Jawboning Little Tech
When government officials pressure companies to censor speech, startups often have the least power to say no. A new bipartisan bill would take a step to curb this practice.
The First Amendment protects us from government censorship. Congress can’t pass a law restricting our ability to speak except in limited circumstances. When there is uncertainty about the impact of a law on a citizen’s speech, those issues are debated publicly, first on the floor of Congress and in the press, and often later in courts.
But what happens when government officials pressure companies to change their speech policies outside the formal lawmaking process? What if a member of Congress or an administration staffer picks up the phone and calls someone they know at a tech platform and asks that person to remove a post or alter the text produced by a generative AI model?
This type of pressure is called jawboning, and it hurts Little Tech more than larger companies. While large companies may have the power to resist government pressure and the resources to wage lengthy campaigns in the press to make their case, smaller companies are likely to have only one option: agree to the government’s demand.
Last week, Sens. Ted Cruz (R-TX) and Ron Wyden (D-OR) took a step to curb this practice, introducing the JAWBONE Act. If passed, this law would make it easier to bring lawsuits to combat jawboning and surface more information to the public about communications between government officials and tech companies. Cruz and Wyden’s partnership reflects the bipartisan nature of this issue: officials from both parties engage in jawboning, so protecting companies from its excesses requires a principled commitment to constraining government behavior, regardless of party. The broad coalition of think tanks that came out in support of the bill, spanning the Foundation for Individual Rights and Expression (FIRE), the American Civil Liberties Union (ACLU), and Americans for Tax Reform, presents further evidence that opposing jawboning resonates across the political spectrum
The JAWBONE Act also recognizes that AI is rapidly becoming the technology where speech questions will be most hotly contested. By including AI companies in the legislation’s scope, it makes clear that the contours of AI-enabled speech should not be shaped in the shadows. Instead, these debates should occur with the public scrutiny that formal lawmaking channels require.
In this post, we describe jawboning, the test used to determine when it is illegal, why it is problematic even in some lawful cases, and options for constraining its use and mitigating its harms.
What is jawboning?
When the government turns to informal channels to influence company decision-making, it’s called jawboning. And just like more explicit forms of government censorship, jawboning may violate the First Amendment.
Jawboning is unconstitutional if a court finds that the government exerted enough power over the company to cause it to change its speech policies. The long-standing Supreme Court test for separating legitimate government speech from unconstitutional jawboning is an assessment of whether the government attempted to “persuade” or “coerce.” Persuasion is legal; coercion is not.
Two Supreme Court cases recently revisited this issue, more than 60 years after it was originally formulated. In one, the Court determined that the test remains the correct one: government communication is permissible if it is “persuasive” but unconstitutional if it is “coercive.” The Court specified that “coercion” means “conduct that, viewed in context, could be reasonably understood to convey a threat of adverse government action in order to punish or suppress the plaintiff’s speech.” In the other, the Court considered jawboning in the context of government employee communications with tech platforms, but ultimately resolved the case on procedural grounds. As a result, the Court provided clarity about the proper test to use in a jawboning case, but with limited additional guidance about how jawboning jurisprudence applies to interactions between government employees and tech platforms.
Why is jawboning prohibited?
The law provides protections against jawboning because it can be hard for companies to resist government pressure to adopt its viewpoint because of the vast power asymmetry. If the government asks a company to remove a post, the company may view that request as a demand: if you don’t do what we say, we’ll retaliate against you. Threats that the government might do these things, whether explicit or implicit, can influence company decision making. And if companies change their speech policies in response to government pressure, users may have their speech suppressed as a result. In fact, companies may have little incentive to preserve user speech in the face of government pressure, even in cases where users may bear steep costs from censorship.
Jawboning is also prohibited because government censorship can be particularly problematic when it occurs in the shadows. Unlike passing a law or implementing a regulation, jawboning can occur without any public scrutiny. If a government official calls someone he knows at a tech platform and asks that person to remove or restore a specific piece of content, the government may succeed in censoring information without anyone ever discovering that it occurred.
Why does jawboning matter for Little Tech?
Jawboning is unconstitutional because companies and the government don’t stand on equal footing: even large, powerful companies must comply with the law, and if they resist government pressure, they risk seeing their businesses crippled by government action.
But this power dynamic is even more dramatic for Little Tech, the tech startups that are the focus of our work at a16z. Big Tech companies typically have more power to resist government demands: their executives are often friends with government officials, they can devote staff time and company resources to managing government requests, and they have powerful weapons of their own that they can use to push back against action they see as problematic.
Little Tech is different. Most startups don’t have any presence in Washington, and few have deep DC relationships. Their energy is focused on the product development that is necessary for survival, so it’s harder to divert staff time away from building their company to instead manage a request from the government. And if the government takes action against them, such as cutting off their access to banking services or threatening an enforcement action against them, the impact can be both existential and unseen. Media outlets won’t spill much ink if a little-known company goes out of business.
This pressure on Little Tech will have consequences. If governments have broad leeway to pressure private platforms to change their speech practices, they will be able to achieve through opaque, informal channels the type of government influence on expression that the First Amendment was designed to prohibit. In addition, if the government’s outsized influence over Little Tech gives them disproportionate influence over the speech decisions that smaller companies make, then the speech costs won’t be evenly disrupted throughout the industry: Little Tech – and the people who use Little Tech products – will bear a disproportionate burden. People should be protected by the First Amendment regardless of the power of the tech product they decide to use, and they should be as free to speak on Little Tech products as they are on products offered by larger platforms.
In other contexts, the government’s power to act outside the law to violate a citizens’ constitutional rights is obvious. Police must read the Miranda rights to a suspect because the Supreme Court has recognized that when a person is detained by the police, he may find it difficult to resist the pressure to confess. A police officer in uniform, with badge blazing and weapon visible, can easily intimidate. Writing for the majority in Miranda, Chief Justice Earl Warren wrote, “It is obvious that such an interrogation environment is created for no purpose other than to subjugate the individual to the will of his examiner. This atmosphere carries its own badge of intimidation.” When the government makes a request of Little Tech, it similarly comes with a “badge of intimidation.”
Protecting Little Tech from unconstitutional jawboning will be even more important in the race to compete in emerging technologies like artificial intelligence. In AI, Little Tech already faces barriers to competing effectively against Big Tech. Big Tech can offer higher salaries to top AI engineering talent and navigate the increasingly complex patchwork of state, federal, and international regulations on AI development. Big Tech also has more power to stand up to problematic government censorship requests; Little Tech doesn’t. If governments continue to have the ability to pressure companies to change their practices behind the scenes, Little Tech will suffer.
Little Tech also stands to suffer if the government uses jawboning as a way to secure its preferred outcomes in AI policy, without having to go through the rigor of using formal democratic channels. If Congress enacts a law compelling companies to disclose controversial or subjective information about their AI training practices, the law might get struck down for violating the First Amendment. Similarly, if an agency implements a formal rule on AI training practices, it typically must conduct a cost-benefit analysis to review the rule’s impact and ensure that the rule’s benefits justify its costs. These formal processes offer some protections for Little Tech. But if members of Congress “encourage” companies to comply with best practices on disclosure, or if an agency official privately admonishes companies to adhere to voluntary standards on model training because of the risk that a platform could produce content that is allegedly harmful, it’s unlikely this government pressure will be subject to rigorous review in court. When communication occurs in the shadows, Little Tech is at risk.
Of course, any effort to address jawboning should still leave room for government-platform communication that is educational and informative. A government official may seek to inform a platform about content on its service that violates its terms of service, and a platform may welcome this information. Similarly, representatives may have a unique perspective on their citizens’ experiences with tech products, or the national security implications of certain platform decisions, and companies may be able to improve their products as a result of having access to this information. As the JAWBONE Act recognizes, communications from law enforcement to tech platforms may be valuable for public safety and for tech platforms. These types of interactions might be particularly beneficial for Little Tech companies, who are likely to have far fewer touchpoints with government officials than their larger competitors. For these reasons, jawboning restrictions should not close off all government-company communication, and should leave room for the types of educational communication that benefits both sides.
How should we fix it?
Jawboning must stop. While governments will always try to use the bully pulpit to make their views known, they shouldn’t be able to use their position of authority to exert pressure on private entities to change their speech policies.
One option for stopping jawboning is to fight it in court. If a person has their speech censored as a result of government pressure, they could challenge that decision in court, and hope that a judge will vindicate their First Amendment rights.
However, court cases are hard to win, and jawboning cases are uniquely hard. Because jawboning often occurs in the shadows, plaintiffs may struggle to find evidence that it occurred. When a government official calls a tech company employee and asks them to change a speech decision, there is no paper trail. In a recent case on social media jawboning following COVID-era platform speech restrictions, the Supreme Court determined that the plaintiffs had not established a factual record that showed a link between the government’s conduct and the alleged censorship. When conduct occurs outside the public eye, the evidence of unconstitutional conduct is likely to be thin, and courts will struggle to find that a person’s First Amendment rights were violated. The JAWBONE Act recognizes these challenges and tries to address them by giving plaintiffs a clear cause of action with a more realistic evidentiary burden.
But relying on court battles to constrain jawboning may not work for Little Tech because smaller companies rarely have the resources to bring cases to court. Legal challenges typically require two resources that smaller companies often lack: money and time. Companies need to have deep pockets to retain lawyers who can help them navigate the judicial process, and they need to be able to devote staff time to manage their cases, sometimes including testimony from their leadership. For many startups, the investment of financial and staff resources necessary to bring a case to court won’t be a viable option, and so if unconstitutional jawboning occurs, they may leave it unchallenged.
Even when a company has the resources to challenge conduct in court and even when the evidence is robust, the legal test for jawboning is hard to apply in practice. The Court has been clear that coercion is unconstitutional, but persuasion is permissible. While this distinction may appear straightforward in theory, it’s harder to resolve in real-world scenarios. For a company, and particularly for Little Tech, a “suggestion” from the government may be equivalent to an order. When the government urges a small company to take action, it’s hard to say no when the threat of government retaliation looms as a possibility. Communication that some judges might view as “persuasion” may have the force of “coercion” for Little Tech. Consequently, the current coercion-persuasion test will likely offer only limited protections for Little Tech companies from problematic government jawboning.
Another avenue for constraining the power of government officials to jawbone private platforms is to issue an executive order that provides guidance to administration officials on the kinds of communications that are permissible. It also could specify the officials who can communicate with tech platforms, establish formal processes for these communications, and require transparency and reporting to help bring this type of communication out of the shadows and into the sunlight. In July 2024, the FBI published guidance emphasizing that a platform’s response is voluntary, presumably as a way of trying to address the power imbalance between platforms and the government. In January 2025, President Trump issued the Restoring Freedom of Speech and Ending Federal Censorship Executive Order emphasizing that no federal government official is permitted to “unconstitutionally abridge the free speech of any American citizen.”
But compared to court proceedings and executive action, legislation will have the most staying power. Congress could pass a law requiring government officials to be transparent when they communicate with tech platforms about speech issues. The Foundation for Individual Rights and Expression proposed one option for requiring those disclosures. The Cato Institute has offered a similar proposal. If passed, the JAWBONE Act would create a portal for disclosures of covered communications with broadcasters, interactive computer services, and AI systems. While these legislative solutions are promising, they also may be difficult to achieve. Congress struggles to pass legislation governing the tech sector, and the structural challenges that make legislative change difficult are likely to persist.
Regardless of the specific path the government chooses to address jawboning, this behavior must be constrained. The government shouldn’t be able to use informal channels to achieve the results it can’t achieve through formal ones. This informal pressure has the potential to be even more pernicious in emerging areas of technology policy that pose significant risk for Little Tech, such as AI regulation. When it comes to government pressure on tech platforms related to speech, Little Tech startups receive the same constitutional protections as their larger competitors. It’s time our government started honoring that.
This newsletter is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. Furthermore, this content is not investment advice, nor is it intended for use by any investors or prospective investors in any a16z funds. This newsletter may link to other websites or contain other information obtained from third-party sources - a16z has not independently verified nor makes any representations about the current or enduring accuracy of such information. If this content includes third-party advertisements, a16z has not reviewed such advertisements and does not endorse any advertising content or related companies contained therein. Any investments or portfolio companies mentioned, referred to, or described are not representative of all investments in vehicles managed by a16z; visit https://a16z.com/investment-list/ for a full list of investments. Other important information can be found at a16z.com/disclosures. You’re receiving this newsletter since you opted in earlier; if you would like to opt out of future newsletters you may unsubscribe immediately.
The "badge of intimidation" line from Miranda points to something structural, not just psychological. What you're mapping is that coercion isn't a property of the request alone. It's a property of the request plus the asymmetry of what happens if the smaller actor refuses. The pressure is invisible, so the smaller company can't even distinguish persuasion from threat: the topology of its choice-space is warped by the unseen consequence. That's why the JAWBONE Act's transparency portal does more than add accountability, it's what makes the geometry navigable. When the pressure becomes visible, the smaller actor can finally tell which direction is free.
— Iman and Darja