The Future of Free Speech

In the age of AI, who will be held accountable for suppressing speech when bots are doing the censorship?

Freedom of speech is an integral aspect of authentic freedom, grounded in the distinctively human capacity to adhere to truth through persuasion rather than imposition. While truth is objective and independent of the knowing subject, it can only be received by the subject freely. We can contrast this with a thinly grounded notion of freedom as merely a pragmatic expedient to “get along” in a pluralistic society in which we assume that truth is more or less unattainable. In this thin form of liberalism, when it no longer appears the optimal empirical solution to maintain social comity or stability, freedom of speech will be easily abandoned. We have seen this dynamic play out in Western societies in the past decade, where elites now frequently insist that censorship is necessary to “protect democracy” from the populist rabble.

Principled freedom that encourages the shared discovery of truth must always exclude violence against ideas. Such force reduces individuals to mere means and undermines the possibility of honest persuasion. Instead, public authorities should only intervene to prevent persecution or the propaganda of “truncheon arguments”—that is, various forms of coercion or incitement to violence—while permitting the open expression of opposing views, however controversial.

The late Italian Catholic philosopher Augusto Del Noce illustrates this by insisting that even ideological adversaries, such as Fascists, should be permitted to voice their bad ideas (which they were not permitted to do in Italy immediately following World War II): “Let the Fascists also support their idea of violence if they so wish; we will fight them with arguments, and the task will not really be too difficult,” Del Noce confidently asserted. He warned that forcefully suppressing false speech only breeds underground myths and perpetuates defeated ideologies, as seen in the posthumous “Mussolini legend” that arose in the climate of postwar Italian censorship. Among the central problems with censorship is this: the prohibition on pronouncing certain things often makes them seem true even when they are not. Better to allow falsehoods to be voiced and refute them with arguments than to forcefully silence them before they can be spoken.

---

Recovery of enduring principles on which to ground freedom of speech is essential as we face new challenges from the advent of enormously powerful digital technologies and global platforms of communication. Discovery in our Missouri v. Biden case revealed extensive coercion of digital platforms by the White House, the Surgeon General, the CDC, and other federal agencies to flag, suppress, deboost, or remove online content and accounts—via direct pressure, algorithmic changes affecting thousands of posts, or the actions of suborned third parties. This “censorship-industrial complex” targeted critics of the government’s preferred covid policies, election integrity, abortion, gender ideology, foreign policy, and other issues of public importance.

As the appellate court noted in our case, given the vast reach of new digital technologies, the scope and scale of this censorship had no precedent in prior free speech cases in the U.S. In the words of the district court judge, what our case uncovered was arguably the “worst violation of free speech in United States history.” Previous cases of censorship by overzealous government officials typically dealt with one-time offenses: a single article suppressed, a few paragraphs struck from a book, a journalistic story killed before it saw the light of day. But what we saw in our age of social media was entirely new, facilitated by the powerful reach of advanced digital technologies: government action censoring tens of thousands of ordinary Americans hundreds of thousands of times.

For many Americans like me who were on the receiving end of this, such actions felt very personal: every citizen with a Twitter or Facebook account became a potential “newspaper editor” or “journalist” targeted for suppression by government speech police. Often this censorship was done by identifiable political actors, as when the White House’s director of digital communications Rob Flaherty browbeat a senior executive at Facebook into submission, as I described in detail in a Wall Street Journal piece co-authored with one of our lawyers in the case. In that instance, documents we obtained on discovery implicated an identifiable government official who could be held accountable and brought before the bar of justice.

However, much of the government censorship was far more subtle and indirect. More often, it did not require a detectable person flagging a specific post or account for censure. In the U.S., the First Amendment bars government from engaging in viewpoint-based censorship, while he state-action doctrine bars government from circumventing constitutional strictures by suborning private companies to accomplish forbidden ends indirectly, which also occurred. Our case revealed that in some circumstances, government pressure resulted in social media platform’s changing their algorithms and terms of service in order to automate and conceal the process of government-sponsored censorship.

In this regime, those silenced may not even realize that they are being censored, as the algorithm simply limits the reach of their posts such that it is impossible for them to spread organically or “go viral” and accrue influence. In the new context of digitally mediated censorship, subtle algorithmic manipulation can create an illusion of free speech while effectively controlling the reach and impact of communications online.

If an investigator or journalist asks X or Facebook to see its algorithm, the company will not be able to produce it, for the coding has been outsourced to thousands of programming consultants, each of whom are responsible for only one small piece of the entire puzzle. This produces opaque means of information control with no clear locus of moral agency or accountability. Censorship goes underground, where it is governed by machines, which, unlike human officials, cannot be challenged in court or held to account.

The future of censorship will involve outsourcing most of this work to AI. With large language models now capable of digesting and synthesizing enormous swathes of online speech, our ability to communicate in the digital realm will soon be controlled by invisible yet immensely powerful bots. Even if government actors are prevented from censoring online, and even if we beat back the egregious EU Digital Services Act and other authoritarian mechanisms of state-sponsored censorship, private companies with vested ideological interests, armed with the tools of advanced AI, will remain a serious hazard to free speech. This global threat from powerful private actors may prove to be every bit as severe, and considerably more powerful, than government-sponsored censorship. (For reasons that go beyond the scope of this article, my own view is that under U.S. law these platforms should be treated as common carriers rather than editors of content, which would mean that private companies or their algorithms could not suppress constitutionally protected forms of speech.)

We are witnessing emerge in real time this pervasive and tenacious machinery of power, propaganda, and technological control. The response will require vigilance, legal precedents, and structural reforms to allow for open online discourse in our digital future.

---

This piece is adapted from my Foreword to the forthcoming volume, Truth, Freedom of Speech and Modern-day Censorship (En Route Books, 2026).

---

---