no longer exists, where expression was distributed and power was diffuse. Whether that framework remains adequate for an internet characterized by AI-mediated information curation, platform concentration, and state censorship infrastructure is one of the most important legal and policy questions of the current era.