Common Ground on Censorship: Two Bills to Check Government Coercion of Speech
Two bills recently introduced in Congress take direct aim at a growing apparatus of government censorship.
Both the JAWBONE Act and the Preventing AI Censorship Act would give Americans a way to push back when federal officials attempt to restrict free speech. The Biden administration pressed social media platforms to suppress pandemic related content, and the current Trump administration has publicly threatened broadcasters over the content of late-night programming. Each side has watched the machinery of informal coercion turn against speech it values, which is why a durable fix has to operate the same way no matter who holds power.
Together these bills offer the clearest sign in years that momentum may be shifting toward the First Amendment. The conduct they target is jawboning, which is the practice of pressuring private intermediaries into suppressing lawful expression. This is often done by government officials with the goal of shaping what most Americans can see and say without ever passing or defending a law in court.
The First Amendment forbids direct censorship, so policymakers instead use indirect channels and methods: a raised eyebrow from a regulator, a pointed letter, or a quiet threat to initiate legal proceedings. While there are some reforms that could help curb this conduct, the deeper problem is that the courts have left targets of jawboning with almost no recourse. Damages actions against federal officials have narrowed to a vanishing point, injunctions evaporate when an official moves on, and the Supreme Court’s recent decisions in cases like Murthy vs. Missouri have made the underlying coercion nearly impossible to prove.
The JAWBONE Act, introduced by Sen. Cruz (R-Texas) and Sen. Wyden (D-Ore.) sets out to give citizens a way to fight back against government coercion. If enacted, it would give citizens the ability to take legal action against jawboning by any agency, officer, or employee who coerces or attempts to coerce a broadcaster, online service, or artificial intelligence (AI) provider into acting against protected speech. It allows a jawboning claim to proceed regardless of whether the coercion succeeded, which answers the causation problem that has sunk so many plaintiffs. It codifies jawboning and makes a designation between permissible persuasion and impermissible threat, carves out legitimate law enforcement, and bars punitive damages so that the remedy stays tethered to actual injury. It also builds a transparency regime around the litigation in the form of a public portal logging government communications with companies the government wishes to speak to, inspector general audits, and a complaint channel for providers who believe they have been pressured.
The bill arrives with backing from a broad ideological and bipartisan coalition, indicating that both sides recognize that the threat changes hands with every election.
The Preventing AI Censorship Act, applies that same logic to the technology that will increasingly mediate public debate. It creates a right of action, which is a legal standing to sue, against federal employees. There is now standing to sue if public officials coerce, induce, or encourage an AI provider to suppress or distort its outputs on the basis of viewpoint. They can also sue if the government attempts to alter a model’s training data or system instructions toward the same end, to degrade a user’s access because of their views, or to harvest users’ prompts for surveillance.
An AI system that has been tuned to disfavor certain views, or whose provider has been pressured into doing so, censors more efficiently and far less visibly than any human content moderator. That danger does not belong to one party. An apparatus capable of shaping what models will say is a loaded weapon that every future administration inherits, and the side that builds it will not always be the side that wields it. For that reason, this bill deserves what it does not yet have, which is cosponsors from both parties. It rests on the identical principle that a coalition spanning left and right has already embraced in the broadcasting and social media context, and that principle does not weaken because the chokepoint is a model rather than a feed.
The AI bill would also be stronger, and more likely to attract that bipartisan support, if it borrowed some features from the JAWBONE Act. JAWBONE pairs its cause of action with disclosure. The AI bill relies on litigation alone. Adding a communications portal and a logging requirement of the kind JAWBONE establishes would surface improper pressure without anyone having to file suit, and it would give the bill the structural, forward-looking safeguard that civil liberties organizations have long preferred to after-the-fact damages. Transparency is a good mechanism to give the government a chance to express legal and important speech but deter coercion before it ever distorts what a model will say.
None of this means the bills are beyond improvement, and the enforcement designs in particular deserve a measure of caution. Both rely on private lawsuits, and private rights of action can be abused; however, neither of these is built to produce a windfall for attorneys. JAWBONE forbids punitive damages and limits recovery to actual harm and reasonable fees, and the AI bill offers no statutory bounty either, only compensatory relief and attorney fees at the court’s discretion.
The sharper concern lies in a single provision of JAWBONE that allows a state attorney general to sue on behalf of an entire state’s residents. Aggregating thousands of claims in the hands of officials who are elected, and increasingly partisan, risks turning a neutral accountability tool into an instrument of political combat, with each party’s attorneys general training the device on the other party’s appointees. That provision, which has no counterpart in the AI bill, is the one most in need of narrowing. Politicized enforcement would squander the very consensus that gives these reforms their authority. A tool that comes to be seen as a partisan weapon forfeits the cross-ideological credibility that is its greatest asset.
For years, the debate over online speech has been a proxy war, each side convinced that government pressure runs in only one direction—against them. Both of these bills reflect a more honest reckoning, that coercion of the companies carrying American expression is a standing temptation no administration has resisted and none should be trusted to resist on its own. The opening to act on that insight will not stay open long.