Federal courts have issued a run of significant First Amendment decisions since 2021, and the recent freedom of speech cases below cover the questions people actually run into: whether the government can ban a foreign-owned app, when a public official can block you online, what happens when officials privately pressure platforms, whether states can force platforms to keep posts up, when a creative professional can refuse a job, what counts as a criminal threat, how much religious expression a public employee gets, what schools can punish off campus, and how AI-generated political content fits into any of this. Each ruling draws a sharper line around what the government can and cannot do when speech is in play.
TikTok, ByteDance, and National Security
In January 2025, the Supreme Court unanimously upheld the federal law requiring ByteDance to divest TikTok or face a U.S. ban. TikTok Inc. v. Garland treated the Protecting Americans from Foreign Adversary Controlled Applications Act as content-neutral because it targeted a foreign adversary’s ability to harvest data from 170 million American users, not anything anyone was posting.1Supreme Court of the United States. TikTok Inc. v. Garland
Applying intermediate scrutiny, the Court found the law sufficiently tailored. A qualified divestiture would let TikTok keep operating once the data-collection risk was eliminated, and nothing in the statute restricts what anyone says on the platform or singles out a viewpoint. The Court gave Congress substantial deference on the national security judgment, part of which rested on classified evidence.1Supreme Court of the United States. TikTok Inc. v. Garland
For users and creators, the practical lesson is that foreign ownership of a platform creates vulnerabilities the government can address, even when doing so disrupts a major speech channel. The incidental burden on speech was not enough to defeat a law whose justification had nothing to do with suppressing a message.
When a Public Official Blocks You on Social Media
A unanimous Supreme Court set the test for this question in March 2024. Lindke v. Freed holds that a public official’s social media activity counts as state action only if the official had actual legal authority to speak for the government on the topic in question and was using that authority in the post or interaction at issue.2Supreme Court of the United States. Lindke v. Freed
Both parts matter. A city manager posting about road closures from an account that references the official title and links to government resources is likely acting in that official capacity. The same person posting vacation photos from a personal account is not, even though everyone knows the office. The question is whether a reasonable viewer would see the post as government communication rather than personal expression.
The companion case, O’Connor-Ratcliff v. Garnier, involved school board trustees who blocked parents from accounts used to discuss board business. The Supreme Court sent it back to the Ninth Circuit to apply Lindke, and the Ninth Circuit did so in 2025.3United States Court of Appeals for the Ninth Circuit. Garnier v. O’Connor-Ratcliff The framework prevents officials from using an “it’s my personal page” defense on accounts that function as government channels, while leaving genuinely private expression alone.
Government Pressure on Platforms Behind the Scenes
Two 2024 decisions addressed the subtler problem of officials privately leaning on social media companies to take content down. The legal term is “jawboning,” and the Court looked at it from opposite directions in the same term.
Murthy v. Missouri
In Murthy v. Missouri, the Court reversed a sweeping injunction that would have barred federal officials from communicating with platforms about content moderation. The ruling never reached the merits. The plaintiffs lacked standing because they could not show that any specific government communication caused a specific platform to remove any specific post of theirs.4Supreme Court of the United States. Murthy v. Missouri
The standing problem was fundamental. Platforms had already begun removing COVID-19 misinformation before the government contacts at issue, and they had independent reasons to moderate that content. A plaintiff asking a court to stop the government from pressuring a third party has to show a direct chain: a particular official pressured a particular platform about a particular topic, and that pressure caused the platform to suppress that plaintiff’s speech. Speculation about influence was not enough.4Supreme Court of the United States. Murthy v. Missouri
The practical effect is that jawboning claims remain hard to bring. Even if officials are leaning on platforms, proving the causal link between a staffer’s phone call and a takedown is a high bar.
NRA v. Vullo
Where Murthy sidestepped the merits, NRA v. Vullo confronted them. The NRA alleged that a New York financial regulator pressured banks and insurance companies to cut ties with the organization because of its gun-rights advocacy. The Court unanimously held that the NRA stated a valid First Amendment claim.5Supreme Court of the United States. National Rifle Association of America v. Vullo
Applying the framework from Bantam Books v. Sullivan, the Court explained that officials cannot coerce private parties into suppressing disfavored speech. A plaintiff must plausibly allege that the official’s conduct, read in context, could reasonably be understood as a threat of adverse government action meant to punish or suppress speech. Relevant factors include the official’s regulatory authority over the target, the tone and wording of the communications, whether the recipient perceived a threat, and whether the official referenced consequences for noncompliance.5Supreme Court of the United States. National Rifle Association of America v. Vullo
Read together, the two cases create an asymmetry. Proving a jawboning violation requires both a coercive threat under Vullo and proof under Murthy that the threat actually caused suppression. The first is about what the government did. The second is about what the platform would have done anyway.
Can States Force Platforms to Keep Content Up?
Texas and Florida passed laws in 2021 trying to stop large social media companies from removing political content. Both reflected a view that platforms were censoring conservative speech. In Moody v. NetChoice and NetChoice v. Paxton, decided together in 2024, the Supreme Court vacated the lower court rulings on both laws and sent them back for a more rigorous analysis.6Supreme Court of the United States. Moody v. NetChoice, LLC
The reasoning sent a strong signal. The majority recognized that when platforms curate feeds by choosing what to promote, label, or remove, they engage in editorial activity protected by the First Amendment. The analogy was to a newspaper choosing which op-eds to print: the government cannot force a private publisher to carry speech it would rather exclude.6Supreme Court of the United States. Moody v. NetChoice, LLC
The Court was skeptical of treating platforms as common carriers like phone companies or railroads. Social media companies build curated experiences through active moderation, which distinguishes them from utilities that passively transmit whatever users send. The Court did acknowledge that its analysis might not apply uniformly to every platform function; a direct messaging service might look more like a common carrier than an algorithmically sorted news feed. The lower courts must now do a proper facial challenge analysis, examining the full scope of the laws and sorting constitutional applications from unconstitutional ones. Those proceedings are ongoing.
Compelled Speech and Creative Professionals
In 303 Creative LLC v. Elenis (2023), the Court ruled 6-3 that Colorado could not use its anti-discrimination law to force a website designer to create custom wedding websites celebrating same-sex marriages when doing so conflicted with her beliefs.7Supreme Court of the United States. 303 Creative LLC v. Elenis
The distinction the majority drew is between refusing to serve a person because of who they are and declining to create a particular message. A bakery that refuses to sell any cake to a gay couple violates public accommodation laws. A custom artist who will serve anyone but will not write a specific message is engaged in protected speech. The websites in the case qualified as pure speech because they involved original, customized content created through close collaboration with the couple, where the designer’s creative choices shaped the overall message.7Supreme Court of the United States. 303 Creative LLC v. Elenis
The ruling does not reach every business interaction. It protects services that involve genuine artistic or intellectual expression (custom design, writing, illustration) rather than off-the-shelf retail. A photographer composing original work for each client has a stronger claim than a printer running standard business cards. Justice Sotomayor’s dissent warned that the logic could extend to businesses refusing service to interracial couples, disabled parents, or any group a business owner disfavors, so long as the service is framed as “expressive.”7Supreme Court of the United States. 303 Creative LLC v. Elenis Where courts draw the line between protected creative refusal and unlawful discrimination will depend on how broadly future cases define “expressive” services.
What Counts as a True Threat
Before 2023, many states could convict someone of criminal threats based solely on how a reasonable listener would read the words, regardless of whether the speaker intended or realized the message sounded threatening. Counterman v. Colorado changed that. In a 7-2 decision, the Court held that the First Amendment requires prosecutors to prove at least recklessness, meaning the speaker consciously disregarded a substantial risk that the words would be perceived as threatening violence.8Supreme Court of the United States. Counterman v. Colorado
The case involved a man who sent hundreds of Facebook messages to a local musician over two years. Many were disturbing, but Counterman argued he had not intended them as threats. Colorado’s law required only that he intended to send the messages, not that he understood their threatening nature. The Court found that insufficient.8Supreme Court of the United States. Counterman v. Colorado
Recklessness sits in the middle of the mental-state spectrum. The prosecution does not have to prove the speaker specifically intended to frighten anyone, but it does have to show the speaker was aware the words could reasonably be taken as a threat and sent them anyway. The Court picked this standard to avoid chilling protected speech while still allowing prosecution of those who knowingly roll the dice with threatening language. The ruling specifically applies to speech-based stalking conduct, not to physical following, surveillance, or other non-expressive stalking behaviors.
Public Employee Religious Expression
In Kennedy v. Bremerton School District (2022), the Court ruled 6-3 that a public school football coach had a constitutional right to pray quietly on the field after games. The school district had suspended Kennedy for refusing to stop, arguing the practice could look like government endorsement of religion. The Court disagreed, finding the prayers were private speech, not government speech, because Kennedy was not performing a coaching duty at the time and was not directing students to join him.9Supreme Court of the United States. Kennedy v. Bremerton School District
The decision also reshaped how courts evaluate Establishment Clause claims. For nearly 50 years, courts applied the Lemon test, a three-part framework from a 1971 case. The Kennedy majority declared that the Court had “long ago abandoned” that approach and replaced it with a standard rooted in historical practices and understandings of the First Amendment.10Congressional Research Service. Kennedy v. Bremerton School District: School Prayer and the Establishment Clause
The shift matters beyond football fields. Under Lemon, courts often found Establishment Clause problems when a government employee’s religious expression was visible to others. Under the new standard, the focus moves to whether the government is actually coercing religious participation. A teacher wearing a cross, a city council member praying before a meeting, a firefighter keeping a prayer book at the station — all become harder to discipline unless the employee is pressuring others to participate.
Student Speech Off Campus
Students don’t shed their speech rights at the schoolhouse gate, but they don’t get unlimited rights off campus either. In Mahanoy Area School District v. B.L. (2021), the Supreme Court held that a school could not punish a student for a vulgar Snapchat post made off campus on a weekend. The Court identified three features of off-campus speech that limit a school’s authority to regulate it.11Supreme Court of the United States. Mahanoy Area School District v. B.L.
First, schools rarely stand in the place of parents when a student is away from school. Second, letting schools regulate off-campus speech on top of on-campus speech would put a student’s expression under institutional control around the clock. Third, schools themselves have an interest in protecting unpopular student expression, because public schools function as “nurseries of democracy.” The Court made clear that schools may still act when off-campus speech involves serious bullying, threats aimed at students or staff, or genuine disruption of classroom activities, but they carry a heavy burden of justification.11Supreme Court of the United States. Mahanoy Area School District v. B.L.
Courts continue to refine the framework. In Leroy v. Livingston Manor Central School District (2025), the Second Circuit ruled that schools cannot discipline students for off-campus speech just because it is offensive, though threatening speech may be a different matter. The court declined to set a bright-line rule, requiring instead a fact-specific analysis of the nature of the speech, where and when the student spoke, and the risk that it would reach and disrupt the school. The court recognized that social media posts travel more readily than a private Snapchat story, but that alone does not give schools unlimited authority to punish what they find objectionable.
AI-Generated Content and Election Deepfakes
The intersection of artificial intelligence and the First Amendment is the newest and least settled area here. As of 2026, twenty-nine states have enacted laws regulating AI-generated deepfakes in political campaigns, using two main approaches: outright prohibitions on deceptive synthetic media near elections and mandatory disclosure labels on AI-generated political content.
Early court challenges suggest these laws face serious constitutional headwinds. In August 2025, a federal judge struck down key provisions of California’s “Defending Democracy from Deepfake Deception Act,” which had required platforms to block or label AI-generated political content in the 120 days before an election. The court found the labeling requirements overly broad, the standard for what constitutes harm to a candidate’s “electoral prospects” unconstitutionally vague, and the private right of action allowing nearly anyone to sue far too expansive. A Hawaii deepfake law was struck down on similar grounds.
At the federal level, the White House released a “National Policy Framework for Artificial Intelligence” in March 2026 recommending that Congress consider protections against unauthorized AI-generated replicas of a person’s voice or likeness. The framework explicitly called for exceptions for parody, satire, news reporting, and other First Amendment-protected expression, an acknowledgment that any federal regulation of synthetic speech will have to navigate the same constitutional constraints that tripped up California’s law.
No court has definitively answered whether AI-generated content itself gets First Amendment protection. If a person uses AI as a tool to create political satire, that output likely carries the same protections as satire created with Photoshop. A fully autonomous system generating and distributing deceptive political content without meaningful human direction raises harder questions courts have not yet reached. Any regulation enacted today faces a real chance of being narrowed or struck down before the next election cycle.