Free speech is a practice with preconditions, and our adversaries have learned to dismantle them. Defending those conditions is the most liberal thing a democracy can do.
In the fall of 2023, Israeli television broadcast footage of hostages taken by Hamas on October 7. The videos showed emaciated, desperate, terrorized captives speaking directly to camera, pleading with the Israeli government to end the war at any cost and bring them home. The footage was produced by Hamas and aired by Israeli broadcasters who, operating under the principles of a free press, saw fit to deliver it into the living rooms of the very society it was designed to break. Meanwhile, as the world accused Israel of genocide, Israeli media outlets were busy publishing claims that made the case. Amplified globally and cited in campaigns to isolate the Israeli state, the libels did their job. Hamas did not need to storm a television studio or a news office. It needed only to trust that a liberal democracy’s free press would do the rest.
This is the paradox at the heart of modern irregular warfare. Liberal democracies are built on freedoms that function as both their greatest strength and their Achilles’ heel. Non-liberal actors, from state sponsors like Iran and Russia to organizations like Hamas and ISIS, have learned that the decisive battlefield is cognitive and the most powerful weapons are narratives. Turning a democracy’s institutions against it costs far less than defeating its armies, and works far better. The response from democratic governments has been halting and contradictory, because the choice looks impossible. To restrict, regulate, or censor is to risk betraying the very values the state claims to defend. To do nothing is to cede the information environment to actors who operate without equivalent restraint.
The dilemma is real, but it rests on a misunderstanding of what liberal norms actually require. Freedom of speech is a practice rather than a self-sustaining absolute, and like all practices it depends on conditions that must be actively maintained. When malevolent actors set out to destroy those conditions, the principled response is to defend them. Targeted, legally constrained censorship, civil rather than criminal, subject to adversarial review, limited to actors exploiting liberal openness for illiberal ends, is not a betrayal of liberalism. It is what liberalism, honestly understood, demands.
A Practiced Inheritance
The phrase “honestly understood” is doing real work, so it deserves unpacking. What matters here is the relationship between liberalism as a theory and liberalism as a practice. Liberal theory, associated principally with Locke, Mill, and Rawls, holds that individual liberty is foundational and near-absolute. Mill’s harm principle permits restrictions on speech and conduct only where they cause direct harm to others. The logical conclusion is a strong presumption against censorship in almost any form.
Liberal practice tells a different story. Liberal democracies have never governed themselves according to the strict dictates of liberal theory. The most striking example is conscription. The military draft is about as direct a violation of personal autonomy as a government can impose, compelling citizens to risk their lives in the state’s defense regardless of their individual will. Yet liberal democracies have used it without apology, and most retain the legal architecture to do so again. Liberal practice not only permits conscription but, in extremis, demands it, precisely on the grounds that the survival of the liberal order is a precondition for the enjoyment of liberal freedoms. Tocqueville saw this clearly. The love of freedom that animates liberal societies was, in his account, a practiced inheritance: embedded in institutions, habits, and shared mores that require active maintenance and, when necessary, defense.
When commentators invoke liberal principles to oppose any restriction on the information environment, they are reasoning from theory alone, mistaking the philosophical scaffolding for the building itself. The question worth asking is whether censorship is compatible with liberalism honestly understood: theory and practice together.
The Marketplace and Its Myths
The most powerful objection to any form of speech regulation is less legal than philosophical. It is the argument, associated most influentially with Mill and given constitutional expression in Justice Oliver Wendell Holmes’s dissent in Abrams v. United States, that the remedy for bad speech is more speech. “The best test of truth,” Holmes wrote, “is the power of the thought to get itself accepted in the competition of the market.” The marketplace of ideas, on this view, is self-correcting. The solution to propaganda is counter-propaganda; the solution to a hostile narrative is a better one. This is a serious argument, and it deserves a serious response. It receives three here.
The first is what internet lore calls Brandolini’s Law: the effort required to refute a piece of misinformation is an order of magnitude greater than the effort required to produce it. A false casualty claim or a hostage video stripped of its coercive context can be disseminated across platforms in seconds. Refuting it requires sources, evidence, expertise, and content that is by its nature less emotionally immediate than the original, for the truth is often more complex. The marketplace structurally advantages the producer of disinformation over the producer of correction. A lie travels halfway around the world before the truth puts on its shoes.
The second objection cuts deeper. The marketplace assumes that human beings are, at baseline, truth-seeking creatures. The assumption is ahistorical. As Yuval Noah Harari argues in Sapiens, the distinguishing feature of human cognition is its orientation toward narrative: we are the only species capable of organizing large numbers of strangers around shared stories. People gravitate toward accounts that are emotionally resonant, socially reinforcing, and identity-confirming, and a well-crafted narrative will consistently outcompete an accurate but emotionally flat one. This is less a failure of intelligence than a feature of the species. Narrative is the native language of human cognition, and it is precisely what adversarial information operations exploit. Hamas does not produce propaganda because it cannot make arguments. It produces propaganda because propaganda works on human beings in ways that facts alone do not.
The third objection concerns the structure of the forum itself. Algorithmic platforms distribute content according to engagement metrics that systematically reward emotional provocation, outrage, and conflict, precisely the qualities that characterize adversarial information operations. The result looks less like a marketplace and more like a rigged auction. Imagine a debate in which one side is given a microphone, thirty minutes, and a stadium full of listeners, and the other is given no microphone, thirty seconds, and an empty room. No serious person would call that a fair exchange of ideas. Yet that is the structural condition of the contemporary information environment. A true market requires equal access. Today’s information environment, systematically tilted toward those willing to exploit it, offers none.
What Free Speech Presupposes
Mill himself came close to acknowledging the limits. In On Liberty, he confines liberty of expression to “human beings in the maturity of their faculties,” sets aside “those backward states of society in which the race itself may be considered as in its nonage,” and allows that “despotism is a legitimate mode of government in dealing with barbarians,” provided it improves them. The colonial prejudice embedded in the passage is obvious and should not be rehabilitated. But the underlying principle survives the discrediting of its particular application. What Mill was identifying is that freedom of speech presupposes certain social and epistemic capacities in those who exercise it; it is a practice with preconditions. The preconditions he identified were wrong. The insight that preconditions exist was not. Three matter most.
The first is constrained pluralism. A healthy marketplace of ideas requires competition between views that, however divergent, share a common end. Two bears fighting over territory are not preying on each other. The contest is fierce, but both are the same kind of creature playing by the same rules, and the population as a whole is strengthened by the selection. A wolf and a deer are a different matter entirely: one’s survival requires the other’s death. The same logic applies to political discourse. American conservatives and liberals disagree sharply about how to address crime, but both want safer communities, a shared end that makes deliberation possible. Hamas operates in a different register altogether. Its foundational objective is the destruction of Israel as a state and a people, foreclosing deliberation from the outset. Extending full participatory rights to actors of that kind is ecological suicide masquerading as pluralism.
The second precondition is good faith. Deliberation rests on the premise that subjecting ideas to rigorous scrutiny produces better outcomes than any single participant could reach alone: iron sharpens iron. Plato understood the alternative well. In the Gorgias, Socrates observes that the rhetorician need not know the truth about things; he need only persuade the ignorant that he knows more than those who do. Sophistry is the rhetorical nuke, available to any trained actor willing to abandon the rules, and devastatingly effective precisely because it works on human cognition in ways that bona fide argument does not. Deliberation therefore depends on a kind of mutually assured destruction: all sides refraining from emotional manipulation and coordinated deception, because deploying those tactics would destroy the purpose of the exchange. There are, in the end, two kinds of actors in any deliberative arena: those who seek to edge closer to truth, and those who enter to win by any means necessary. The liberal tradition extended free speech protections on the assumption of the former. It has no principled answer to the latter.
The third precondition is a suitable forum. The Athenian agora, the early American town hall, even the twentieth-century newspaper had structural features that, however imperfectly, constrained the worst tendencies of public discourse: physical proximity, editorial gatekeeping, reputational accountability, temporal pacing. The contemporary social media environment has none of them. It is anonymous, instantaneous, algorithmically amplified, and globally accessible, a forum designed as a matter of structural incentive to maximize emotional engagement rather than rational deliberation. Applying free speech norms developed for the town hall to the social media platform is a category error. The forum has changed; the norms have not caught up.
Karl Popper named the stakes in The Open Society and Its Enemies: tolerance, claimed in its own name, includes the right not to tolerate the intolerant. Tolerance, like free speech, is a practice: a living arrangement sustained by the mutual commitment of its participants to preserving the conditions that make it possible. When an actor enters the liberal order to destroy rather than to participate, tolerance straps a suicide vest onto the very principles it claims to uphold. Defensive liberalism, the position this essay defends, is the hand that cuts the wire.
The View from Israel
The Israel-Hamas war is the sharpest contemporary case of what happens when liberal openness meets irregular warfare and loses. Two episodes illustrate the mechanism.
The first is the hostage videos. In the months following October 7, Hamas released a stream of videos of hostages taken during the attack. The pattern was consistent: captive men, women, and the elderly speaking directly to camera, pleading with the Israeli government to end the war, calling on the Israeli public to march in the streets, urging acceptance of ceasefire terms regardless of their military implications. Each statement was a scripted performance, extracted from people held at gunpoint by an organization that had no reservation pulling the trigger.
As Noga Halevi of the Washington Institute for Near East Policy has documented, Hamas conceived of hostage-taking as an instrument of cognitive warfare: a means of paralyzing Israeli leadership, fracturing public unity through emotional attrition, and manipulating international narratives by weaponizing empathy. Beyond their value as bargaining chips, the hostages served as broadcast assets, released at politically sensitive moments to deepen internal divisions. Halevi is explicit that the fragmentation was the objective, not a by-product.
The perversity of the mechanism is easy to overlook. A hostage appears on camera. She is alive. She is speaking. She is imploring Israel to stop the war. What the viewer does not see is the gun. What the viewer does not hear is the coercion that preceded the filming. Laundered as testimony, the video is a weapon produced by Hamas and aimed directly at the Israeli public’s emotional capacity to sustain support for military operations.
Israeli news channels, including Kan 11, the national public broadcaster, aired the videos repeatedly, often without adequate contextualization of their coercive origin. The decision was legally unimpeachable. A free press has the right to broadcast newsworthy footage. But the effect was to deliver Hamas’s psychological operation directly into the Israeli information environment, with results that were predictable and, from Hamas’s perspective, calibrated: mass street protests, sustained pressure to accept strategically catastrophic terms, a progressive erosion of public consensus. Gil Merom argued in How Democracies Lose Small Wars that liberal states lose irregular conflicts in the domestic public sphere, where insurgents convert moral sensitivity and public opinion into pressure on the state. The war bore this out in real time. Hamas could lose every battle and still win, so long as Israel lost the will to keep fighting.
The second episode involves domestic media. In August 2025, the Israeli outlet +972 Magazine published an investigation claiming that 83 percent of Palestinians killed in Gaza were civilians. The claim rested on a classified military intelligence database showing that, as of May 2025, the IDF had identified 8,900 Hamas and Palestinian Islamic Jihad operatives killed by name. With the Gaza Health Ministry reporting roughly 53,000 total deaths at the time, +972 concluded that the remainder must be civilians.
The flaw requires no expertise in military affairs to identify. The database tracked only named, pre-identified operatives. It explicitly excluded fighters killed but not identifiable by name, and members of armed factions outside Hamas and Islamic Jihad. Treating a partial dataset of pre-catalogued operatives as exhaustive of all combatant deaths, and assigning every unaccounted death to the civilian category, is a logical fallacy too elementary to have been an honest mistake.
The downstream effect was considerable. Within seventy-two hours the figure had accumulated an estimated 6.6 billion social media impressions, carried by the credibility chain of liberal media institutions with no mechanism for distinguishing a fraudulent factual claim from legitimate investigative reporting. On September 16, 2025, the UN Human Rights Council’s Independent Commission of Inquiry cited the statistic in its genocide determination against Israel and recommended that member states cease transferring arms. On October 8, Spain’s Congress of Deputies ratified a total arms embargo on Israel. The legal scaffolding for that vote ran directly through the Commission report.
+972 operates inside Israel with full legal protection, under the same press freedom norms that protect every Israeli journalist. That is precisely the point. The liberal framework that permitted the publication could not distinguish between journalism and the strategic deployment of a fraudulent claim against a democracy at war.
Defamation, Not Wartime Censorship
What, then, should liberal democracies do? The spectrum of answers runs from libertarian absolutism, which holds that any filtering mechanism is more dangerous than the threat it addresses, to the authoritarian model of Viktor Orbán’s Hungary and Vladimir Putin’s Russia, where the language of information security becomes a license to consolidate media ownership and criminalize independent journalism. The first stands paralyzed before information warfare; by refusing any filter on principle, it leaves the environment open to systematic exploitation and so endangers the very continuity of free speech it claims to protect. The second is illiberal in kind, aimed at eliminating competition altogether rather than protecting the conditions for it. Defensive liberalism occupies the principled middle, and its model already exists within liberal law: defamation.
Defamation law establishes that freedom of expression does not extend to the deliberate or reckless publication of false statements of fact that cause demonstrable harm. Crucially, it is civil rather than criminal: the remedy is damages, the burden of proof rests on the claimant, and a judge, not the government, makes the determination. Defensive liberalism applies the same logic to information warfare through two mechanisms.
First, legislatures should define adversarial information operations narrowly and precisely: the involvement of or collusion with a foreign or non-liberal actor, the deliberate deployment of demonstrably false content, and the specific intent to undermine democratic decision-making or military effectiveness. The definition must not extend to criticism of government policy, dissenting journalism, or the publication of inconvenient truths. Jurisdiction should rest with courts, under a high evidentiary burden and full adversarial process. Court supervision rather than executive discretion is the structural safeguard that distinguishes defensive liberalism from what Hungary and Turkey practice.
Second, a civil liability framework should reach domestic outlets and information actors who knowingly or recklessly serve as vectors for such operations. The framework targets veracity, not viewpoint, leaving opinion, analysis, and political commentary entirely untouched. A journalist who argues that Israel’s military campaign is immoral is expressing a view; a journalist who publishes a statistic derived from a methodology they know to be fraudulent is making a false factual claim. The framework reaches only the latter.
The Slope Has Already Been Traversed
The strongest objection holds that targeted censorship, however well intentioned, creates the legal infrastructure for broader suppression: exceptions harden into rules, emergency powers become permanent. The concern is real and should not be waved away. But carried to its logical conclusion, it proves far too much. The power to arrest can be used to imprison political opponents. The power to prosecute can be weaponized against dissidents. Tax authority can be deployed selectively to punish disfavored groups. None of this has led serious people to conclude that governments should have no police, no prosecutors, no tax authority. The real challenge of democratic governance is not whether to grant power but how to grant it responsibly: with what constraints, what oversight, what accountability. To prohibit speech regulation entirely because it could be abused is to apply to this one domain a standard that no functioning democracy applies to any other.
The historical record is also more mixed than absolutists imply. The expansive First Amendment protections Americans know today are largely a product of post-1960s jurisprudence. For most of its history, the United States operated under considerably more restrictive speech laws, and the republic survived. Germany has prohibited Holocaust denial since the post-war period; France bans terrorist recruitment content. Democracies that maintain such constraints have not eliminated free speech. They have, at most, marked its edges.
Absolutists also display a curious blind spot. They worry, rightly, about censorship as an instrument of authoritarian consolidation, while overlooking the inverse: disinformation as an instrument of authoritarian ascent. Authoritarians flood the information environment with propaganda, scapegoating, and fabricated grievance, and impose formal censorship only after power is consolidated. Mussolini built his movement through Il Popolo d’Italia and the mythmaking of Italy’s open liberal press; formal censorship came after the march on Rome. Rodrigo Duterte rode Facebook’s algorithmic openness to the Philippine presidency before dismantling the country’s free press. In each case, the information environment was the first battlefield.
And the slope, in any case, has already been traversed, from both directions, without any of the safeguards proposed here. In 2021, senior White House officials pressured Facebook to remove COVID-19 content, including satire. The FBI’s warnings to platforms ahead of the 2020 election helped suppress reporting on the Hunter Biden laptop, a story that turned out to be authentic. From the other side of the aisle, Attorney General Pam Bondi declared in September 2025 that the Justice Department would “absolutely target you, go after you, if you are targeting anyone with hate speech.” The real choice is between formal, transparent, judicially supervised frameworks with evidentiary standards and adversarial review, and the ad hoc executive pressure that currently passes for policy. The alternative to a principled framework is not the absence of censorship. It is censorship by whoever happens to hold executive power, answerable to no one.
Why Education Is Not Enough
The gentler objection holds that the true answer to propaganda is education: media literacy, critical thinking, the patient cultivation of discerning citizens. But on the available evidence it is insufficient. Education is a generational project requiring sustained investment and long lead times. Adversarial information operations work in real time. The fraudulent casualty statistic reached billions of people within hours of publication. No curriculum responds at that speed.
The deeper problem is that the education argument assumes people fall for misinformation because they lack analytical tools. The evidence inverts the assumption. Dan Kahan’s research on motivated numeracy found that subjects with the strongest quantitative skills showed greater polarization on politically charged problems, deploying their abilities to construct sophisticated defenses of what they already believed. A 2024 meta-analysis in the Proceedings of the National Academy of Sciences found the same pattern at scale: higher analytical ability predicted greater susceptibility to partisan bias, a phenomenon the researchers call motivated reflection. Education does not produce more open-minded citizens on contested political questions. It produces citizens who are better at defending closed minds. Adversarial propaganda targets politically charged beliefs, about civilian casualties, military conduct, the legitimacy of the state, precisely because those beliefs are the most resistant to factual correction. Education matters, and nothing in this argument diminishes it. But it is a complement to a legal framework, and democracies have spent decades trying to get by on the complement alone.
A Failure of Nerve
The argument advanced here is paradoxical only on its surface. Censoring in the name of free speech sounds absurd, until one recognizes that free speech has never been an unconditional absolute, that liberal democracies have always maintained a protective shell around their open societies, and that the shell has been allowed to atrophy just as adversaries have learned to exploit what lies beneath it. Defensive liberalism is the recognition that liberalism, honestly practiced rather than merely theorized, has always required the willingness to defend itself.
The chasm between the promise of free speech and its observable reality can only grow so wide before the populace ceases to defend it with much conviction. When free speech produces a chronically misinformed electorate rather than an enlightened one; when a single fraudulent statistic accumulates billions of impressions while its correction reaches almost no one; when hostile actors use domestic broadcasters to deliver coerced propaganda into our living rooms, indiscriminate censorship stops sounding like such a bad idea. That is the strongest argument for a principled, court-supervised alternative: doing nothing produces the demand for something far worse.
