What does AI have to do with the First Amendment? 

1983 was an exhilarating year for technology. The modern internet was created. Microsoft Word was launched. The new Mac had its first graphical user interface. In keeping with that visionary period, a scholar of technology and history at MIT named Ithiel de Sola Pool wrote of emerging technology’s great promise for freedom and liberty. 

But he also had a grave worry. 

During World War II, Pool had studied how Nazi and Soviet regimes exploited information and the communications technologies of the day. Fascinated, he spent the rest of his career examining how information technology, societies, and governments interact — and he saw something in the 20th century that pointed to a disturbing forecast for free expression.

He noticed that free speech protections had developed around a particular set of media: “pamphlets, platforms, and periodicals.” But the communications landscape was changing. The traditional media around which free speech protections had developed were being “relegated . . . to a corner of the public forum,” while electronic forms of communication that “have not inherited all the legal immunities that were won for the old” were moving to center stage.

His conclusion was stark: “The five-century growth of an unabridged right of citizens to speak without controls may be endangered.”

Pool worried that shifting to new mediums of communication and information-sharing left a window — a period where the free expression implications of the new technological mediums were not obvious — for the state to establish a foothold of influence before the First Amendment could catch up.

Applying First Amendment principles to AI-generated expression is both a natural extension of those principles and a necessary step to prevent those principles’ weakening in other applications, including in areas like newspapers and video games.

His most important observation about these shifts was that they would occur without society realizing what was happening. Because these new technologies were understood as requiring technical oversight, “regulation seemed to be a technical necessity.” The emerging new centers of information-sharing would be brought under government control through what lawmakers and regulators believed to be technocratic, democratic, seemingly unobjectionable decisions. Only later, when those technologies were fully woven into our social fabric, would society realize the control over information it ceded to the government — perhaps to actors with malign intentions, like those Pool had carefully studied all those years earlier.

So what led Pool to his grave conclusion? Bitter experience.

Pool had long observed that regulators and courts often missed the expressive implications of new communications technologies by analogizing them to infrastructure rather than expressive tools, like the press. The telegraph, for example, was regulated like a kind of transportation network rather than a new vehicle for speech; its regulatory structure borrowed from the common-carrier model developed for railroads.

The sidelining of free expression was as much an outcome of social unawareness as government determination. Because early telegrams “carried so few words” at “such a high cost” and were primarily business-oriented, he said, censorship was not on anyone’s mind. In fact, “there were virtually no cases in which First Amendment rights were claimed or such precedents applied.”

Lawmakers likewise analogized the telephone to the telegraph and imposed a similar legal framework. Broadcasting followed the same pattern, ushering in nine decades (and counting) of direct government censorship of the content spread over the airwaves, showing the consequences of failing to recognize the full breadth of First Amendment protections for new expressive technologies.

And with the world of computers and the internet that was materializing at the time of Pool’s warning, heeding his message by extending First Amendment principles to those new technologies has been a long and ongoing fight.

Now, as legislative attention turns towards the next evolution in expressive technology, artificial intelligence, Pool’s warning is of paramount importance.

A Look Forward

In this series, we’ll dive deep into the First Amendment questions raised by artificial intelligence. The idea that text seemingly spun from the machinations of an incomprehensible black box enjoys protection similar to the words out of your mouth can be bewildering. A federal judge in Florida noted as much in a case for which FIRE filed an amicus curiae brief, the judge struggling with the notion that “words strung together by an LLM are speech.”

We’ll argue the technology is less inscrutable than it may first appear, and that applying First Amendment principles to AI-generated expression is both a natural extension of those principles and a necessary step to prevent those principles’ weakening in other applications, including in areas like newspapers and video games.

FIRE to court: AI speech is still speech — and the First Amendment still applies

Is AI-generated speech speech? In a new amicus brief, FIRE says yes — and warns that when it comes to free speech and emerging tech, early missteps can echo for decades.

Read More

Opinions will differ on some points, like whose speech is implicated by restrictions on AI tools and where liability should be assigned. But, heeding Pool’s message, these debates must be grounded in the fact that AI systems are rapidly becoming a dominant way people seek information, ask questions, and develop ideas. From a First Amendment perspective, that is sacred terrain which we cannot afford to lose to government control.

The Supreme Court has made clear that technological change does not dilute that interest. As it reiterated in 2024 when talking about social media platforms, “whatever the challenges of applying the Constitution to ever-advancing technology, the basic principles of the First Amendment do not vary.”

Those words have a straightforward implication here. However novel AI-generated speech may seem, the First Amendment remains consistently concerned with state efforts to shape or control the flow of information and ideas, whatever form it takes. That’s our bedrock principle as we work through the particulars of this live issue.

It’s a principle that we believe is intuitive. Imagine if a Republican administration required AI outputs to align with conservative views on contested social issues, or if a Democratic administration required them to align with progressive views on those same issues. It would activate the same classically American aversion to censorship the respective sides would feel if the government was acting on a newspaper or a cable channel.

That instinct drives constitutional analysis, too. The Supreme Court has long refused to let labels muddle the essence in defining the scope of constitutional rights: “The Constitution deals with substance, not shadows.” That matters here. If the substance of an AI regulation controls what ideas an expressive technology may be used to generate, transmit, suppress, or prioritize, the First Amendment interest can not be overridden by framing AI as novel, technical, or as a mere ‘product.’ The medium may be novel. The constitutional problem is not.

Abandon that principle, and Pool’s warning becomes reality: a world where Americans will remain free to speak to their heart's content in yesterday’s forums, but when it comes to the forums of tomorrow, expression will exist only by government permission.

Tune in next week where we’ll explain the basics of how the First Amendment applies to AI. See you then!