This is the third article in a weekly series on AI and Free Speech. You can read the first article explaining why the First Amendment is so important in the age of AI here.
In February 2024, FIRE President and CEO Greg Lukianoff took a short walk from our D.C. office to Capitol Hill to testify before the House Judiciary Committee. He delivered an important message: “the most chilling threat that the government poses in the context of emerging AI is government overreach that limits its potential as a tool for contributing to human knowledge.” A legislative panic, he warned, “could result in a small number of Americans deciding for everyone else what speech, ideas, and even questions are permitted.”
His warning to lawmakers marked an early milestone in FIRE’s growing engagement with AI as a free speech issue — work that has since taken us before legislators and state officials, and led us to fund grants and conduct proprietary research. But before we get into that work, it’s worth asking: How did a campus free speech organization end up testifying about artificial intelligence?
It started with our principles. We believe knowledge grows through open inquiry, debate, and the free exchange of ideas — and for decades, FIRE defended that principle on college campuses. Longtime fans know that from 1999 to 2022, our acronym stood for Foundation for Individual Rights in Education. But we always understood our principles extended beyond the university gates, and so, in 2022, we expanded to become the Foundation for Individual Rights and Expression, ready to defend free speech wherever it’s threatened, including in AI.
The expansion saw the rapid growth of our Legislative, Research, and Litigation Departments, and a robust engagement with tech policy issues soon followed. In 2022, we launched our first legal challenge to a social media regulation. Then, in 2023, we dropped our reports on social media and AI. The next year, with Greg’s testimony to Congress, we dove headfirst into free speech’s next frontier.
Our state legislative efforts
The 2024 to 2026 stretch has seen literally thousands of AI-related bills introduced in state legislatures across the country. That includes 1,208 bills in 2025 alone — not even counting the bills that have been introduced at the federal level.
Thankfully, most of these bills don’t directly implicate First Amendment interests, but there’s been more than enough to keep FIRE’s Legislative team busy as we faced exactly the wave of activity Greg’s testimony had feared.
We have flown to state capitols, worked with legislators, and fought efforts to let the government decide what AI systems can say and how Americans can use them.
We’ll take Washington state as a snapshot to give you a sense of just how many AI-related bills we’ll see march through a state simultaneously, and the First Amendment threats which often get lost amid the rush to legislate.
In January of 2026, FIRE delivered testimony to Washington’s state legislature on three different bills advancing through committee at once.
HB 1170 belonged to a broad category of bills we’ve seen in other states regulating AI-generated images and videos. The bill mandated that developers embed AI-generated content with government-prescribed disclosures.
We’d seen it before. It was FIRE’s second time in front of the Washington legislature on the bill, and that is only the tip of the iceberg in terms of the nationwide landscape.
In fact, FIRE has intervened to oppose over 20 bills targeting AI-generated content or “synthetic media,” including proposals in states like Virginia, Texas, Arkansas, North Dakota, Maryland, Rhode Island, Vermont, Connecticut, North Carolina, and Missouri. Those bills often aim at AI-generated political content, including “deepfakes,” and many sweep far beyond fraud or defamation and burden satire, parody, commentary, journalism, and simple criticism of candidates.
Next, Washington’s HB 2157, an “algorithmic discrimination” bill, would have made AI developers liable if their models were used by others to engage in discrimination. Such duties incentivize AI developers to handicap their models to avoid any possibility of offering recommendations that some might deem discriminatory or simply offensive.
Our fears stemmed directly from our higher education roots: “FIRE has spent decades defending students and faculty from vague and overbroad anti-discrimination standards that cause institutions to suppress speech rather than risk punishment.” Overlaying that kind of legal regime on AI models, rather than relying on existing law, threatens their capability as an information tool. We’ve kept an eye on similar bills, including a bill in Texas the previous year. We warned lawmakers about its free speech implications as it marched to the governor’s desk.
The last bill in Washington was HB 2225, which belongs to a broad group of AI chatbot-focused regulations that attack the First Amendment on multiple fronts. They restrict how chatbots like ChatGPT can respond to certain prompts and impose or encourage burdens on access, like age verification. It’s one of the fastest growing categories of AI restrictions at the time of this writing, and we’ve even had to fight it at the federal level with the GUARD Act.
Now, those three examples are only the Washington bills from 2026 that we had time to engage on. Among the 37 AI bills introduced in the Washington legislature that session, several more had clear First Amendment implications. With this flurry of activity, we met with Washington’s AI Task Force after the previous legislative session concluded to give a presentation laying out how the First Amendment applies to AI.
Again, this is just one state. We barely scratched the surface of the action in the other 49.
Federal legislation
While states had been the primary battleground, activity has now picked up at the federal level too. We mentioned the GUARD Act, but chatbot regulation is also advancing at the federal level in the CHATBOT Act and the KIDS Act. We’re also fighting the NO FAKES Act, a bill that would expose people to lawsuits for speech involving a digital version of somebody else’s likeness without their permission. This would give politicians and other public figures new leverage over how they’re portrayed in today’s media.
Moves from Congress have been paired with aggressive action from the executive branch, where FIRE has opposed attempts to create an arbitrary licensing regime over advanced “frontier” AI models that would give the White House unrestrained power over them and open the door to ideological censorship.
Building the case
Our interventions have all been grounded in a stalwart opposition to the government substituting its own beliefs for those of developers, and a fear of grave consequences if government censorship is allowed to gain a foothold just because the technology is new.
That gets at a point we’ve made again and again: The government cannot restrict lawful expression just because a person used a new tool to create or share it.
As for addressing the real harms that may result from the use of AI, we also make clear that new tools are still covered by the older laws. Forgery, defamation, discrimination, and election-related crimes are already addressed under existing law, whether AI is involved or not. Fraud is still fraud whether it is committed with a pen, a keyboard, or a chatbot. Thankfully, some policymakers recognize that point.
The same themes animating our legislative efforts have also informed our litigation work. Because AI-generated expression is still expression, FIRE has filed amicus briefs in early cases where courts or government officials have treated AI-assisted speech as if it falls outside ordinary First Amendment protection. In 2025, we filed an amicus “friend of the court” brief urging appellate review of a federal court decision that voiced uncertainty about whether “words strung together by an LLM” could be speech. It was one of the first courts to weigh those questions, and these early cases can cast a long shadow in determining how the First Amendment operates in new terrain.
Our early amicus efforts have also included a brief filed in California concerning a bill mandating the removal of “deceptively” edited digital media and deepfakes of political figures. As we wrote, this violates the First Amendment’s “core function of protecting democratic self-governance by enabling free and unfettered discussion of political affairs.”
When the Pentagon designated Anthropic a “supply-chain risk” for insisting on keeping limits on the permitted uses of Claude, Anthropic brought a legal challenge and we soon filed an amicus brief in support of it. We criticized the government for “openly and shamelessly retaliating against Anthropic for [its] protected speech.” We further explained our concerns with the government’s use of AI for domestic surveillance and the grave risk of capabilities that could be exploited to monitor public discourse or preemptively squelch dissent.
The world of AI litigation remains in an early stage, but we’re actively reviewing new laws, identifying potential cases, and building capacity to bring legal challenges to help develop the First Amendment doctrines that will protect expressive freedoms in the age of AI.
Research
Our mission has taken us beyond the legislative and legal landscape. We’ve also been studying how Americans think about AI, free speech, and regulation so advocates and lawmakers have a better sense of the public’s mindset.
In January 2026, FIRE included AI for the first time in our National Speech Index — our continuously updated survey of Americans’ free speech attitudes. The results were clear: Americans are worried about AI, but they are also worried about the government using AI regulation as a backdoor for censorship. Seventy-two percent said they were at least somewhat concerned about everyday AI use — but the same percentage said they were concerned about the government regulating human expression that involves AI. Most strikingly, 92% said it is at least somewhat important for governments to protect free speech when regulating AI.
That polling confirms our previous findings: Americans reject the false choice between addressing harms and protecting free speech.
Funding new AI models
Protecting free speech in the AI age will take more than stopping bad laws. It will also take actively building the future where AI is a tool for finding truth and expanding access to knowledge rather than limiting or restricting it. In May 2025, FIRE and the Cosmos Institute launched the Truth-Seeking AI Grants Program, a $1 million initiative to support open-source projects that build freedom, inquiry, and viewpoint diversity into AI systems.
The first and second cohorts included 51 builders working on projects like Steven Molotnikov’s Priori, which changes the usual prompt-and-response experience by revealing the assumptions behind an AI’s answer, allowing users to modify those assumptions before generating a new response. Rather than treating AI as a black box, this innovative tool gives users insight into — and control over — how the model reasons. Other tools include an “Argument Debugger” that finds gaps in reasoning, tools that compare AI model outputs for signs of censorship, systems that surface competing perspectives, and AI thought partners that push users to ask better questions.
This may sound different from FIRE’s traditional work, but it grows out of the same mission. If AI tools are going to shape what we read, write, ask, and know, then FIRE wants to work to both protect those tools from government interference and build examples that demonstrate a positive alternative vision of AI to one of centralized power and control.
As Philipp Koralus, senior research fellow at the Cosmos Institute, explained, “AI can already steer our thoughts. The future is AI that expands them, not controls them.” The grants program is FIRE and the Cosmos Institute’s bet on those who expand them: “We’re giving them the capital, computing resources, and community they need to seize that opportunity.”
Whether the venue is a state legislature, Congress, a courtroom, or even in the AI labs themselves, FIRE’s aim is to ensure AI technology widens the universe of what people can know and share, rather than empowering the government and a small number of corporations to control that universe to fit their own ends.
 | A guest post by
|